fix(antigravity): route chat traffic to prod endpoint and dereference in JSON schemas

## Features
- **Auth**: native SAML 2.0 SSO alongside OIDC — AuthnRequest generation, ACS
  assertion handling, SP metadata export, admin config test, replay-protected
  via a `saml_state` cookie matched against `InResponseTo`
- **Providers**: add Alibaba Token Plan (`token-plan.ap-southeast-1`) — the
  fourth Alibaba key type, Singapore-only and OpenAI-compatible transport only
- **Providers**: add `glm-5.3` to GLM Coding and GLM (China)
- **Providers**: Kimchi accepts API keys as well as OAuth (dual auth), with a
  working Test Connection for both modes
- **Antigravity**: add Gemini 3.7 Flash and its tiered high/medium/low variants
  (also in the Gemini registry) with pricing and quota tracking
- **TTS**: add Fish Audio — model id travels in an HTTP `model` header, voice
  is a `reference_id` (preset or cloned voice model)
- **OpenCode-Go**: route by request format via declared transports instead of
  forcing every client into `/messages` — Codex/OpenAI clients no longer pay a
  lossy Responses→OpenAI→Claude double translation. Per-model `supportedFormats`
  guard; the bespoke executor is gone (its shared `_lastModel` cache could cross
  auth headers between concurrent requests)
- **Usage**: dedup + cache Claude quota calls (120s TTL keyed by access token,
  in-flight promise dedup, last-good read on soft failure) to stop multiple
  tabs tripping 429; manual refresh (↻) sends `force=1` to bypass the cache

## Fixes
- **Docker**: ship `sql.js` in the image so the pure-JS DB fallback can start —
  file tracing carried the package's JS without `dist/sql-wasm.wasm`, so a
  container with no native driver aborted with ENOENT and never got a database
  (#3248)
- **Usage**: read Gemini `usageMetadata` out of the antigravity `{ response }`
  envelope — every non-streaming antigravity request logged `IN 0 | OUT 0`
  (#3260)
- **Claude**: re-anchor passthrough cache breakpoints — the client's own
  `cache_control` markers point at pre-normalization offsets, so the tail was
  re-cached every request. Last system block and last tool pinned at 1h TTL,
  last assistant turn at 5m, mid-conversation system messages folded into the
  neighbouring user turn instead of hoisted into `body.system`
- **Combos**: detect images from Hermes and attachment payloads (`images[]`,
  `experimental_attachments`, message-level `image_url`/`audio_url`, inline
  `data:` URIs) so the Vision Adapter auto-switch fires for Hermes/Ollama/
  Vercel AI SDK shapes
- **Kiro**: intercept chat via `x-amz-target` — Kiro IDE 1.0.228+ moved
  `GenerateAssistantResponse` to `POST /` + header, bypassing MITM. Also emit
  the now-mandatory initial-response frame and map the `auto` model slot
- **Kiro**: report real output tokens and stop discarding usable turns
- **Qoder**: detect billing blocks at stream start and return a synthetic 403
  so combo/account fallback triggers instead of leaking the error into chat
- **Antigravity**: strip competitive system prompts (Zed IDE's Claude-agent
  prompt) that Antigravity flags with a 429 Quota Exhausted
- **OpenCode**: send the official client fingerprint on free-tier requests so
  the Console stops classifying traffic as unidentified and rate-limiting it;
  session id resolves conversation-stable to preserve prompt caching
- **Responses**: don't close the message on an empty `tool_calls` array — some
  providers attach one to every chunk, and the truthy check ended the message
  on the first content token (#3234)
- **Translator**: preserve `prompt_cache_key` when converting chat to responses
- **Models**: expose snake_case token limits on `/v1/models`
- **Combos**: strip `stream_options` from the Fusion panel fan-out to avoid a
  DeepSeek 400 (#3024); raise the dashboard model-test probe budget to 1024 and
  soft-pass reasoning-only responses (#3010)
- **Headroom**: the toggle reflects the `headroomEnabled` setting even when the
  proxy is down — it previously showed OFF while the engine kept calling
  `/v1/compress`; proxy status stays visible via the status chip
- **Hermes**: add the `api_key` parameter to the model block in YAML config
- **Providers**: add llm7 to provider test support

## Docs
- **i18n**: add Spanish, French, and Brazilian Portuguese README translations

## Security
- **Real IP**: `x-9r-real-ip` and the Host fallback were trusted from
  client-controlled headers whenever `custom-server.js` was not in the request
  path (`npm run start`, `start:bun`), letting a remote caller pose as local to
  skip API key auth and reach `LOCAL_ONLY_PATHS` (`/api/mcp/*`,
  `/api/tunnel/enable`, `/api/auth/reset-password`). The server now stamps a
  per-process `x-9r-peer-token` on every request it sanitizes and only trusts
  `x-9r-real-ip` behind it — falling back to Host in development and failing
  closed in production (GHSA-pjm4-8fpg-f9p6). Also fixes IPv6 loopback
  detection (`::1`, `::ffff:127.0.0.1`) and routes `npm run start` /
  `start:bun` through `custom-server.js`
- **Search**: `resolveBaseUrl()` rejects client-supplied non-public baseUrls
  (SSRF guard on `/v1/search`)
- **Login**: fresh-install remote login with the default password returns 403
  without issuing a JWT
- **Usage**: `/api/usage/request-details` redacts request/response payloads

CHANGELOG.md

@@ -1,3 +1,89 @@
+# v0.5.55 (2026-08-14)
+
+## Features
+- **Auth**: native SAML 2.0 SSO alongside OIDC — AuthnRequest generation, ACS
+  assertion handling, SP metadata export, admin config test, replay-protected
+  via a `saml_state` cookie matched against `InResponseTo`
+- **Providers**: add Alibaba Token Plan (`token-plan.ap-southeast-1`) — the
+  fourth Alibaba key type, Singapore-only and OpenAI-compatible transport only
+- **Providers**: add `glm-5.3` to GLM Coding and GLM (China)
+- **Providers**: Kimchi accepts API keys as well as OAuth (dual auth), with a
+  working Test Connection for both modes
+- **Antigravity**: add Gemini 3.7 Flash and its tiered high/medium/low variants
+  (also in the Gemini registry) with pricing and quota tracking
+- **TTS**: add Fish Audio — model id travels in an HTTP `model` header, voice
+  is a `reference_id` (preset or cloned voice model)
+- **OpenCode-Go**: route by request format via declared transports instead of
+  forcing every client into `/messages` — Codex/OpenAI clients no longer pay a
+  lossy Responses→OpenAI→Claude double translation. Per-model `supportedFormats`
+  guard; the bespoke executor is gone (its shared `_lastModel` cache could cross
+  auth headers between concurrent requests)
+- **Usage**: dedup + cache Claude quota calls (120s TTL keyed by access token,
+  in-flight promise dedup, last-good read on soft failure) to stop multiple
+  tabs tripping 429; manual refresh (↻) sends `force=1` to bypass the cache
+
+## Fixes
+- **Docker**: ship `sql.js` in the image so the pure-JS DB fallback can start —
+  file tracing carried the package's JS without `dist/sql-wasm.wasm`, so a
+  container with no native driver aborted with ENOENT and never got a database
+  (#3248)
+- **Usage**: read Gemini `usageMetadata` out of the antigravity `{ response }`
+  envelope — every non-streaming antigravity request logged `IN 0 | OUT 0`
+  (#3260)
+- **Claude**: re-anchor passthrough cache breakpoints — the client's own
+  `cache_control` markers point at pre-normalization offsets, so the tail was
+  re-cached every request. Last system block and last tool pinned at 1h TTL,
+  last assistant turn at 5m, mid-conversation system messages folded into the
+  neighbouring user turn instead of hoisted into `body.system`
+- **Combos**: detect images from Hermes and attachment payloads (`images[]`,
+  `experimental_attachments`, message-level `image_url`/`audio_url`, inline
+  `data:` URIs) so the Vision Adapter auto-switch fires for Hermes/Ollama/
+  Vercel AI SDK shapes
+- **Kiro**: intercept chat via `x-amz-target` — Kiro IDE 1.0.228+ moved
+  `GenerateAssistantResponse` to `POST /` + header, bypassing MITM. Also emit
+  the now-mandatory initial-response frame and map the `auto` model slot
+- **Kiro**: report real output tokens and stop discarding usable turns
+- **Qoder**: detect billing blocks at stream start and return a synthetic 403
+  so combo/account fallback triggers instead of leaking the error into chat
+- **Antigravity**: strip competitive system prompts (Zed IDE's Claude-agent
+  prompt) that Antigravity flags with a 429 Quota Exhausted
+- **OpenCode**: send the official client fingerprint on free-tier requests so
+  the Console stops classifying traffic as unidentified and rate-limiting it;
+  session id resolves conversation-stable to preserve prompt caching
+- **Responses**: don't close the message on an empty `tool_calls` array — some
+  providers attach one to every chunk, and the truthy check ended the message
+  on the first content token (#3234)
+- **Translator**: preserve `prompt_cache_key` when converting chat to responses
+- **Models**: expose snake_case token limits on `/v1/models`
+- **Combos**: strip `stream_options` from the Fusion panel fan-out to avoid a
+  DeepSeek 400 (#3024); raise the dashboard model-test probe budget to 1024 and
+  soft-pass reasoning-only responses (#3010)
+- **Headroom**: the toggle reflects the `headroomEnabled` setting even when the
+  proxy is down — it previously showed OFF while the engine kept calling
+  `/v1/compress`; proxy status stays visible via the status chip
+- **Hermes**: add the `api_key` parameter to the model block in YAML config
+- **Providers**: add llm7 to provider test support
+
+## Docs
+- **i18n**: add Spanish, French, and Brazilian Portuguese README translations
+
+## Security
+- **Real IP**: `x-9r-real-ip` and the Host fallback were trusted from
+  client-controlled headers whenever `custom-server.js` was not in the request
+  path (`npm run start`, `start:bun`), letting a remote caller pose as local to
+  skip API key auth and reach `LOCAL_ONLY_PATHS` (`/api/mcp/*`,
+  `/api/tunnel/enable`, `/api/auth/reset-password`). The server now stamps a
+  per-process `x-9r-peer-token` on every request it sanitizes and only trusts
+  `x-9r-real-ip` behind it — falling back to Host in development and failing
+  closed in production (GHSA-pjm4-8fpg-f9p6). Also fixes IPv6 loopback
+  detection (`::1`, `::ffff:127.0.0.1`) and routes `npm run start` /
+  `start:bun` through `custom-server.js`
+- **Search**: `resolveBaseUrl()` rejects client-supplied non-public baseUrls
+  (SSRF guard on `/v1/search`)
+- **Login**: fresh-install remote login with the default password returns 403
+  without issuing a JWT
+- **Usage**: `/api/usage/request-details` redacts request/response payloads
+
 # v0.5.50 (2026-08-05)
 
 ## Features

Dockerfile

@@ -37,6 +37,9 @@ COPY --from=builder /app/src/mitm ./src/mitm
 COPY --from=builder /app/node_modules/node-forge ./node_modules/node-forge
 # Ensure `next` is available at runtime in case tracing did not include it.
 COPY --from=builder /app/node_modules/next ./node_modules/next
+# sql.js loads dist/sql-wasm.wasm by path at runtime; tracing only follows JS imports,
+# so the last-resort DB driver would abort with ENOENT on the missing binary.
+COPY --from=builder /app/node_modules/sql.js ./node_modules/sql.js
 
 RUN mkdir -p /app/data && chown -R node:node /app && \
   mkdir -p /app/data-home && chown node:node /app/data-home && \

README.md

@@ -17,7 +17,7 @@
 
 [🚀 Quick Start](#-quick-start) • [💡 Features](#-key-features) • [📖 Setup](#-setup-guide) • [🌐 Website](https://9router.com)
 
-[🇻🇳 Tiếng Việt](./i18n/README.vi.md) • [🇨🇳 中文](./i18n/README.zh-CN.md) • [🇯🇵 日本語](./i18n/README.ja-JP.md) • [🇷🇺 Русский](./i18n/README.ru.md) • [🇹🇭 ไทย](./i18n/README.th.md) • [🇮🇷 فارسی](./i18n/README.fa_IR.md) • [🇮🇩 Indonesia](./i18n/README.id-ID.md)
+[🇧🇷 Português (Brasil)](./i18n/README.pt-BR.md) • [🇻🇳 Tiếng Việt](./i18n/README.vi.md) • [🇨🇳 中文](./i18n/README.zh-CN.md) • [🇯🇵 日本語](./i18n/README.ja-JP.md) • [🇷🇺 Русский](./i18n/README.ru.md) • [🇹🇭 ไทย](./i18n/README.th.md) • [🇮🇷 فارسی](./i18n/README.fa_IR.md) • [🇮🇩 Indonesia](./i18n/README.id-ID.md) • [🇪🇸 Español](./i18n/README.es.md) • [🇫🇷 Français](./i18n/README.fr.md)
 
 </div>
 

cli/package.json

@@ -1,6 +1,6 @@
 {
   "name": "9router",
-  "version": "0.5.50",
+  "version": "0.5.55",
   "description": "9Router CLI - Start and manage 9Router server",
   "bin": {
     "9router": "./cli.js"

cli/scripts/build-cli.js

@@ -216,7 +216,9 @@ function buildCliPackage() {
     fs.copyFileSync(customServerSrc, path.join(cliAppDir, "custom-server.js"));
     console.log("✅ Copied custom-server.js\n");
   } else {
-    console.warn("⚠️  custom-server.js not found — server will run without real-IP injection\n");
+    console.error("❌ custom-server.js not found — without it no request can be proven local,");
+    console.error("   so the packaged CLI would demand an API key for its own dashboard and /v1.");
+    process.exit(1);
   }
 
   // Step 3b: Ensure sql.js (pure JS fallback) bundled in app/cli/app/node_modules.

cli/src/cli/menus/providers.js

@@ -53,6 +53,9 @@ const PROVIDER_MODELS = {
     { id: "glm-4.7" },
   ],
   ag: [
+    { id: "gemini-3.7-flash-high" },
+    { id: "gemini-3.7-flash-medium" },
+    { id: "gemini-3.7-flash-low" },
     { id: "gemini-3.6-flash-high" },
     { id: "gemini-3.6-flash-medium" },
     { id: "gemini-3.6-flash-low" },

custom-server.js

@@ -1,9 +1,18 @@
 const http = require("http");
 const path = require("path");
+const fs = require("fs");
+const crypto = require("crypto");
 const { pathToFileURL } = require("url");
 
 const origCreate = http.createServer.bind(http);
 
+// Per-process secret proving x-9r-real-ip was stamped below rather than sent by the client.
+// A bare `next start` / `next dev` never loads this file, so it cannot produce a matching
+// header even though the env var is inherited by child processes. Named like x-9r-cli-token
+// so the request-detail header sanitizer redacts it too.
+const PEER_TOKEN = crypto.randomBytes(24).toString("hex");
+process.env.NINEROUTER_PEER_TOKEN = PEER_TOKEN;
+
 let backgroundRefreshStarted = false;
 
 function startBackgroundTokenRefreshFromCustomServer() {
@@ -57,7 +66,9 @@ http.createServer = (...args) => {
     delete req.headers["x-9r-real-ip"];
     delete req.headers["x-forwarded-for"];
     delete req.headers["x-9r-via-proxy"];
+    delete req.headers["x-9r-peer-token"];
     req.headers["x-9r-real-ip"] = ip;
+    req.headers["x-9r-peer-token"] = PEER_TOKEN;
     if (viaProxy) req.headers["x-9r-via-proxy"] = "1";
     return handler(req, res);
   };
@@ -114,4 +125,15 @@ http.createServer = (...args) => {
   return server;
 };
 
-if (require.main === module) require("./server.js");
+if (require.main === module) {
+  const standalone = path.join(__dirname, "server.js");
+  if (fs.existsSync(standalone)) {
+    require(standalone);
+  } else {
+    // Repo checkout has no standalone build next to us. `next start` builds its HTTP
+    // server in-process, so the wrapper above still sanitizes every request.
+    const nextBin = require.resolve("next/dist/bin/next");
+    process.argv = [process.argv[0], nextBin, "start", ...process.argv.slice(2)];
+    require(nextBin);
+  }
+}

open-sse/config/providerModels.js

@@ -2,7 +2,7 @@ import { PROVIDERS } from "./providers.js";
 import REGISTRY from "../providers/registry/index.js";
 // PROVIDER_MODELS now built from providers/registry (transport + models co-located)
 import { PROVIDER_MODELS } from "../providers/index.js";
-import { modelQuotaFamily, modelStrip, modelTargetFormat, normalizeModelId } from "../providers/models/schema.js";
+import { modelQuotaFamily, modelStrip, modelTargetFormat, modelSupportedFormats, normalizeModelId } from "../providers/models/schema.js";
 import { CODEX_REVIEW_SUFFIX } from "../providers/models/helpers.js";
 export { PROVIDER_MODELS };
 
@@ -54,6 +54,14 @@ export function getModelTargetFormat(aliasOrId, modelId) {
   return modelTargetFormat(findModel(models, modelId, aliasOrId));
 }
 
+// Declared upstream formats for a model (registry `supportedFormats`). Drives the
+// per-model guard on the sourceFormat-matched transport; null when undeclared.
+export function getModelSupportedFormats(aliasOrId, modelId) {
+  const models = PROVIDER_MODELS[aliasOrId];
+  if (!models) return null;
+  return modelSupportedFormats(findModel(models, modelId, aliasOrId));
+}
+
 export function getModelType(aliasOrId, modelId) {
   const models = PROVIDER_MODELS[aliasOrId];
   if (!models) return null;

open-sse/executors/antigravity.js

@@ -245,6 +245,18 @@ export class AntigravityExecutor extends BaseExecutor {
     // Strip tools/toolConfig (handled separately) and blacklisted fields that Google rejects
     const { tools: _originalTools, toolConfig: _originalToolConfig, ...requestWithoutTools } = body.request || {};
     stripBlacklisted(requestWithoutTools);
+    
+    // Rewrite competitive system prompts (e.g. Zed IDE's Claude prompt) to prevent Antigravity from 
+    // flagging the request and immediately blocking it with a 429 Quota Exhausted response.
+    if (requestWithoutTools.systemInstruction?.parts) {
+      const oldText = "You are a Claude agent, built on Anthropic's Claude Agent SDK.";
+      for (const part of requestWithoutTools.systemInstruction.parts) {
+        if (typeof part.text === "string" && part.text.includes(oldText)) {
+          part.text = part.text.split(oldText).join("");
+        }
+      }
+    }
+
     const generationConfig = { ...(requestWithoutTools.generationConfig || {}) };
     if (generationConfig.maxOutputTokens > MAX_ANTIGRAVITY_OUTPUT_TOKENS) {
       generationConfig.maxOutputTokens = MAX_ANTIGRAVITY_OUTPUT_TOKENS;

open-sse/executors/index.js

@@ -10,7 +10,6 @@ import { CodexExecutor } from "./codex.js";
 import { CursorExecutor } from "./cursor.js";
 import { VertexExecutor } from "./vertex.js";
 import { OpenCodeExecutor } from "./opencode.js";
-import { OpenCodeGoExecutor } from "./opencode-go.js";
 import { GrokWebExecutor } from "./grok-web.js";
 import { GrokCliExecutor } from "./grok-cli.js";
 import { PerplexityWebExecutor } from "./perplexity-web.js";
@@ -41,7 +40,6 @@ const executors = {
   vertex: new VertexExecutor("vertex"),
   "vertex-partner": new VertexExecutor("vertex-partner"),
   opencode: new OpenCodeExecutor(),
-  "opencode-go": new OpenCodeGoExecutor(),
   "grok-web": new GrokWebExecutor(),
   "grok-cli": new GrokCliExecutor(),
   gcli: new GrokCliExecutor(), // Alias
@@ -86,7 +84,6 @@ export { CursorExecutor } from "./cursor.js";
 export { VertexExecutor } from "./vertex.js";
 export { DefaultExecutor } from "./default.js";
 export { OpenCodeExecutor } from "./opencode.js";
-export { OpenCodeGoExecutor } from "./opencode-go.js";
 export { GrokWebExecutor } from "./grok-web.js";
 export { GrokCliExecutor } from "./grok-cli.js";
 export { PerplexityWebExecutor } from "./perplexity-web.js";

open-sse/executors/kiro.js

@@ -144,6 +144,12 @@ function normalizeStopReason(value) {
   return reason || null;
 }
 
+// Of the reasons stopDisposition() folds into "terminal_incomplete", only these
+// mean "usable as far as it got, then the budget ran out" -- the case
+// finish_reason "length" exists for. cancelled / pause_turn are abandoned turns
+// whose partial content must stay private, so they are deliberately absent.
+const KIRO_TRUNCATION_STOP_REASONS = new Set(["model_context_window_exceeded", "max_tokens"]);
+
 function stopDisposition(stopReason, hasToolCalls) {
   if (["malformed_model_output", "invalid_model_output"].includes(stopReason)) return "retryable_protocol_failure";
   if (["cancelled", "pause_turn", "model_context_window_exceeded"].includes(stopReason)) return "terminal_incomplete";
@@ -711,14 +717,25 @@ export class KiroExecutor extends BaseExecutor {
     };
     const emitTools = (controller) => {
       for (const tool of state.tools.values()) {
-        const input = parsedToolInput(tool);
-        if (tool.name === "tool_call") {
-          if (typeof input.name !== "string" || !input.name.trim()) {
-            throw new Error("Invalid Kiro tool_call payload: missing nested MCP tool name");
-          }
-          if (!Object.prototype.hasOwnProperty.call(input, "arguments")) {
-            throw new Error("Invalid Kiro tool_call payload: missing nested MCP tool arguments");
+        // Validate per tool, not per turn: one unusable fragment used to throw out
+        // of emitTools and take every other complete tool call in the same turn
+        // with it, which the client saw as a turn that answered nothing.
+        let input;
+        try {
+          input = parsedToolInput(tool);
+          if (tool.name === "tool_call") {
+            if (typeof input.name !== "string" || !input.name.trim()) {
+              throw new Error("Invalid Kiro tool_call payload: missing nested MCP tool name");
+            }
+            if (!Object.prototype.hasOwnProperty.call(input, "arguments")) {
+              throw new Error("Invalid Kiro tool_call payload: missing nested MCP tool arguments");
+            }
           }
+        } catch (error) {
+          state.droppedTools = (state.droppedTools || 0) + 1;
+          state.toolValidationError ||= error.message;
+          console.error(`[Kiro] dropping unusable tool call ${tool.id} (${tool.name}): ${error.message}`);
+          continue;
         }
         const index = state.toolCounter++;
         emitDelta(controller, {
@@ -729,14 +746,26 @@ export class KiroExecutor extends BaseExecutor {
             function: { name: tool.name, arguments: "" }
           }]
         });
+        const serializedInput = JSON.stringify(input);
         emitDelta(controller, {
-          tool_calls: [{ index, function: { arguments: JSON.stringify(input) } }]
+          tool_calls: [{ index, function: { arguments: serializedInput } }]
         });
+        // Tool arguments are billed output like any other completion bytes. They
+        // were never added to totalContentLength, so the /4 estimator in finish()
+        // reported OUT 0 -- or the Math.max floor of 1 -- for every turn whose
+        // entire answer was a tool call.
+        state.totalContentLength += tool.name.length + serializedInput.length;
         state.hasToolCalls = true;
       }
       state.tools.clear();
       state.bufferedToolBytes = 0;
-      if (state.stopReason === "tool_use" && !state.hasToolCalls) {
+      // A declared tool turn that emitted no usable call is only fatal when the
+      // turn produced nothing else. Throwing unconditionally here escaped
+      // emitTools() with provenance "invalid_tool_call", which the integrity gate
+      // re-derived into a repair retry -- discarding text the client had already
+      // been promised.
+      if (state.stopReason === "tool_use" && !state.hasToolCalls &&
+          !state.hasText && !state.hasReasoning && !state.hasCode) {
         throw new Error("Kiro tool_use stop reason did not include a complete tool call");
       }
     };
@@ -796,7 +825,6 @@ export class KiroExecutor extends BaseExecutor {
         emitDelta(controller, { content: event.payload.content });
       } else if (eventType === "toolUseEvent") {
         state.sawToolUse = true;
-        if (state.toolValidationError) return true;
         const values = Array.isArray(event.payload) ? event.payload : [event.payload];
         if (!values[0]) throw new Error("Kiro toolUseEvent is empty");
         for (const value of values) {
@@ -924,9 +952,10 @@ export class KiroExecutor extends BaseExecutor {
         } catch (error) {
           const bufferExceeded = error.code === "KIRO_BUFFER_EXCEEDED";
           if (!bufferExceeded) {
+            // Keep whatever is already buffered: the rejected fragment belongs to
+            // one tool, and clearing the map dropped the complete calls too.
             state.toolValidationError ||= error.message;
-            state.tools.clear();
-            state.bufferedToolBytes = 0;
+            console.error(`[Kiro] tool fragment rejected, keeping ${state.tools.size} buffered tool(s): ${error.message}`);
             continue;
           }
           fail(
@@ -958,7 +987,16 @@ export class KiroExecutor extends BaseExecutor {
       }
       state.transportState = "clean_eof";
       const declaredDisposition = stopDisposition(state.stopReason, state.sawToolUse);
-      if (["retryable_protocol_failure", "terminal_incomplete", "terminal_refusal", "unknown_failure"].includes(declaredDisposition)) {
+      // model_context_window_exceeded / max_tokens map to terminal_incomplete. When
+      // they arrive after the model already streamed content, fail() threw away a
+      // complete-enough answer; a truncated turn is what finish_reason "length" is
+      // for. chunkIndex > 0 means at least one delta already reached the client.
+      const declaredTruncatedAfterOutput = declaredDisposition === "terminal_incomplete" &&
+        KIRO_TRUNCATION_STOP_REASONS.has(state.stopReason) && state.chunkIndex > 0;
+      if (declaredTruncatedAfterOutput) {
+        console.error(`[Kiro] truncated after ${state.chunkIndex} chunk(s) (stop_reason=${state.stopReason}); keeping output`);
+      }
+      if (!declaredTruncatedAfterOutput && ["retryable_protocol_failure", "terminal_incomplete", "terminal_refusal", "unknown_failure"].includes(declaredDisposition)) {
         const code = declaredDisposition === "retryable_protocol_failure"
           ? "kiro_retryable_protocol_failure"
           : declaredDisposition === "terminal_refusal"
@@ -975,16 +1013,6 @@ export class KiroExecutor extends BaseExecutor {
         );
         return;
       }
-      if (state.toolValidationError) {
-        fail(
-          controller,
-          "invalid_tool_call",
-          "invalid_kiro_tool_call",
-          state.toolValidationError,
-          { transport_state: state.transportState, stop_disposition: "retryable_protocol_failure" }
-        );
-        return;
-      }
       try {
         emitTools(controller);
       } catch (error) {
@@ -997,6 +1025,22 @@ export class KiroExecutor extends BaseExecutor {
         );
         return;
       }
+      // Fail only when the turn has nothing usable left. emitTools() validates
+      // per tool and drops just the unusable ones, so this has to run AFTER it:
+      // before, the rejected tool was still buffered and tools.size was never 0.
+      // A turn that also produced text keeps that text -- the dropped call is
+      // logged, not fatal.
+      if (state.toolValidationError && !state.hasToolCalls &&
+          !state.hasText && !state.hasReasoning && !state.hasCode) {
+        fail(
+          controller,
+          "invalid_tool_call",
+          "invalid_kiro_tool_call",
+          state.toolValidationError,
+          { transport_state: state.transportState, stop_disposition: "retryable_protocol_failure" }
+        );
+        return;
+      }
 
       const hasOutput = state.hasText || state.hasReasoning || state.hasCode || state.hasToolCalls;
       if (!hasOutput && !state.explicitStop) {
@@ -1011,7 +1055,13 @@ export class KiroExecutor extends BaseExecutor {
       }
 
       const disposition = stopDisposition(state.stopReason, state.hasToolCalls);
-      if (["retryable_protocol_failure", "terminal_incomplete", "terminal_refusal", "unknown_failure"].includes(disposition)) {
+      // Same reasoning as declaredTruncatedAfterOutput above.
+      const truncatedAfterOutput = disposition === "terminal_incomplete" &&
+        KIRO_TRUNCATION_STOP_REASONS.has(state.stopReason) && state.chunkIndex > 0;
+      if (truncatedAfterOutput) {
+        console.error(`[Kiro] truncated after ${state.chunkIndex} chunk(s) (stop_reason=${state.stopReason}); closing as length`);
+      }
+      if (!truncatedAfterOutput && ["retryable_protocol_failure", "terminal_incomplete", "terminal_refusal", "unknown_failure"].includes(disposition)) {
         const code = disposition === "retryable_protocol_failure"
           ? "kiro_retryable_protocol_failure"
           : disposition === "terminal_refusal"
@@ -1041,18 +1091,24 @@ export class KiroExecutor extends BaseExecutor {
           total_tokens: prompt + completion
         };
       }
-      const finishReason = state.hasToolCalls
-        ? "tool_calls"
-        : disposition === "length"
-          ? "length"
-          : "stop";
+      const finishReason = truncatedAfterOutput
+        ? "length"
+        : state.hasToolCalls
+          ? "tool_calls"
+          : disposition === "length"
+            ? "length"
+            : "stop";
       controller.enqueue(sseChunk({}, finishReason, state.usage));
       controller.enqueue(encoder.encode(SSE_DONE));
       state.finished = true;
       options.onTerminalState?.(diagnostics({
         terminal_provenance: state.terminalProvenance || "clean_eventstream_eof",
         transport_state: state.transportState,
-        stop_disposition: disposition
+        // Report what this exit actually did, not the raw disposition. The
+        // integrity gate re-derives its verdict from stop_disposition, so
+        // reporting "terminal_incomplete" for a turn we deliberately kept made
+        // it discard the very bytes we just released to the client.
+        stop_disposition: truncatedAfterOutput ? "length" : disposition
       }));
     };
 

open-sse/executors/opencode-go.js

@@ -1,49 +0,0 @@
-import { BaseExecutor } from "./base.js";
-import { PROVIDERS } from "../config/providers.js";
-import { injectReasoningContent } from "../utils/reasoningContentInjector.js";
-import { ANTHROPIC_API_VERSION } from "../providers/shared.js";
-
-// Models that use /zen/go/v1/messages (Anthropic/Claude format + x-api-key auth)
-const MESSAGES_FORMAT_MODELS = new Set([
-  "minimax-m3",
-  "minimax-m2.7",
-  "minimax-m2.5",
-  "qwen3.7-max",
-  "qwen3.7-plus",
-  "qwen3.6-plus",
-]);
-
-const BASE = "https://opencode.ai/zen/go/v1";
-
-export class OpenCodeGoExecutor extends BaseExecutor {
-  constructor() {
-    super("opencode-go", PROVIDERS["opencode-go"]);
-  }
-
-  // buildUrl runs before buildHeaders in BaseExecutor.execute, cache model here
-  buildUrl(model) {
-    this._lastModel = model;
-    return MESSAGES_FORMAT_MODELS.has(model)
-      ? `${BASE}/messages`
-      : `${BASE}/chat/completions`;
-  }
-
-  buildHeaders(credentials, stream = true) {
-    const key = credentials?.apiKey || credentials?.accessToken;
-    const headers = { "Content-Type": "application/json" };
-
-    if (MESSAGES_FORMAT_MODELS.has(this._lastModel)) {
-      headers["x-api-key"] = key;
-      headers["anthropic-version"] = ANTHROPIC_API_VERSION;
-    } else {
-      headers["Authorization"] = `Bearer ${key}`;
-    }
-
-    if (stream) headers["Accept"] = "text/event-stream";
-    return headers;
-  }
-
-  transformRequest(model, body) {
-    return injectReasoningContent({ provider: this.provider, model, body });
-  }
-}

open-sse/executors/opencode.js

@@ -1,16 +1,43 @@
+import crypto from "crypto";
 import { BaseExecutor } from "./base.js";
 import { PROVIDERS } from "../config/providers.js";
 import { injectReasoningContent } from "../utils/reasoningContentInjector.js";
+import { resolveSessionId } from "../utils/sessionManager.js";
 
-// Models that use /zen/v1/messages (claude format)
+const OPENCODE_UA = "opencode";
 const MESSAGES_MODELS = new Set();
 
+function generateRequestId() {
+  return `msg_${crypto.randomUUID().replace(/-/g, "")}`;
+}
+
+function generateSessionId() {
+  return `ses_${crypto.randomUUID().replace(/-/g, "")}`;
+}
+
+// Normalize any resolved id into opencode's ses_ format (stable per-conversation)
+function toOpencodeSession(id) {
+  const stripped = String(id || "").replace(/^ses_/, "").replace(/-/g, "");
+  return stripped ? `ses_${stripped}` : null;
+}
+
+function resolveOpencodeSession(body, credentials) {
+  return toOpencodeSession(resolveSessionId({
+    headers: credentials?.rawHeaders,
+    body,
+    connectionId: credentials?.connectionId,
+    scope: "opencode",
+  }));
+}
+
 export class OpenCodeExecutor extends BaseExecutor {
   constructor() {
     super("opencode", PROVIDERS.opencode);
+    this._currentSessionId = null;
   }
 
-  transformRequest(model, body) {
+  transformRequest(model, body, stream, credentials) {
+    this._currentSessionId = resolveOpencodeSession(body, credentials);
     return injectReasoningContent({ provider: this.provider, model, body });
   }
 
@@ -21,12 +48,23 @@ export class OpenCodeExecutor extends BaseExecutor {
       : `${base}/zen/v1/chat/completions`;
   }
 
-  buildHeaders() {
+  buildHeaders(credentials, stream = true) {
+    const raw = credentials?.rawHeaders || {};
+    const lower = {};
+    for (const [k, v] of Object.entries(raw)) lower[k.toLowerCase()] = v;
+
+    const downstreamUa = lower["user-agent"] || "";
+    const isOpencodeDownstream = downstreamUa.toLowerCase().includes("opencode");
+
     return {
       "Content-Type": "application/json",
       "Authorization": "Bearer public",
-      "x-opencode-client": "desktop",
-      "Accept": "text/event-stream"
+      "User-Agent": isOpencodeDownstream ? downstreamUa : OPENCODE_UA,
+      "x-opencode-client": lower["x-opencode-client"] || "desktop",
+      "x-opencode-session": lower["x-opencode-session"] || this._currentSessionId || generateSessionId(),
+      "x-opencode-request": lower["x-opencode-request"] || generateRequestId(),
+      "x-opencode-project": lower["x-opencode-project"] || "global",
+      "Accept": stream ? "text/event-stream" : "*/*",
     };
   }
 }

open-sse/executors/qoder.js

@@ -215,6 +215,52 @@ async function buildQoderRequestBody({ model, body, credentials, log, proxyOptio
   };
 }
 
+/**
+ * Check if a qoder error message indicates a billing/quota block.
+ * Signatures: code 112 (quota exhausted), code 10605 (queue throttle), pricingUrl field.
+ */
+function isBillingBlock(inner) {
+  if (!inner || typeof inner !== "string") return false;
+  const lowerMsg = inner.toLowerCase();
+  // Match: {"code":"112",...}, {"code":"10605",...}, or pricingUrl field
+  return /\"code\"\s*:\s*\"(112|10605)\"/.test(inner) || lowerMsg.includes("pricingurl");
+}
+
+/**
+ * Peek the first SSE frame to detect billing errors before piping.
+ * Returns { isBilling, statusVal, message, consumed } — `consumed` is every
+ * byte read so far (including the peeked line) so the caller can re-process
+ * it and nothing is dropped from the stream.
+ */
+async function peekFirstQoderFrame(reader, decoder) {
+  let consumed = "";
+  while (true) {
+    const { done, value } = await reader.read();
+    if (done) return { isBilling: false, consumed, upstreamDone: true };
+
+    consumed += decoder.decode(value, { stream: true });
+    const nl = consumed.indexOf("\n");
+    if (nl === -1) continue; // need a full line first
+
+    const line = consumed.slice(0, nl).replace(/\r$/, "").trim();
+    if (!line.startsWith("data:")) continue;
+
+    const data = line.slice(5).trimStart();
+    if (data === "[DONE]") return { isBilling: false, consumed };
+
+    let envelope;
+    try { envelope = JSON.parse(data); } catch { return { isBilling: false, consumed }; }
+
+    const statusVal = typeof envelope.statusCodeValue === "number" ? envelope.statusCodeValue : 200;
+    const inner = typeof envelope.body === "string" ? envelope.body : "";
+
+    if (statusVal !== 200 && isBillingBlock(inner)) {
+      return { isBilling: true, statusVal, message: inner || `qoder billing block (${statusVal})` };
+    }
+    return { isBilling: false, consumed };
+  }
+}
+
 /**
  * Wrap the upstream's `{statusCodeValue, body}` SSE envelope into plain
  * OpenAI SSE chunks the rest of the chatCore pipeline understands.
@@ -229,16 +275,34 @@ async function buildQoderRequestBody({ model, body, credentials, log, proxyOptio
  * [DONE]/error frame (agent keepalive). Non-streaming clients drain via
  * response.text() which hangs until the socket closes — so on terminal
  * events we cancel the upstream reader and close our stream immediately.
+ *
+ * NEW: Peek first frame to detect billing blocks (code 112/10605/pricingUrl).
+ * If detected, return 403 response so chatCore marks connection unavailable
+ * and triggers combo fallback instead of leaking error text into chat.
  */
-function wrapQoderSSE(response, model) {
+async function wrapQoderSSE(response, model) {
   if (!response.ok || !response.body) return response;
 
   const decoder = new TextDecoder();
-  const encoder = new TextEncoder();
-  let buffer = "";
-  let doneEmitted = false;
   const reader = response.body.getReader();
 
+  // Peek first frame to detect billing block
+  const peek = await peekFirstQoderFrame(reader, decoder);
+  if (peek?.isBilling) {
+    // Billing block detected — return 403 so chatCore fails this connection
+    await reader.cancel().catch(() => {});
+    return new Response(
+      JSON.stringify({ error: { message: peek.message, code: peek.statusVal } }),
+      { status: 403, headers: { "Content-Type": "application/json" } }
+    );
+  }
+
+  // Normal flow: re-process every byte the peek consumed, then continue.
+  let buffer = peek.consumed || "";
+  const upstreamDrained = peek.upstreamDone === true;
+  const encoder = new TextEncoder();
+  let doneEmitted = false;
+
   // Process one already-extracted SSE line (no trailing newline).
   const processLine = (line, controller) => {
     const trimmed = line.replace(/\r$/, "").trim();
@@ -287,7 +351,28 @@ function wrapQoderSSE(response, model) {
     // enqueueing would never be re-invoked, hanging consumers like .text().
     async start(controller) {
       try {
-        while (!doneEmitted) {
+        // Drain whatever the peek already pulled off the socket first.
+        let nlSeed;
+        while ((nlSeed = buffer.indexOf("\n")) !== -1) {
+          const line = buffer.slice(0, nlSeed);
+          buffer = buffer.slice(nlSeed + 1);
+          processLine(line, controller);
+          if (doneEmitted) {
+            await reader.cancel().catch(() => {});
+            controller.close();
+            return;
+          }
+        }
+        if (upstreamDrained) {
+          // Peek hit end-of-stream: flush any trailing partial line.
+          buffer += decoder.decode();
+          if (buffer.length > 0) {
+            processLine(buffer, controller);
+            buffer = "";
+          }
+        }
+
+        while (!doneEmitted && !upstreamDrained) {
           const { done, value } = await reader.read();
           if (done) {
             buffer += decoder.decode();
@@ -472,7 +557,7 @@ export class QoderExecutor extends BaseExecutor {
       return { response, url, headers, transformedBody: payload };
     }
 
-    const wrapped = wrapQoderSSE(response, `qoder/${qoderKey}`);
+    const wrapped = await wrapQoderSSE(response, `qoder/${qoderKey}`);
     return { response: wrapped, url, headers, transformedBody: payload };
   }
 
@@ -496,4 +581,5 @@ export const __test__ = {
   normalizeMessages,
   wrapQoderSSE,
   buildQoderRequestBody,
+  isBillingBlock,
 };

open-sse/handlers/chatCore.js

@@ -2,11 +2,11 @@ import { detectFormat, getTargetFormat, resolveTransport } from "../services/pro
 import { translateRequest } from "../translator/index.js";
 import { applyThinking, extractThinking, stripThinkingSuffix } from "../translator/concerns/thinkingUnified.js";
 import { FORMATS } from "../translator/formats.js";
-import { normalizeClaudePassthrough } from "../translator/formats/claude.js";
+import { normalizeClaudePassthrough, anchorClaudeCache } from "../translator/formats/claude.js";
 import { createStreamController } from "../utils/streamHandler.js";
 import { refreshWithRetry } from "../services/tokenRefresh.js";
 import { createRequestLogger } from "../utils/requestLogger.js";
-import { getModelTargetFormat, getModelStrip, getModelUpstreamId, getModelType, PROVIDER_ID_TO_ALIAS } from "../config/providerModels.js";
+import { getModelTargetFormat, getModelSupportedFormats, getModelStrip, getModelUpstreamId, getModelType, PROVIDER_ID_TO_ALIAS } from "../config/providerModels.js";
 import { PROVIDERS } from "../config/providers.js";
 import { createErrorResult, parseUpstreamError, formatProviderError } from "../utils/error.js";
 import { HTTP_STATUS, TOKEN_SAVER_HEADER } from "../config/runtimeConfig.js";
@@ -78,10 +78,20 @@ export async function handleChatCore({ body, modelInfo, credentials, log, onCred
 
   const alias = PROVIDER_ID_TO_ALIAS[provider] || provider;
   const modelTargetFormat = getModelTargetFormat(alias, model);
-  // Multi-endpoint providers: pick transport matching sourceFormat → zero translation
+  // Multi-endpoint providers: pick transport matching sourceFormat → zero translation.
+  // Per-model guard: only use the transport when the model declares support for that
+  // sourceFormat — opencode-go models differ in endpoint support (kimi/glm only do
+  // /chat/completions), so without this guard a claude-format request would wrongly
+  // route kimi to /messages.
+  const modelSupportedFormats = getModelSupportedFormats(alias, model);
   const runtimeTransport = resolveTransport(provider, sourceFormat);
-  const targetFormat = modelTargetFormat || runtimeTransport?.format || getTargetFormat(provider, credentials);
-  if (runtimeTransport && credentials) credentials.runtimeTransport = runtimeTransport;
+  // Per-model guard: when a model declares supportedFormats, only use the
+  // sourceFormat-matched transport if that format is declared (opencode-go models
+  // differ — kimi/glm only do /chat/completions). Undeclared models keep the
+  // upstream default (use the transport), preserving behavior for glm/deepseek/...
+  const useTransport = (!modelSupportedFormats || modelSupportedFormats.includes(sourceFormat)) ? runtimeTransport : null;
+  const targetFormat = modelTargetFormat || useTransport?.format || getTargetFormat(provider, credentials);
+  if (useTransport && credentials) credentials.runtimeTransport = useTransport;
   const stripList = getModelStrip(alias, model);
   const upstreamModel = getModelUpstreamId(alias, model);
 
@@ -275,6 +285,10 @@ export async function handleChatCore({ body, modelInfo, credentials, log, onCred
 
   if (xf.length && log?.line) log.line(reqTag, "⚙", xf.join(" · "));
 
+  // Pin cache breakpoints to the final body — every saver above can reshape
+  // system/tools/messages, and a stale anchor costs a full prefix rewrite.
+  if (passthrough && clientTool === "claude") anchorClaudeCache(translatedBody);
+
   const executor = getExecutor(provider);
   trackPendingRequest(model, provider, connectionId, true);
   appendRequestLog({ model, provider, connectionId, status: "PENDING" }).catch(() => { });

open-sse/handlers/chatCore/requestDetail.js

@@ -44,13 +44,14 @@ export function extractUsageFromResponse(responseBody) {
     };
   }
 
-  // Gemini format
-  if (responseBody.usageMetadata) {
+  // Gemini format. Antigravity / gemini-cli wrap the payload in { response: {...} }.
+  const usageMetadata = responseBody.usageMetadata || responseBody.response?.usageMetadata;
+  if (usageMetadata) {
     return {
-      prompt_tokens: responseBody.usageMetadata.promptTokenCount || 0,
-      completion_tokens: responseBody.usageMetadata.candidatesTokenCount || 0,
-      cached_tokens: responseBody.usageMetadata.cachedContentTokenCount || 0,
-      reasoning_tokens: responseBody.usageMetadata.thoughtsTokenCount || 0
+      prompt_tokens: usageMetadata.promptTokenCount || 0,
+      completion_tokens: usageMetadata.candidatesTokenCount || 0,
+      cached_tokens: usageMetadata.cachedContentTokenCount || 0,
+      reasoning_tokens: usageMetadata.thoughtsTokenCount || 0
     };
   }
 

open-sse/handlers/search/callers.js

@@ -29,6 +29,8 @@
  * @property {Record<string,unknown>} [providerSpecificData]
  */
 
+import { assertPublicUrl } from "../../../src/shared/utils/ssrfGuard.js";
+
 // ── Helpers ─────────────────────────────────────────────────────────────
 
 /**
@@ -63,12 +65,31 @@ export function getProviderSetting(params, key) {
 
 /**
  * Resolve base URL with optional override from providerOptions.baseUrl.
+ *
+ * The override is client-controlled and therefore SSRF-hardened: only public
+ * http(s) URLs are accepted (internal/private/loopback/metadata addresses are
+ * rejected via assertPublicUrl). The provider's own configured baseUrl is
+ * trusted as-is (admin-controlled).
+ *
  * @param {SearchProviderConfig} config
  * @param {SearchRequestParams} params
  * @returns {string}
  */
 export function resolveBaseUrl(config, params) {
   const override = getProviderSetting(params, "baseUrl");
+  if (override) {
+    // SSRF guard: client-supplied base URLs must be public http(s) only.
+    let parsed;
+    try {
+      parsed = new URL(override);
+    } catch {
+      throw new Error(`Invalid baseUrl: ${override}`);
+    }
+    if (parsed.protocol !== "http:" && parsed.protocol !== "https:") {
+      throw new Error(`Invalid baseUrl protocol: ${parsed.protocol}`);
+    }
+    assertPublicUrl(override);
+  }
   return (override || config.baseUrl).replace(/\/+$/, "");
 }
 

open-sse/handlers/ttsProviders/genericFormats.js

@@ -51,6 +51,25 @@ async function huggingface({ baseUrl, apiKey, text, modelId }) {
   return responseToBase64(res, "wav");
 }
 
+// Fish Audio: model travels in an HTTP header, the voice is a reference_id, returns binary
+async function fishAudio({ baseUrl, apiKey, text, modelId, voiceId }) {
+  const res = await fetch(baseUrl, {
+    method: "POST",
+    headers: {
+      "Content-Type": "application/json",
+      "Authorization": `Bearer ${apiKey}`,
+      "model": modelId || "s2.1-pro-free",
+    },
+    body: JSON.stringify({
+      text,
+      format: "mp3",
+      ...(voiceId ? { reference_id: voiceId } : {}),
+    }),
+  });
+  if (!res.ok) await throwUpstreamError(res);
+  return responseToBase64(res, "mp3");
+}
+
 // Inworld: Basic auth, JSON { audioContent }
 async function inworld({ baseUrl, apiKey, text, modelId, voiceId }) {
   const res = await fetch(baseUrl, {
@@ -166,4 +185,5 @@ export const FORMAT_HANDLERS = {
   tortoise,
   openai: openaiCompat,
   "minimax-tts": minimaxTts,
+  "fish-audio": fishAudio,
 };

open-sse/providers/capabilities.js

@@ -205,6 +205,7 @@ export const PATTERN_CAPABILITIES = [
 
   // ── Gemini (all 2.0+ multimodal + google_search grounding, 1M ctx) ─
   { pattern: "*gemini*image*",  caps: { vision: true, imageOutput: true, contextWindow: 1048576 } },
+  { pattern: "*gemini-3.7*",    caps: { vision: true, audioInput: true, videoInput: true, reasoning: true, search: true, thinkingFormat: "gemini-level", thinkingCanDisable: false, contextWindow: 1048576, maxOutput: 65536 } },
   { pattern: "*gemini-3*pro*",  caps: { vision: true, audioInput: true, videoInput: true, reasoning: true, search: true, thinkingFormat: "gemini-level", thinkingCanDisable: false, contextWindow: 1048576, maxOutput: 65535 } },
   { pattern: "*gemini-3*",      caps: { vision: true, audioInput: true, videoInput: true, reasoning: true, search: true, thinkingFormat: "gemini-level", thinkingCanDisable: false, contextWindow: 1048576, maxOutput: 65536 } },
   { pattern: "*gemini-2.5*",    caps: { vision: true, audioInput: true, videoInput: true, reasoning: true, search: true, thinkingFormat: "gemini-budget", thinkingRange: { min: 0, max: 24576 }, contextWindow: 1048576, maxOutput: 65536 } },

open-sse/providers/models/schema.js

@@ -38,3 +38,11 @@ export function modelStrip(model) {
 export function modelTargetFormat(model) {
   return model?.targetFormat || MODEL_DEFAULTS.targetFormat;
 }
+
+// Per-model declared upstream formats (e.g. ["openai", "claude"]). Guards the
+// sourceFormat-matched transport for multi-endpoint providers whose models differ
+// in endpoint support (opencode-go: kimi/glm only do /chat/completions, minimax/qwen
+// also do /messages, deepseek also does /responses).
+export function modelSupportedFormats(model) {
+  return model?.supportedFormats || null;
+}

open-sse/providers/pricing.js

@@ -57,6 +57,10 @@ export const MODEL_PRICING = {
   "o1-mini":                      { input: 3.00,  output: 12.00, cached: 1.50,  reasoning: 18.00,  cache_creation: 3.00  },
 
   // === Gemini ===
+  "gemini-3.7-flash":              { input: 1.50,  output: 7.50,  cached: 0.15,  reasoning: 11.25,  cache_creation: 1.875 },
+  "gemini-3.7-flash-high":         { input: 1.50,  output: 7.50,  cached: 0.15,  reasoning: 11.25,  cache_creation: 1.875 },
+  "gemini-3.7-flash-medium":       { input: 1.50,  output: 7.50,  cached: 0.15,  reasoning: 11.25,  cache_creation: 1.875 },
+  "gemini-3.7-flash-low":          { input: 1.50,  output: 7.50,  cached: 0.15,  reasoning: 11.25,  cache_creation: 1.875 },
   "gemini-3.6-flash":              { input: 1.50,  output: 7.50,  cached: 0.15,  reasoning: 11.25,  cache_creation: 1.875 },
   "gemini-3.6-flash-high":         { input: 1.50,  output: 7.50,  cached: 0.15,  reasoning: 11.25,  cache_creation: 1.875 },
   "gemini-3.6-flash-medium":       { input: 1.50,  output: 7.50,  cached: 0.15,  reasoning: 11.25,  cache_creation: 1.875 },

open-sse/providers/registry/alitp-intl.js

@@ -0,0 +1,35 @@
+// Token Plan — credit subscription keys on token-plan.<region>.maas.aliyuncs.com.
+// Fourth Alibaba key type: Coding Plan (alicode/alicode-intl) and Model Studio
+// (alims-intl) both reject these keys, and they reject Model Studio keys back.
+// Singapore is the only region that serves the plan; eu-central-1 answers
+// IllegalEndpoint. The Anthropic surface (/apps/anthropic/v1/messages) is not
+// authorized for this plan, so OpenAI-compatible mode is the only transport.
+export default {
+  id: "alitp-intl",
+  priority: 11,
+  alias: "alitp-intl",
+  display: {
+    name: "Alibaba Token Plan",
+    icon: "cloud",
+    color: "#FF6A00",
+    textIcon: "ATP",
+    website: "https://www.alibabacloud.com/campaign/ai-landing-page-token",
+    notice: {
+      apiKeyUrl: "https://modelstudio.console.alibabacloud.com/?apiKey=1",
+    },
+  },
+  category: "apikey",
+  transport: {
+    baseUrl: "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions",
+    headers: {},
+    quirks: { preserveCacheControl: true },
+  },
+  models: [
+    { id: "qwen3.8-max-preview", name: "Qwen3.8 Max Preview" },
+    { id: "qwen3.7-max", name: "Qwen3.7 Max" },
+    { id: "qwen3.7-plus", name: "Qwen3.7 Plus" },
+    { id: "qwen3.6-flash", name: "Qwen3.6 Flash" },
+    { id: "glm-5.2", name: "GLM 5.2" },
+    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
+  ],
+};

open-sse/providers/registry/antigravity.js

@@ -45,6 +45,9 @@ export default {
     clientSecret: "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",
   },
   models: [
+    { id: "gemini-3.7-flash-high", name: "Gemini 3.7 Flash (High)", upstreamModelId: "gemini-3.7-flash-tiered(high)" },
+    { id: "gemini-3.7-flash-medium", name: "Gemini 3.7 Flash (Medium)", upstreamModelId: "gemini-3.7-flash-tiered(medium)" },
+    { id: "gemini-3.7-flash-low", name: "Gemini 3.7 Flash (Low)", upstreamModelId: "gemini-3.7-flash-tiered(low)" },
     { id: "gemini-3.6-flash-high", name: "Gemini 3.6 Flash (High)", upstreamModelId: "gemini-3.6-flash-tiered(high)" },
     { id: "gemini-3.6-flash-medium", name: "Gemini 3.6 Flash (Medium)", upstreamModelId: "gemini-3.6-flash-tiered(medium)" },
     { id: "gemini-3.6-flash-low", name: "Gemini 3.6 Flash (Low)", upstreamModelId: "gemini-3.6-flash-tiered(low)" },
@@ -72,7 +75,7 @@ export default {
       "https://www.googleapis.com/auth/cclog",
       "https://www.googleapis.com/auth/experimentsandconfigs",
     ],
-    apiEndpoint: "https://daily-cloudcode-pa.googleapis.com",
+    apiEndpoint: "https://cloudcode-pa.googleapis.com",
     apiVersion: "v1internal",
     loadCodeAssistEndpoint: "https://cloudcode-pa.googleapis.com/v1internal:loadCodeAssist",
     onboardUserEndpoint: "https://cloudcode-pa.googleapis.com/v1internal:onboardUser",

open-sse/providers/registry/fish-audio.js

@@ -0,0 +1,31 @@
+// Fish Audio TTS — the model id travels in an HTTP `model` header rather than the
+// JSON body, and the voice is a reference_id (a cloned or preset voice model).
+export default {
+  id: "fish-audio",
+  alias: "fish",
+  display: {
+    name: "Fish Audio",
+    icon: "record_voice_over",
+    color: "#1E9BF0",
+    textIcon: "FA",
+    website: "https://fish.audio",
+    notice: {
+      apiKeyUrl: "https://fish.audio/app/api-keys/",
+    },
+  },
+  category: "apikey",
+  authType: "apikey",
+  serviceKinds: ["tts"],
+  ttsConfig: {
+    baseUrl: "https://api.fish.audio/v1/tts",
+    authType: "apikey",
+    authHeader: "bearer",
+    format: "fish-audio",
+    models: [
+      { id: "s2.1-pro-free", name: "S2.1 Pro Free" },
+      { id: "s2.1-pro", name: "S2.1 Pro" },
+      { id: "s2-pro", name: "S2 Pro" },
+      { id: "s1", name: "S1" },
+    ],
+  },
+};

open-sse/providers/registry/gemini.js

@@ -36,6 +36,7 @@ export default {
     },
   },
   models: [
+    { id: "gemini-3.7-flash", name: "Gemini 3.7 Flash" },
     { id: "gemini-3.6-flash", name: "Gemini 3.6 Flash" },
     { id: "gemini-3.5-flash-lite", name: "Gemini 3.5 Flash Lite" },
     { id: "gemini-3.1-pro-preview", name: "Gemini 3.1 Pro Preview" },

open-sse/providers/registry/glm-cn.js

@@ -21,6 +21,7 @@ export default {
     },
   },
   models: [
+    { id: "glm-5.3", name: "GLM 5.3" },
     { id: "glm-5.2", name: "GLM 5.2" },
     { id: "glm-5.1", name: "GLM 5.1" },
     { id: "glm-5", name: "GLM 5" },

open-sse/providers/registry/glm.js

@@ -45,6 +45,7 @@ export default {
     },
   ],
   models: [
+    { id: "glm-5.3", name: "GLM 5.3" },
     { id: "glm-5.2", name: "GLM 5.2" },
     { id: "glm-5.1", name: "GLM 5.1" },
     { id: "glm-5", name: "GLM 5" },

open-sse/providers/registry/index.js

@@ -119,6 +119,8 @@ import p116 from "./tokenrouter.js";
 import p117 from "./selfhosted-stt.js";
 import p118 from "./selfhosted-tts.js";
 import p119 from "./selfhosted-embedding.js";
+import p120 from "./fish-audio.js";
+import p121 from "./alitp-intl.js";
 
 export default [
   p0,
@@ -239,4 +241,6 @@ export default [
   p117,
   p118,
   p119,
+  p120,
+  p121,
 ];

open-sse/providers/registry/kimchi.js

@@ -14,7 +14,7 @@ export default {
     },
   },
   category: "freeTier",
-  authModes: ["oauth"],
+  authModes: ["oauth", "apikey"],
   hasOAuth: true,
   transport: {
     baseUrl: "https://llm.kimchi.dev/openai/v1/chat/completions",

open-sse/providers/registry/opencode-go.js

@@ -22,20 +22,28 @@ export default {
     baseUrl: "https://opencode.ai/zen/go/v1/chat/completions",
     headers: {},
   },
+  // Multi-endpoint: pick the transport matching the client sourceFormat to skip
+  // translation. Guarded per-model by `supportedFormats` (see chatCore) because
+  // opencode-go models differ in endpoint support.
+  transports: [
+    { format: "openai", baseUrl: "https://opencode.ai/zen/go/v1/chat/completions", auth: { combined: true, header: "Authorization", scheme: "bearer" } },
+    { format: "claude", baseUrl: "https://opencode.ai/zen/go/v1/messages", auth: { combined: true, header: "x-api-key", scheme: "raw", anthropicVersion: true } },
+    { format: "openai-responses", baseUrl: "https://opencode.ai/zen/go/v1/responses", auth: { combined: true, header: "Authorization", scheme: "bearer" } },
+  ],
   models: [
-    { id: "glm-5.2", name: "GLM 5.2" },
-    { id: "glm-5.1", name: "GLM 5.1" },
-    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code" },
-    { id: "kimi-k2.6", name: "Kimi K2.6" },
-    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro" },
-    { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash" },
-    { id: "mimo-v2.5", name: "MiMo V2.5" },
-    { id: "mimo-v2.5-pro", name: "MiMo V2.5 Pro" },
-    { id: "minimax-m3", name: "MiniMax M3", targetFormat: "claude" },
-    { id: "minimax-m2.7", name: "MiniMax M2.7", targetFormat: "claude" },
-    { id: "minimax-m2.5", name: "MiniMax M2.5", targetFormat: "claude" },
-    { id: "qwen3.7-max", name: "Qwen 3.7 Max", targetFormat: "claude" },
-    { id: "qwen3.7-plus", name: "Qwen 3.7 Plus", targetFormat: "claude" },
-    { id: "qwen3.6-plus", name: "Qwen 3.6 Plus", targetFormat: "claude" },
+    { id: "glm-5.2", name: "GLM 5.2", supportedFormats: ["openai"] },
+    { id: "glm-5.1", name: "GLM 5.1", supportedFormats: ["openai"] },
+    { id: "kimi-k2.7-code", name: "Kimi K2.7 Code", supportedFormats: ["openai"] },
+    { id: "kimi-k2.6", name: "Kimi K2.6", supportedFormats: ["openai"] },
+    { id: "deepseek-v4-pro", name: "DeepSeek V4 Pro", supportedFormats: ["openai", "claude", "openai-responses"] },
+    { id: "deepseek-v4-flash", name: "DeepSeek V4 Flash", supportedFormats: ["openai", "claude", "openai-responses"] },
+    { id: "mimo-v2.5", name: "MiMo V2.5", supportedFormats: ["openai"] },
+    { id: "mimo-v2.5-pro", name: "MiMo V2.5 Pro", supportedFormats: ["openai"] },
+    { id: "minimax-m3", name: "MiniMax M3", supportedFormats: ["openai", "claude"] },
+    { id: "minimax-m2.7", name: "MiniMax M2.7", supportedFormats: ["openai", "claude"] },
+    { id: "minimax-m2.5", name: "MiniMax M2.5", supportedFormats: ["openai", "claude"] },
+    { id: "qwen3.7-max", name: "Qwen 3.7 Max", supportedFormats: ["openai", "claude"] },
+    { id: "qwen3.7-plus", name: "Qwen 3.7 Plus", supportedFormats: ["openai", "claude"] },
+    { id: "qwen3.6-plus", name: "Qwen 3.6 Plus", supportedFormats: ["openai", "claude"] },
   ],
 };

open-sse/providers/shared.js

@@ -78,7 +78,7 @@ export const ANTHROPIC_COMPAT_BASE = "https://api.anthropic.com/v1";
 // Keep this static even when 9router runs on Linux: the provider profile is
 // intentionally matching the IDE client, not the server host.
 export const ANTIGRAVITY_IDE_VERSION = "2.1.1";
-export const ANTIGRAVITY_IDE_BASE_URL = "https://daily-cloudcode-pa.googleapis.com";
+export const ANTIGRAVITY_IDE_BASE_URL = "https://cloudcode-pa.googleapis.com";
 export const ANTIGRAVITY_IDE_USER_AGENT = `antigravity/ide/${ANTIGRAVITY_IDE_VERSION} darwin/arm64`;
 
 // Antigravity OAuth client credentials (public CLI client — duplicated in usage.js + src/lib/oauth)

open-sse/services/combo.js

@@ -138,8 +138,42 @@ export function detectRequiredCapabilities(body) {
     if (Array.isArray(content)) for (const b of content) scanBlock(b);
   };
 
+  const scanMessage = (m) => {
+    if (!m || typeof m !== "object") return;
+
+    // Ollama / Hermes images array (strings or objects)
+    if (Array.isArray(m.images) && m.images.length > 0) {
+      required.add("vision");
+    }
+
+    // Vercel AI SDK / Hermes attachments / experimental_attachments
+    const attachments = m.experimental_attachments || m.attachments;
+    if (Array.isArray(attachments)) {
+      for (const att of attachments) {
+        if (!att) continue;
+        const mime = att.contentType || att.mediaType || (typeof att.url === "string" && att.url.match(/^data:([^;,]+)/)?.[1]);
+        if (mime) addByMime(mime);
+        else if (att.url || att.data) required.add("vision");
+      }
+    }
+
+    // Direct message-level modality properties
+    if (m.image_url || m.image) required.add("vision");
+    if (m.audio_url || m.audio) required.add("audioInput");
+
+    // Scan array content blocks
+    scanContent(m.content);
+
+    // Scan string content for embedded data URIs
+    if (typeof m.content === "string") {
+      if (m.content.includes("data:image/")) required.add("vision");
+      else if (m.content.includes("data:audio/")) required.add("audioInput");
+      else if (m.content.includes("data:application/pdf")) required.add("pdf");
+    }
+  };
+
   // Modalities: current user turn only (trailing user run across each known shape).
-  for (const m of trailingUserItems(body.messages)) scanContent(m.content);      // openai / claude
+  for (const m of trailingUserItems(body.messages)) scanMessage(m);              // openai / claude / hermes / ollama
   for (const it of trailingUserItems(body.input)) scanContent(it.content);       // responses
   const contents = body.contents || body.request?.contents;                      // gemini / antigravity
   for (const c of trailingUserItems(contents)) scanContent(c.parts);
@@ -530,7 +564,10 @@ export async function handleFusionChat({ body, models, handleSingleModel, log, c
   log.info("FUSION", `Combo "${comboName}" | panel=${panel.length} [${panel.join(", ")}] | judge=${judge} | quorum=${minPanel}`);
 
   // 1. Fan out to the panel in parallel: non-streaming, tools stripped (we want prose).
-  const { tools, tool_choice, ...rest } = body;
+  const { tools, tool_choice, stream_options, ...rest } = body;
+  // Fusion runs panel models non-streaming; drop stream_options too, or providers
+  // like DeepSeek reject it with "stream_options should be set along with stream = true".
+  // See issue #3024.
   const panelBody = { ...rest, stream: false };
 
   // Flatten tool turns to prose so panel models keep context without emitting tool_calls.

open-sse/services/usage.js

@@ -33,7 +33,7 @@ const USAGE_HANDLERS = {
   github: (c) => getGitHubUsage(c.accessToken, c.providerSpecificData, c.proxyOptions),
   "gemini-cli": (c) => getGeminiUsage(c.accessToken, c.providerDataWithProjectId, c.proxyOptions),
   antigravity: (c) => getAntigravityUsage(c.accessToken, c.providerSpecificData, c.proxyOptions),
-  claude: (c) => getClaudeUsage(c.accessToken, c.proxyOptions),
+  claude: (c) => getClaudeUsage(c.accessToken, c.proxyOptions, { force: c.force }),
   codex: (c) => getCodexUsage(c.accessToken, c.proxyOptions),
   kiro: (c) => getKiroUsage(c.accessToken, c.providerSpecificData, c.proxyOptions),
   qoder: async (c) => {
@@ -56,7 +56,7 @@ const USAGE_HANDLERS = {
   deepseek: (c) => getDeepseekUsage(c.apiKey, c.proxyOptions),
 };
 
-export async function getUsageForProvider(connection, proxyOptions = null) {
+export async function getUsageForProvider(connection, proxyOptions = null, options = {}) {
   const { provider, accessToken, apiKey, providerSpecificData, projectId } = connection;
   const providerDataWithProjectId = {
     ...(providerSpecificData || {}),
@@ -65,5 +65,13 @@ export async function getUsageForProvider(connection, proxyOptions = null) {
 
   const handler = USAGE_HANDLERS[provider];
   if (!handler) return { message: `Usage API not implemented for ${provider}` };
-  return await handler({ provider, accessToken, apiKey, providerSpecificData, providerDataWithProjectId, proxyOptions });
+  return await handler({
+    provider,
+    accessToken,
+    apiKey,
+    providerSpecificData,
+    providerDataWithProjectId,
+    proxyOptions,
+    force: options.force === true,
+  });
 }

open-sse/services/usage/claude.js

@@ -19,7 +19,43 @@ const CLAUDE_CONFIG = {
 const OAUTH_429_COOLDOWN_MS = 180000;
 const oauthCooldown = new Map();
 
-export async function getClaudeUsage(accessToken, proxyOptions = null) {
+// Dedup + short TTL cache per access token. Many tabs / many accounts / auto-refresh
+// all funnel through here; without this each call hits Anthropic and triggers 429.
+const USAGE_CACHE_TTL_MS = 300000;
+const usageCache = new Map(); // token -> { promise } | { result, expiresAt }
+
+export async function getClaudeUsage(accessToken, proxyOptions = null, options = {}) {
+  const force = options?.force === true;
+
+  // Serve in-flight or fresh cached result (skip on manual force)
+  if (!force && accessToken) {
+    const hit = usageCache.get(accessToken);
+    if (hit?.promise) return hit.promise;
+    if (hit && hit.expiresAt > Date.now()) return hit.result;
+  }
+
+  const stale = (!force && accessToken && usageCache.get(accessToken)?.result) || null;
+
+  const promise = (async () => {
+    const result = await fetchClaudeUsageRaw(accessToken, proxyOptions);
+    // Only cache real quota data, not soft-failure {message: ...} payloads
+    if (accessToken && result?.quotas) {
+      usageCache.set(accessToken, {
+        result,
+        expiresAt: Date.now() + USAGE_CACHE_TTL_MS,
+      });
+      return result;
+    }
+    // Soft failure (429/error): prefer the last good read over a transient error
+    if (stale) return stale;
+    return result;
+  })();
+
+  if (accessToken) usageCache.set(accessToken, { promise });
+  return promise;
+}
+
+async function fetchClaudeUsageRaw(accessToken, proxyOptions = null) {
   try {
     // Skip OAuth usage call while this token is cooling down from a recent 429
     const cooldownUntil = oauthCooldown.get(accessToken);

open-sse/services/usage/google.js

@@ -161,6 +161,9 @@ export async function getAntigravityUsage(accessToken, providerSpecificData, pro
     if (data.models) {
       // Filter only recommended/important models (must match PROVIDER_MODELS ag ids)
       const importantModels = [
+        'gemini-3.7-flash-high',
+        'gemini-3.7-flash-medium',
+        'gemini-3.7-flash-low',
         'gemini-3.6-flash-high',
         'gemini-3.6-flash-medium',
         'gemini-3.6-flash-low',

open-sse/translator/concerns/modality.js

@@ -62,6 +62,19 @@ function stripOpenAI(body, caps) {
   if (!Array.isArray(body.messages)) return;
   const last = body.messages.length - 1;
   body.messages.forEach((msg, i) => {
+    if (caps.vision === false) {
+      if (Array.isArray(msg.images)) delete msg.images;
+      if (Array.isArray(msg.experimental_attachments)) {
+        msg.experimental_attachments = msg.experimental_attachments.filter(
+          (a) => !(a?.contentType?.startsWith("image/") || (typeof a?.url === "string" && a.url.startsWith("data:image/")))
+        );
+      }
+      if (Array.isArray(msg.attachments)) {
+        msg.attachments = msg.attachments.filter(
+          (a) => !(a?.contentType?.startsWith("image/") || (typeof a?.url === "string" && a.url.startsWith("data:image/")))
+        );
+      }
+    }
     if (!Array.isArray(msg.content)) return;
     const removed = new Set();
     msg.content = filterBlocks(msg.content, capForOpenAIBlock, caps, removed, i === last);

open-sse/translator/formats/claude.js

@@ -9,6 +9,9 @@ import { PROVIDERS } from "../../providers/index.js";
 import { getCapabilitiesForModel } from "../../providers/capabilities.js";
 import { DEFAULT_MAX_TOKENS } from "../../config/runtimeConfig.js";
 
+const CACHE_CONTROL_5M = { type: "ephemeral" };
+const CACHE_CONTROL_1H = { type: "ephemeral", ttl: "1h" };
+
 // Check if message has valid non-empty content
 export function hasValidContent(msg) {
   if (typeof msg.content === "string" && msg.content.trim()) return true;
@@ -124,32 +127,38 @@ export function normalizeClaudePassthrough(body, model = "") {
     if (Object.keys(body.output_config).length === 0) delete body.output_config;
   }
 
-  // 2. Hoist mid-conversation system messages into the top-level system field
+  // 2. Fold mid-conversation system messages into the neighbouring turn.
+  // Hoisting them into body.system would insert volatile content (token counters,
+  // reminders) ahead of the whole conversation and invalidate the prefix cache on
+  // every request. Folding in place keeps the cached prefix stable.
   if (Array.isArray(body.messages)) {
-    const systemBlocks = [];
     const messages = [];
     for (const msg of body.messages) {
-      if (msg.role === ROLE.SYSTEM) {
-        const text = typeof msg.content === "string"
-          ? msg.content
-          : Array.isArray(msg.content)
-            ? msg.content.map(b => (typeof b === "string" ? b : b?.text || "")).join("\n")
-            : "";
-        if (text.trim()) systemBlocks.push({ type: CLAUDE_BLOCK.TEXT, text });
+      if (msg.role !== ROLE.SYSTEM) {
+        messages.push(msg);
         continue;
       }
-      messages.push(msg);
-    }
+      const text = typeof msg.content === "string"
+        ? msg.content
+        : Array.isArray(msg.content)
+          ? msg.content.map(b => (typeof b === "string" ? b : b?.text || "")).join("\n")
+          : "";
+      if (!text.trim()) continue;
 
-    if (systemBlocks.length > 0) {
-      const existing = Array.isArray(body.system)
-        ? body.system
-        : typeof body.system === "string" && body.system.trim()
-          ? [{ type: "text", text: body.system }]
-          : [];
-      body.system = [...existing, ...systemBlocks];
-      body.messages = messages;
+      // Copy-on-write: the caller's body is reused across account-fallback
+      // attempts, so folding must never mutate the original message.
+      const block = { type: CLAUDE_BLOCK.TEXT, text };
+      const prev = messages[messages.length - 1];
+      if (prev?.role === ROLE.USER) {
+        const content = typeof prev.content === "string"
+          ? [{ type: CLAUDE_BLOCK.TEXT, text: prev.content }]
+          : Array.isArray(prev.content) ? [...prev.content] : [];
+        messages[messages.length - 1] = { ...prev, content: [...content, block] };
+        continue;
+      }
+      messages.push({ role: ROLE.USER, content: [block] });
     }
+    body.messages = messages;
   }
 
   // 3. Drop thinking blocks whose signature is not Claude's (combo mixes models,
@@ -182,6 +191,70 @@ export function normalizeClaudePassthrough(body, model = "") {
   return body;
 }
 
+// Put a 5m breakpoint on the last cache-eligible block of a message.
+// thinking/redacted_thinking blocks do not accept cache_control.
+function markLastCacheableBlock(msg) {
+  if (!Array.isArray(msg?.content)) return false;
+  for (let i = msg.content.length - 1; i >= 0; i--) {
+    const block = msg.content[i];
+    if (typeof block !== "object" || block === null) continue;
+    if (block.type === CLAUDE_BLOCK.THINKING || block.type === CLAUDE_BLOCK.REDACTED_THINKING) continue;
+    block.cache_control = { ...CACHE_CONTROL_5M };
+    return true;
+  }
+  return false;
+}
+
+// Re-anchor cache breakpoints on a Claude passthrough body (same policy as
+// prepareClaudeRequest): last tool + last system block at 1h, last assistant at 5m.
+// The client's own markers point at pre-normalization offsets, so they are dropped.
+// Must run LAST, after every step that can reshape system/tools/messages
+// (normalize, tool dedupe, token savers) — otherwise the anchor drifts off the tail.
+export function anchorClaudeCache(body) {
+  if (!body || typeof body !== "object") return body;
+
+  if (Array.isArray(body.system)) {
+    const last = body.system.length - 1;
+    body.system.forEach((block, i) => {
+      if (typeof block !== "object" || block === null) return;
+      if (i === last) block.cache_control = { ...CACHE_CONTROL_1H };
+      else delete block.cache_control;
+    });
+  }
+
+  if (Array.isArray(body.tools)) {
+    const last = body.tools.length - 1;
+    body.tools.forEach((tool, i) => {
+      if (i === last) tool.cache_control = { ...CACHE_CONTROL_1H };
+      else delete tool.cache_control;
+    });
+  }
+
+  if (Array.isArray(body.messages)) {
+    let anchored = null;
+    for (let i = body.messages.length - 1; i >= 0; i--) {
+      const msg = body.messages[i];
+      if (!Array.isArray(msg.content)) continue;
+      for (const block of msg.content) delete block.cache_control;
+
+      // Prefer the last assistant turn: it ends a completed exchange, so the
+      // prefix up to it stays byte-stable across the following requests.
+      if (anchored || msg.role !== ROLE.ASSISTANT) continue;
+      anchored = markLastCacheableBlock(msg);
+    }
+
+    // First turn of a conversation has no assistant yet — anchor the final
+    // message instead, so the opening prompt is cached rather than paid twice.
+    if (!anchored) {
+      for (let i = body.messages.length - 1; i >= 0 && !anchored; i--) {
+        anchored = markLastCacheableBlock(body.messages[i]);
+      }
+    }
+  }
+
+  return body;
+}
+
 // Prepare request for Claude format endpoints
 // - Cleanup cache_control
 // - Filter empty messages

open-sse/translator/formats/gemini.js

@@ -311,6 +311,26 @@ function ensureObjectType(obj) {
 // Clean JSON Schema for Antigravity API compatibility - removes unsupported keywords recursively
 export function cleanJSONSchemaForAntigravity(schema) {
   if (!schema || typeof schema !== "object") return schema;
+  const defs = schema.$defs || schema.definitions || {};
+  function deref(obj) {
+    if (!obj || typeof obj !== "object") return obj;
+    if (Array.isArray(obj)) return obj.map(deref);
+    if (obj.$ref && typeof obj.$ref === "string") {
+      const name = obj.$ref.split("/").pop();
+      if (defs[name]) return deref(Object.assign({}, defs[name]));
+    }
+    const res = {};
+    for (const [k, v] of Object.entries(obj)) {
+      if (k === "$defs" || k === "definitions") continue;
+      res[k] = deref(v);
+    }
+    return res;
+  }
+  return _orig_cleanJSONSchemaForAntigravity(deref(schema));
+}
+
+function _orig_cleanJSONSchemaForAntigravity(schema) {
+  if (!schema || typeof schema !== "object") return schema;
 
   // Mutate directly (schema is only used once per request)
   let cleaned = schema;

open-sse/translator/request/claude-to-kiro.js

@@ -287,6 +287,18 @@ export function claudeToKiroRequest(model, body, stream, credentials) {
     toolSpecs,
     nameMap,
   });
+  // canonicalizeKiroConversation() already ran its second-chance repair (flatten
+  // every structured tool turn to text, then re-validate). A body that is STILL
+  // invalid here cannot be made shippable, and Kiro answers it with
+  // 400 {"message":"Improperly formed request.","reason":"REQUEST_BODY_INVALID"}.
+  // Fail locally instead: chatCore turns a falsy return into a 400 without
+  // spending an upstream call or a per-account cooldown. The taxonomy
+  // (role:N | pair:N | id:N | spec:N | orphan:0 | current) names the offending
+  // turn so the shape can be diagnosed from the log alone.
+  if (!canonical.valid) {
+    console.error(`[Kiro] refusing invalid conversation (claude → kiro): ${(canonical.errors || []).join(", ") || "unknown"} | turns=${(canonical.history || []).length + 1}`);
+    return null;
+  }
   const replayCurrent = canonical.currentMessage.userInputMessage;
   const userInputMessage = {
     content: replayCurrent.content || "",

open-sse/translator/request/openai-responses.js

@@ -421,6 +421,7 @@ export function openaiToOpenAIResponsesRequest(model, body, stream, credentials)
   if (body.reasoning !== undefined) result.reasoning = body.reasoning;
   if (body.reasoning_effort !== undefined) result.reasoning = { effort: body.reasoning_effort, summary: "auto" };
   if (body.service_tier !== undefined) result.service_tier = body.service_tier;
+  if (body.prompt_cache_key !== undefined) result.prompt_cache_key = body.prompt_cache_key;
 
   return result;
 }

open-sse/translator/request/openai-to-kiro.js

@@ -379,6 +379,18 @@ export function openaiToKiroRequest(model, body, stream, credentials) {
     toolSpecs,
     nameMap,
   });
+  // canonicalizeKiroConversation() already ran its second-chance repair (flatten
+  // every structured tool turn to text, then re-validate). A body that is STILL
+  // invalid here cannot be made shippable, and Kiro answers it with
+  // 400 {"message":"Improperly formed request.","reason":"REQUEST_BODY_INVALID"}.
+  // Fail locally instead: chatCore turns a falsy return into a 400 without
+  // spending an upstream call or a per-account cooldown. The taxonomy
+  // (role:N | pair:N | id:N | spec:N | orphan:0 | current) names the offending
+  // turn so the shape can be diagnosed from the log alone.
+  if (!canonical.valid) {
+    console.error(`[Kiro] refusing invalid conversation (openai → kiro): ${(canonical.errors || []).join(", ") || "unknown"} | turns=${(canonical.history || []).length + 1}`);
+    return null;
+  }
   const replayCurrent = canonical.currentMessage.userInputMessage;
 
   const payload = {

open-sse/translator/response/kiro-to-claude.js

@@ -75,6 +75,15 @@ export function kiroToClaudeResponse(chunk, state) {
         ? data.usage.completion_tokens
         : 0;
     state.usage = { input_tokens: promptTokens, output_tokens: outputTokens };
+    // Claude clients read cache_read/cache_creation to price a turn and to size
+    // their prompt cache. Both spellings are accepted because the Kiro executor
+    // emits the Chat shape and passthrough responses use the nested details form.
+    const cacheRead = data.usage.cache_read_input_tokens
+      ?? data.usage.prompt_tokens_details?.cached_tokens;
+    const cacheCreation = data.usage.cache_creation_input_tokens
+      ?? data.usage.prompt_tokens_details?.cache_creation_tokens;
+    if (typeof cacheRead === "number") state.usage.cache_read_input_tokens = cacheRead;
+    if (typeof cacheCreation === "number") state.usage.cache_creation_input_tokens = cacheCreation;
   }
 
   // First chunk → emit message_start.
@@ -254,6 +263,13 @@ export function kiroToClaudeNonStreaming(data) {
     usage: {
       input_tokens: usage.prompt_tokens || 0,
       output_tokens: usage.completion_tokens || 0,
+      // Same cache preservation as the streaming path above.
+      ...(typeof (usage.cache_read_input_tokens ?? usage.prompt_tokens_details?.cached_tokens) === "number"
+        ? { cache_read_input_tokens: usage.cache_read_input_tokens ?? usage.prompt_tokens_details.cached_tokens }
+        : {}),
+      ...(typeof (usage.cache_creation_input_tokens ?? usage.prompt_tokens_details?.cache_creation_tokens) === "number"
+        ? { cache_creation_input_tokens: usage.cache_creation_input_tokens ?? usage.prompt_tokens_details.cache_creation_tokens }
+        : {}),
     },
   };
 }

open-sse/translator/response/openai-responses.js

@@ -99,8 +99,8 @@ export function openaiToOpenAIResponsesResponse(chunk, state) {
     }
   }
 
-  // Handle tool_calls
-  if (delta.tool_calls) {
+  // Handle tool_calls (empty array is truthy; require a real call)
+  if (delta.tool_calls && delta.tool_calls.length) {
     closeMessage(state, emit, idx);
     for (const tc of delta.tool_calls) {
       emitToolCall(state, emit, tc);

package.json

@@ -1,6 +1,6 @@
 {
   "name": "9router-app",
-  "version": "0.5.50",
+  "version": "0.5.55",
   "description": "9Router web dashboard",
   "private": true,
   "scripts": {
@@ -9,10 +9,10 @@
     "build": "next build --webpack",
     "postbuild": "node scripts/copy-standalone-assets.mjs",
     "postbuild:bun": "node scripts/copy-standalone-assets.mjs",
-    "start": "next start --port 20127",
+    "start": "node custom-server.js --port 20127",
     "dev:bun": "bun --bun next dev --webpack --port 20127",
     "build:bun": "bun --bun next build --webpack",
-    "start:bun": "bun ./.next/standalone/server.js",
+    "start:bun": "bun ./.next/standalone/custom-server.js",
     "cli:pack": "npm --prefix cli run pack:cli",
     "cli:publish": "npm --prefix cli run publish:cli"
   },
@@ -23,8 +23,10 @@
     "@dnd-kit/utilities": "^3.2.2",
     "@monaco-editor/react": "^4.7.0",
     "@next/third-parties": "^16.2.9",
+    "@node-saml/node-saml": "^5.1.0",
     "@xyflow/react": "^12.10.1",
     "bcryptjs": "^3.0.3",
+    "chalk": "^5.6.2",
     "confbox": "^0.2.4",
     "express": "^5.2.1",
     "http-proxy-middleware": "^3.0.5",
@@ -37,6 +39,7 @@
     "node-machine-id": "^1.1.12",
     "open": "^11.0.0",
     "ora": "^9.1.0",
+    "prop-types": "^15.8.1",
     "react": "19.2.4",
     "react-dom": "19.2.4",
     "react-is": "^16.13.1",

scripts/copy-standalone-assets.mjs

@@ -29,6 +29,14 @@ export function copyStandaloneAssets({ projectRoot = process.cwd(), distDir = pr
     cpSync(publicSource, publicDestination, { recursive: true, force: true });
     console.log(`[standalone-assets] Copied public assets to ${publicDestination}`);
   }
+
+  // Without it beside server.js the standalone build serves requests unsanitized.
+  const serverWrapperSource = resolve(projectRoot, "custom-server.js");
+  const serverWrapperDestination = resolve(standaloneDir, "custom-server.js");
+  if (existsSync(serverWrapperSource)) {
+    cpSync(serverWrapperSource, serverWrapperDestination, { force: true });
+    console.log(`[standalone-assets] Copied custom-server.js to ${serverWrapperDestination}`);
+  }
 }
 
 if (process.argv[1] && resolve(process.argv[1]) === resolve(dirname(fileURLToPath(import.meta.url)), "copy-standalone-assets.mjs")) {

src/app/(dashboard)/dashboard/cli-tools/components/HermesToolCard.js

@@ -170,7 +170,7 @@ export default function HermesToolCard({
       ? selectedApiKey
       : (!cloudEnabled ? "sk_9router" : "<API_KEY_FROM_DASHBOARD>");
 
-    const yamlContent = `model:\n  default: "${selectedModel || "provider/model-id"}"\n  provider: "custom"\n  base_url: "${getEffectiveBaseUrl()}"\n`;
+    const yamlContent = `model:\n  default: "${selectedModel || "provider/model-id"}"\n  provider: "custom"\n  base_url: "${getEffectiveBaseUrl()}"\n  api_key: \${OPENAI_API_KEY}\n`;
     const envContent = `OPENAI_API_KEY=${keyToUse}\n`;
 
     return [

src/app/(dashboard)/dashboard/profile/page.js

@@ -21,7 +21,7 @@ function getLocaleFromCookie() {
 
 export default function ProfilePage() {
   const { theme, setTheme, isDark } = useTheme();
-  const [locale, setLocale] = useState("en");
+  const [locale, setLocale] = useState(() => getLocaleFromCookie());
   const [langOpen, setLangOpen] = useState(false);
   const [shutdownOpen, setShutdownOpen] = useState(false);
   const [isShuttingDown, setIsShuttingDown] = useState(false);
@@ -46,8 +46,31 @@ export default function ProfilePage() {
   const [oidcLoading, setOidcLoading] = useState(false);
   const [oidcTestLoading, setOidcTestLoading] = useState(false);
   const [oidcTestStatus, setOidcTestStatus] = useState({ type: "", message: "" });
-  const [oidcRedirectUri, setOidcRedirectUri] = useState("/api/auth/oidc/callback");
   const [oidcExpanded, setOidcExpanded] = useState(false);
+
+  const origin = typeof window !== "undefined" ? window.location.origin : "";
+  const oidcRedirectUri = origin ? `${origin}/api/auth/oidc/callback` : "/api/auth/oidc/callback";
+  const samlAcsUrl = origin ? `${origin}/api/auth/saml/acs` : "/api/auth/saml/acs";
+  const samlMetadataUrl = origin ? `${origin}/api/auth/saml/metadata` : "/api/auth/saml/metadata";
+  
+  // SAML State
+  const [ssoTypeTab, setSsoTypeTab] = useState("saml");
+  const [samlForm, setSamlForm] = useState({
+    samlEntryPoint: "",
+    samlIssuer: "urn:9router:sp",
+    samlCert: "",
+    samlLoginLabel: "Sign in with SAML SSO",
+    samlAttributeEmail: "email",
+    samlAttributeName: "name",
+  });
+  const [samlStatus, setSamlStatus] = useState({ type: "", message: "" });
+  const [samlLoading, setSamlLoading] = useState(false);
+  const [samlTestLoading, setSamlTestLoading] = useState(false);
+  const [samlTestStatus, setSamlTestStatus] = useState({ type: "", message: "" });
+  const [showSamlGuide, setShowSamlGuide] = useState(false);
+  const idpMetadataFileRef = useRef(null);
+  const certFileRef = useRef(null);
+
   const importFileRef = useRef(null);
   const [proxyForm, setProxyForm] = useState({
     outboundProxyEnabled: false,
@@ -58,10 +81,6 @@ export default function ProfilePage() {
   const [proxyLoading, setProxyLoading] = useState(false);
   const [proxyTestLoading, setProxyTestLoading] = useState(false);
 
-  useEffect(() => {
-    setLocale(getLocaleFromCookie());
-  }, [langOpen]);
-
   useEffect(() => {
     fetch("/api/settings")
       .then((res) => res.json())
@@ -75,7 +94,23 @@ export default function ProfilePage() {
           oidcLoginLabel: data?.oidcLoginLabel || "Sign in with OIDC",
         });
         setOidcClientSecret("");
-        if (data?.authMode === "oidc" || data?.authMode === "both") setOidcExpanded(true);
+        setSsoTypeTab(data?.ssoType || "saml");
+        setSamlForm({
+          samlEntryPoint: data?.samlEntryPoint || "",
+          samlIssuer: data?.samlIssuer || "urn:9router:sp",
+          samlCert: data?.samlCert || "",
+          samlLoginLabel: data?.samlLoginLabel || "Sign in with SAML SSO",
+          samlAttributeEmail: data?.samlAttributeEmail || "email",
+          samlAttributeName: data?.samlAttributeName || "name",
+        });
+        if (
+          data?.authMode === "sso" ||
+          data?.authMode === "saml" ||
+          data?.authMode === "oidc" ||
+          data?.authMode === "both"
+        ) {
+          setOidcExpanded(true);
+        }
         setProxyForm({
           outboundProxyEnabled: data?.outboundProxyEnabled === true,
           outboundProxyUrl: data?.outboundProxyUrl || "",
@@ -89,12 +124,6 @@ export default function ProfilePage() {
       });
   }, []);
 
-  useEffect(() => {
-    if (typeof window !== "undefined") {
-      setOidcRedirectUri(`${window.location.origin}/api/auth/oidc/callback`);
-    }
-  }, []);
-
   const updateOutboundProxy = async (e) => {
     e.preventDefault();
     if (settings.outboundProxyEnabled !== true) return;
@@ -331,6 +360,7 @@ export default function ProfilePage() {
     try {
       const payload = {
         authMode,
+        ssoType: "oidc",
         oidcIssuerUrl: issuerUrl,
         oidcClientId: clientId,
         oidcScopes: scopes || "openid profile email",
@@ -445,6 +475,159 @@ export default function ProfilePage() {
     }
   };
 
+  const updateSamlForm = (field, value) => {
+    setSamlForm((prev) => ({ ...prev, [field]: value }));
+  };
+
+  const handleIdpMetadataUpload = (event) => {
+    const file = event.target.files?.[0];
+    if (idpMetadataFileRef.current) idpMetadataFileRef.current.value = "";
+    if (!file) return;
+
+    const reader = new FileReader();
+    reader.onload = (e) => {
+      try {
+        const xmlText = e.target?.result || "";
+        const parser = new DOMParser();
+        const doc = parser.parseFromString(xmlText, "text/xml");
+        const parserError = doc.querySelector("parsererror");
+        if (parserError) {
+          setSamlStatus({ type: "error", message: "Unable to parse valid SAML IdP metadata from XML file" });
+          return;
+        }
+
+        const entityID = doc.documentElement.getAttribute("entityID") || "";
+        const ssoNodes = Array.from(doc.querySelectorAll("SingleSignOnService, *|SingleSignOnService"));
+        let ssoUrl = "";
+        for (const node of ssoNodes) {
+          const binding = node.getAttribute("Binding") || "";
+          const location = node.getAttribute("Location") || "";
+          if (location) {
+            ssoUrl = location;
+            if (binding.includes("HTTP-Redirect")) break;
+          }
+        }
+
+        const certNodes = Array.from(doc.querySelectorAll("X509Certificate, *|X509Certificate"));
+        let certStr = "";
+        if (certNodes.length > 0) {
+          certStr = certNodes[0].textContent.trim();
+        }
+
+        setSamlForm((prev) => ({
+          ...prev,
+          samlEntryPoint: ssoUrl || prev.samlEntryPoint,
+          samlIssuer: prev.samlIssuer || "urn:9router:sp",
+          samlCert: certStr || prev.samlCert,
+        }));
+
+        setSamlStatus({
+          type: "success",
+          message: `IdP Metadata imported! (SSO URL: ${ssoUrl ? "found" : "not found"}, EntityID: ${entityID ? "found" : "not found"}, Cert: ${certStr ? "found" : "not found"})`,
+        });
+      } catch (err) {
+        setSamlStatus({ type: "error", message: "Error reading IdP Metadata XML file" });
+      }
+    };
+    reader.readAsText(file);
+  };
+
+  const handleCertFileUpload = (event) => {
+    const file = event.target.files?.[0];
+    if (certFileRef.current) certFileRef.current.value = "";
+    if (!file) return;
+
+    const reader = new FileReader();
+    reader.onload = (e) => {
+      const text = e.target?.result || "";
+      setSamlForm((prev) => ({ ...prev, samlCert: text.trim() }));
+      setSamlStatus({ type: "success", message: "Certificate file loaded into configuration." });
+    };
+    reader.readAsText(file);
+  };
+
+  const saveSamlSettings = async (targetAuthMode = oidcForm.authMode || "password") => {
+    setSamlLoading(true);
+    setSamlStatus({ type: "", message: "" });
+    setSamlTestStatus({ type: "", message: "" });
+
+    try {
+      const payload = {
+        authMode: targetAuthMode,
+        ssoType: "saml",
+        samlEntryPoint: samlForm.samlEntryPoint.trim(),
+        samlIssuer: samlForm.samlIssuer.trim() || "urn:9router:sp",
+        samlCert: samlForm.samlCert.trim(),
+        samlLoginLabel: samlForm.samlLoginLabel.trim() || "Sign in with SAML SSO",
+        samlAttributeEmail: samlForm.samlAttributeEmail.trim() || "email",
+        samlAttributeName: samlForm.samlAttributeName.trim() || "name",
+      };
+
+      const res = await fetch("/api/settings", {
+        method: "PATCH",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify(payload),
+      });
+
+      const data = await res.json();
+      if (res.ok) {
+        setSettings((prev) => ({ ...prev, ...data }));
+        setSamlForm({
+          samlEntryPoint: data?.samlEntryPoint || payload.samlEntryPoint,
+          samlIssuer: data?.samlIssuer || payload.samlIssuer,
+          samlCert: data?.samlCert || payload.samlCert,
+          samlLoginLabel: data?.samlLoginLabel || payload.samlLoginLabel,
+          samlAttributeEmail: data?.samlAttributeEmail || payload.samlAttributeEmail,
+          samlAttributeName: data?.samlAttributeName || payload.samlAttributeName,
+        });
+        setSamlStatus({
+          type: "success",
+          message:
+            targetAuthMode === "sso" || targetAuthMode === "saml"
+              ? "SAML SSO login enabled"
+              : targetAuthMode === "both"
+                ? "Password and SAML SSO login enabled"
+                : "SAML 2.0 settings saved",
+        });
+      } else {
+        setSamlStatus({ type: "error", message: data.error || "Failed to save SAML settings" });
+      }
+    } catch {
+      setSamlStatus({ type: "error", message: "An error occurred while saving SAML settings" });
+    } finally {
+      setSamlLoading(false);
+    }
+  };
+
+  const testSamlConnection = async () => {
+    setSamlTestLoading(true);
+    setSamlStatus({ type: "", message: "" });
+    setSamlTestStatus({ type: "", message: "" });
+
+    try {
+      const res = await fetch("/api/auth/saml/test", {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({
+          samlEntryPoint: samlForm.samlEntryPoint.trim(),
+          samlIssuer: samlForm.samlIssuer.trim(),
+          samlCert: samlForm.samlCert.trim(),
+        }),
+      });
+
+      const data = await res.json();
+      if (res.ok && data.ok) {
+        setSamlTestStatus({ type: "success", message: data.message || "SAML configuration verified!" });
+      } else {
+        setSamlTestStatus({ type: "error", message: data.error || "SAML configuration test failed" });
+      }
+    } catch {
+      setSamlTestStatus({ type: "error", message: "An error occurred while testing SAML configuration" });
+    } finally {
+      setSamlTestLoading(false);
+    }
+  };
+
   const updateObservabilityEnabled = async (enabled) => {
     try {
       const res = await fetch("/api/settings", {
@@ -752,7 +935,7 @@ export default function ProfilePage() {
           </div>
         </Card>
 
-        {/* OIDC */}
+        {/* Single Sign-On (SSO) */}
         <Card>
           <button
             type="button"
@@ -763,9 +946,13 @@ export default function ProfilePage() {
               <span className="material-symbols-outlined text-[20px]">lock_open</span>
             </div>
             <div className="flex-1 min-w-0">
-              <h3 className="text-base sm:text-lg font-semibold">OIDC Dashboard Login</h3>
+              <h3 className="text-base sm:text-lg font-semibold">Single Sign-On (SSO)</h3>
               <p className="text-xs text-text-muted">
-                {settings.authMode === "oidc" ? "OIDC active" : settings.authMode === "both" ? "Password + OIDC active" : "Optional SSO via Authentik/Keycloak/Google"}
+                {settings.authMode === "sso" || settings.authMode === "oidc" || settings.authMode === "saml"
+                  ? `${settings.ssoType === "saml" ? "SAML 2.0" : "OIDC"} SSO active`
+                  : settings.authMode === "both"
+                    ? `Password + ${settings.ssoType === "saml" ? "SAML 2.0" : "OIDC"} active`
+                    : "Optional SSO via Okta, Entra ID, Keycloak, or OIDC"}
               </p>
             </div>
             <span className="material-symbols-outlined text-text-muted shrink-0">
@@ -773,145 +960,472 @@ export default function ProfilePage() {
             </span>
           </button>
           {oidcExpanded && (
-          <div className="flex flex-col gap-4 mt-4">
-            <p className="text-xs sm:text-sm text-text-muted">
-              Use Authentik or any OIDC provider to sign in to the dashboard. You can enable password-only, OIDC-only, or both for the dashboard; model API access still uses API keys.
-            </p>
+            <div className="flex flex-col gap-4 mt-4">
+              <p className="text-xs sm:text-sm text-text-muted">
+                Configure enterprise Single Sign-On (SSO) for dashboard access using SAML 2.0 or OIDC.
+              </p>
 
-            <div className="flex flex-col gap-2">
-              <label className="font-medium text-sm sm:text-base">Auth Mode</label>
-              <div className="grid grid-cols-1 sm:grid-cols-3 gap-2">
-                {[
-                  {
-                    value: "password",
-                    title: "Password only",
-                    desc: "Keep the legacy password login.",
-                  },
-                  {
-                    value: "oidc",
-                    title: "OIDC only",
-                    desc: "Require OIDC for dashboard access.",
-                  },
-                  {
-                    value: "both",
-                    title: "Both",
-                    desc: "Allow either password or OIDC.",
-                  },
-                ].map((option) => {
-                  const active = oidcForm.authMode === option.value;
-                  return (
+              {/* SSO Protocol Switcher Tabs */}
+              <div className="flex flex-col gap-2">
+                <label className="font-medium text-sm sm:text-base">SSO Protocol</label>
+                <div className="flex p-1 rounded-lg bg-black/5 dark:bg-white/5 border border-border">
+                  <button
+                    type="button"
+                    onClick={() => setSsoTypeTab("saml")}
+                    className={cn(
+                      "flex-1 py-1.5 px-3 rounded-md font-medium text-xs sm:text-sm transition-all text-center",
+                      ssoTypeTab === "saml"
+                        ? "bg-white dark:bg-white/10 text-text-main shadow-sm"
+                        : "text-text-muted hover:text-text-main"
+                    )}
+                  >
+                    SAML 2.0
+                  </button>
+                  <button
+                    type="button"
+                    onClick={() => setSsoTypeTab("oidc")}
+                    className={cn(
+                      "flex-1 py-1.5 px-3 rounded-md font-medium text-xs sm:text-sm transition-all text-center",
+                      ssoTypeTab === "oidc"
+                        ? "bg-white dark:bg-white/10 text-text-main shadow-sm"
+                        : "text-text-muted hover:text-text-main"
+                    )}
+                  >
+                    OIDC
+                  </button>
+                </div>
+              </div>
+
+              {/* Auth Mode selection */}
+              <div className="flex flex-col gap-2">
+                <label className="font-medium text-sm sm:text-base">Auth Mode</label>
+                <div className="grid grid-cols-1 sm:grid-cols-3 gap-2">
+                  {[
+                    {
+                      value: "password",
+                      title: "Password only",
+                      desc: "Keep legacy password login.",
+                    },
+                    {
+                      value: "sso",
+                      title: `${ssoTypeTab === "saml" ? "SAML" : "OIDC"} only`,
+                      desc: "Require SSO for dashboard access.",
+                    },
+                    {
+                      value: "both",
+                      title: "Both",
+                      desc: "Allow password or SSO login.",
+                    },
+                  ].map((option) => {
+                    const currentMode = oidcForm.authMode;
+                    const active =
+                      option.value === "password"
+                        ? currentMode === "password"
+                        : option.value === "sso"
+                          ? currentMode === "sso" || currentMode === "saml" || currentMode === "oidc"
+                          : currentMode === "both";
+                    return (
+                      <button
+                        key={option.value}
+                        type="button"
+                        onClick={() => updateOidcForm("authMode", option.value)}
+                        className={cn(
+                          "text-left rounded-lg border p-3 transition-colors",
+                          active
+                            ? "border-primary bg-primary/5"
+                            : "border-border bg-bg hover:bg-black/5 dark:hover:bg-white/5"
+                        )}
+                        disabled={loading || oidcLoading || samlLoading}
+                      >
+                        <p className="font-medium text-sm sm:text-base">{option.title}</p>
+                        <p className="text-xs sm:text-sm text-text-muted mt-1">{option.desc}</p>
+                      </button>
+                    );
+                  })}
+                </div>
+              </div>
+
+              {ssoTypeTab === "saml" ? (
+                /* SAML Configuration Panel */
+                <div className="flex flex-col gap-4 pt-2 border-t border-border/50">
+                  {/* IdP Setup Guidelines Banner & Collapsible Drawer */}
+                  <div className="rounded-lg border border-border bg-bg/80 overflow-hidden">
                     <button
-                      key={option.value}
                       type="button"
-                      onClick={() => updateOidcForm("authMode", option.value)}
-                      className={cn(
-                        "text-left rounded-lg border p-3 transition-colors",
-                        active
-                          ? "border-primary bg-primary/5"
-                          : "border-border bg-bg hover:bg-black/5 dark:hover:bg-white/5"
-                      )}
-                      disabled={loading || oidcLoading}
+                      onClick={() => setShowSamlGuide((prev) => !prev)}
+                      className="w-full p-3 flex items-center justify-between gap-2 text-left hover:bg-surface/50 transition-colors"
                     >
-                      <p className="font-medium text-sm sm:text-base">{option.title}</p>
-                      <p className="text-xs sm:text-sm text-text-muted mt-1">{option.desc}</p>
+                      <div className="flex items-center gap-2">
+                        <span className="material-symbols-outlined text-primary text-lg">menu_book</span>
+                        <div>
+                          <p className="font-semibold text-xs sm:text-sm text-text-main">
+                            IdP Setup Guidelines & Provider Configuration Instructions
+                          </p>
+                          <p className="text-[11px] text-text-muted">
+                            Click to view setup steps for AWS IAM Identity Center, Okta, Entra ID, Keycloak, & Authentik
+                          </p>
+                        </div>
+                      </div>
+                      <span
+                        className="material-symbols-outlined text-text-muted transition-transform text-lg"
+                        style={{ transform: showSamlGuide ? "rotate(180deg)" : "none" }}
+                      >
+                        expand_more
+                      </span>
                     </button>
-                  );
-                })}
-              </div>
+
+                    {showSamlGuide && (
+                      <div className="p-4 border-t border-border bg-surface/30 text-xs text-text-main flex flex-col gap-3">
+                        <div className="p-2.5 rounded border border-primary/20 bg-primary/5 text-primary text-xs">
+                          <p className="font-semibold mb-1">🔑 Required Service Provider (SP) Values for your IdP Setup:</p>
+                          <ul className="list-disc pl-4 space-y-1 font-mono text-[11px]">
+                            <li>
+                              <b>Assertion Consumer Service (ACS) URL:</b>{" "}
+                              <code className="bg-bg px-1 py-0.5 rounded break-all">{samlAcsUrl}</code>
+                            </li>
+                            <li>
+                              <b>SP Entity ID / Audience URI:</b>{" "}
+                              <code className="bg-bg px-1 py-0.5 rounded break-all">{samlForm.samlIssuer || "urn:9router:sp"}</code>
+                            </li>
+                            <li>
+                              <b>NameID Format:</b>{" "}
+                              <code className="bg-bg px-1 py-0.5 rounded">EmailAddress</code> or <code className="bg-bg px-1 py-0.5 rounded">Unspecified</code>
+                            </li>
+                          </ul>
+                        </div>
+
+                        <div className="grid grid-cols-1 md:grid-cols-2 gap-3 pt-1">
+                          <div className="p-3 rounded border border-border bg-bg/50 flex flex-col gap-1.5">
+                            <p className="font-semibold text-text-main flex items-center gap-1.5">
+                              <span>☁️</span> AWS IAM Identity Center
+                            </p>
+                            <ol className="list-decimal pl-4 text-text-muted space-y-1">
+                              <li>Applications → <b>Add application</b> → Select <b>Add custom SAML 2.0 application</b>.</li>
+                              <li>Set <b>Application ACS URL</b> to <code className="text-text-main font-mono">{samlAcsUrl}</code>.</li>
+                              <li>Set <b>Application SAML audience</b> to <code className="text-text-main font-mono">{samlForm.samlIssuer || "urn:9router:sp"}</code>.</li>
+                              <li>Under <i>Attribute mappings</i>, map <code className="text-text-main font-mono">Subject</code> or <code className="text-text-main font-mono">email</code> to <code className="text-text-main font-mono">${`{user:email}`}</code>.</li>
+                              <li>Download <b>IAM Identity Center SAML metadata XML</b> file and use 1-Click Import below!</li>
+                            </ol>
+                          </div>
+
+                          <div className="p-3 rounded border border-border bg-bg/50 flex flex-col gap-1.5">
+                            <p className="font-semibold text-text-main flex items-center gap-1.5">
+                              <span>🔷</span> Microsoft Entra ID (Azure AD)
+                            </p>
+                            <ol className="list-decimal pl-4 text-text-muted space-y-1">
+                              <li>Enterprise Applications → <b>New application</b> → <b>Create your own application</b>.</li>
+                              <li>Select <b>Single sign-on</b> → <b>SAML</b>.</li>
+                              <li><b>Identifier (Entity ID):</b> <code className="text-text-main font-mono">{samlForm.samlIssuer || "urn:9router:sp"}</code></li>
+                              <li><b>Reply URL (ACS):</b> <code className="text-text-main font-mono">{samlAcsUrl}</code></li>
+                              <li>Download <b>Federation Metadata XML</b> and import or copy X.509 Certificate.</li>
+                            </ol>
+                          </div>
+
+                          <div className="p-3 rounded border border-border bg-bg/50 flex flex-col gap-1.5">
+                            <p className="font-semibold text-text-main flex items-center gap-1.5">
+                              <span>🟢</span> Okta / Auth0
+                            </p>
+                            <ol className="list-decimal pl-4 text-text-muted space-y-1">
+                              <li>Applications → <b>Create App Integration</b> → Select <b>SAML 2.0</b>.</li>
+                              <li><b>Single Sign-On URL:</b> <code className="text-text-main font-mono">{samlAcsUrl}</code></li>
+                              <li><b>Audience URI (SP Entity ID):</b> <code className="text-text-main font-mono">{samlForm.samlIssuer || "urn:9router:sp"}</code></li>
+                              <li>Name ID format: <i>EmailAddress</i>.</li>
+                              <li>Download Identity Provider metadata XML or copy the X.509 cert.</li>
+                            </ol>
+                          </div>
+
+                          <div className="p-3 rounded border border-border bg-bg/50 flex flex-col gap-1.5">
+                            <p className="font-semibold text-text-main flex items-center gap-1.5">
+                              <span>🛡️</span> Keycloak / Authentik
+                            </p>
+                            <ol className="list-decimal pl-4 text-text-muted space-y-1">
+                              <li>Clients → <b>Create client</b> → Select <b>SAML</b>.</li>
+                              <li><b>Client ID:</b> <code className="text-text-main font-mono">{samlForm.samlIssuer || "urn:9router:sp"}</code></li>
+                              <li><b>Master SAML Processing URL:</b> <code className="text-text-main font-mono">{samlAcsUrl}</code></li>
+                              <li>Export SAML Descriptor XML or copy IDP Certificate PEM.</li>
+                            </ol>
+                          </div>
+                        </div>
+                      </div>
+                    )}
+                  </div>
+
+                  {/* Quick Import Card */}
+                  <div className="p-3 rounded-lg border border-dashed border-primary/40 bg-primary/5 flex flex-col sm:flex-row sm:items-center justify-between gap-3">
+                    <div>
+                      <p className="font-medium text-sm text-text-main">1-Click IdP Metadata XML Import</p>
+                      <p className="text-xs text-text-muted">Auto-fill SSO URL, Issuer & Cert from XML metadata</p>
+                    </div>
+                    <Button
+                      type="button"
+                      variant="outline"
+                      size="sm"
+                      icon="upload_file"
+                      onClick={() => idpMetadataFileRef.current?.click()}
+                    >
+                      Upload Metadata XML
+                    </Button>
+                    <input
+                      ref={idpMetadataFileRef}
+                      type="file"
+                      accept=".xml,application/xml,text/xml"
+                      className="hidden"
+                      onChange={handleIdpMetadataUpload}
+                    />
+                  </div>
+
+                  <div className="grid grid-cols-1 gap-4">
+                    <div className="flex flex-col gap-2">
+                      <label className="font-medium text-sm sm:text-base">Single Sign-On Service URL (samlEntryPoint)</label>
+                      <Input
+                        placeholder="https://idp.example.com/app/saml/sso/..."
+                        value={samlForm.samlEntryPoint}
+                        onChange={(e) => updateSamlForm("samlEntryPoint", e.target.value)}
+                        disabled={loading || samlLoading}
+                      />
+                    </div>
+
+                    <div className="flex flex-col gap-2">
+                      <label className="font-medium text-sm sm:text-base">SP Entity ID / Audience (samlIssuer)</label>
+                      <Input
+                        placeholder="urn:9router:sp"
+                        value={samlForm.samlIssuer}
+                        onChange={(e) => updateSamlForm("samlIssuer", e.target.value)}
+                        disabled={loading || samlLoading}
+                      />
+                    </div>
+
+                    <div className="flex flex-col gap-2">
+                      <div className="flex items-center justify-between">
+                        <label className="font-medium text-sm sm:text-base">IdP X.509 Certificate (samlCert)</label>
+                        <Button
+                          type="button"
+                          variant="outline"
+                          size="sm"
+                          icon="file_upload"
+                          onClick={() => certFileRef.current?.click()}
+                        >
+                          Upload Certificate
+                        </Button>
+                        <input
+                          ref={certFileRef}
+                          type="file"
+                          accept=".crt,.pem,.cer,text/plain"
+                          className="hidden"
+                          onChange={handleCertFileUpload}
+                        />
+                      </div>
+                      <textarea
+                        rows={4}
+                        placeholder="-----BEGIN CERTIFICATE-----&#10;MIIC...&#10;-----END CERTIFICATE-----"
+                        value={samlForm.samlCert}
+                        onChange={(e) => updateSamlForm("samlCert", e.target.value)}
+                        className="w-full p-2.5 rounded-lg border border-border bg-bg text-xs font-mono text-text-main focus:outline-none focus:border-primary"
+                        disabled={loading || samlLoading}
+                      />
+                      <p className="text-xs text-text-muted">Paste raw Base64 certificate or PEM block.</p>
+                    </div>
+
+                    <div className="grid grid-cols-1 sm:grid-cols-3 gap-4">
+                      <div className="flex flex-col gap-2">
+                        <label className="font-medium text-sm sm:text-base">Login Button Label</label>
+                        <Input
+                          placeholder="Sign in with SAML SSO"
+                          value={samlForm.samlLoginLabel}
+                          onChange={(e) => updateSamlForm("samlLoginLabel", e.target.value)}
+                          disabled={loading || samlLoading}
+                        />
+                      </div>
+
+                      <div className="flex flex-col gap-2">
+                        <label className="font-medium text-sm sm:text-base">Email Claim Attribute</label>
+                        <Input
+                          placeholder="email"
+                          value={samlForm.samlAttributeEmail}
+                          onChange={(e) => updateSamlForm("samlAttributeEmail", e.target.value)}
+                          disabled={loading || samlLoading}
+                        />
+                      </div>
+
+                      <div className="flex flex-col gap-2">
+                        <label className="font-medium text-sm sm:text-base">Display Name Claim</label>
+                        <Input
+                          placeholder="name"
+                          value={samlForm.samlAttributeName}
+                          onChange={(e) => updateSamlForm("samlAttributeName", e.target.value)}
+                          disabled={loading || samlLoading}
+                        />
+                      </div>
+                    </div>
+                  </div>
+
+                  <div className="flex flex-col gap-2 p-3 rounded-lg border border-border bg-bg text-xs sm:text-sm text-text-muted">
+                    <div className="flex items-center justify-between gap-2">
+                      <div>
+                        <p className="font-medium text-text-main">ACS Callback URL</p>
+                        <code className="block break-all font-mono text-xs">{samlAcsUrl}</code>
+                      </div>
+                      <Button
+                        type="button"
+                        variant="outline"
+                        size="sm"
+                        icon="content_copy"
+                        onClick={() => {
+                          navigator.clipboard.writeText(samlAcsUrl);
+                          setSamlStatus({ type: "success", message: "ACS URL copied to clipboard!" });
+                        }}
+                      >
+                        Copy
+                      </Button>
+                    </div>
+                    <div className="flex items-center justify-between gap-2 pt-2 border-t border-border/50">
+                      <div>
+                        <p className="font-medium text-text-main">SP XML Metadata</p>
+                        <code className="block break-all font-mono text-xs">{samlMetadataUrl}</code>
+                      </div>
+                      <a
+                        href={samlMetadataUrl}
+                        target="_blank"
+                        rel="noopener noreferrer"
+                        download="9router-sp-metadata.xml"
+                        className="inline-flex items-center gap-1 text-xs font-medium text-primary hover:underline"
+                      >
+                        <span className="material-symbols-outlined text-[16px]">download</span>
+                        Download XML
+                      </a>
+                    </div>
+                  </div>
+
+                  <div className="flex flex-col sm:flex-row gap-2 pt-2 border-t border-border/50">
+                    <Button
+                      type="button"
+                      variant="primary"
+                      loading={samlLoading}
+                      onClick={() => saveSamlSettings(oidcForm.authMode)}
+                      className="w-full sm:w-auto"
+                    >
+                      Save SAML settings
+                    </Button>
+                    <Button
+                      type="button"
+                      variant="outline"
+                      loading={samlTestLoading}
+                      onClick={testSamlConnection}
+                      className="w-full sm:w-auto"
+                    >
+                      Test SAML settings
+                    </Button>
+                  </div>
+
+                  {samlTestStatus.message && (
+                    <p className={`text-xs sm:text-sm ${samlTestStatus.type === "error" ? "text-red-500" : "text-green-500"}`}>
+                      {samlTestStatus.message}
+                    </p>
+                  )}
+
+                  {samlStatus.message && (
+                    <p className={`text-xs sm:text-sm ${samlStatus.type === "error" ? "text-red-500" : "text-green-500"}`}>
+                      {samlStatus.message}
+                    </p>
+                  )}
+                </div>
+              ) : (
+                /* OIDC Panel */
+                <div className="flex flex-col gap-4 pt-2 border-t border-border/50">
+                  <div className="grid grid-cols-1 gap-4">
+                    <div className="flex flex-col gap-2">
+                      <label className="font-medium text-sm sm:text-base">Issuer URL</label>
+                      <Input
+                        placeholder="https://auth.example.com/application/o/9router/"
+                        value={oidcForm.oidcIssuerUrl}
+                        onChange={(e) => updateOidcForm("oidcIssuerUrl", e.target.value)}
+                        disabled={loading || oidcLoading}
+                      />
+                    </div>
+
+                    <div className="flex flex-col gap-2">
+                      <label className="font-medium text-sm sm:text-base">Client ID</label>
+                      <Input
+                        placeholder="9router-dashboard"
+                        value={oidcForm.oidcClientId}
+                        onChange={(e) => updateOidcForm("oidcClientId", e.target.value)}
+                        disabled={loading || oidcLoading}
+                      />
+                    </div>
+
+                    <div className="flex flex-col gap-2">
+                      <label className="font-medium text-sm sm:text-base">Client Secret</label>
+                      <Input
+                        type="password"
+                        placeholder="Leave blank to keep existing secret"
+                        value={oidcClientSecret}
+                        onChange={(e) => setOidcClientSecret(e.target.value)}
+                        disabled={loading || oidcLoading}
+                      />
+                      <p className="text-xs sm:text-sm text-text-muted">This value is write-only after saving.</p>
+                    </div>
+
+                    <div className="flex flex-col gap-2">
+                      <label className="font-medium text-sm sm:text-base">Scopes</label>
+                      <Input
+                        placeholder="openid profile email"
+                        value={oidcForm.oidcScopes}
+                        onChange={(e) => updateOidcForm("oidcScopes", e.target.value)}
+                        disabled={loading || oidcLoading}
+                      />
+                    </div>
+
+                    <div className="flex flex-col gap-2">
+                      <label className="font-medium text-sm sm:text-base">Login Button Label</label>
+                      <Input
+                        placeholder="Sign in with OIDC"
+                        value={oidcForm.oidcLoginLabel}
+                        onChange={(e) => updateOidcForm("oidcLoginLabel", e.target.value)}
+                        disabled={loading || oidcLoading}
+                      />
+                    </div>
+                  </div>
+
+                  <div className="rounded-lg border border-border bg-bg p-3 text-xs sm:text-sm text-text-muted">
+                    <p className="font-medium text-text-main mb-1">Redirect URI</p>
+                    <code className="block break-all font-mono">{oidcRedirectUri}</code>
+                  </div>
+
+                  <div className="flex flex-col sm:flex-row gap-2 pt-2 border-t border-border/50">
+                    <Button type="button" variant="primary" loading={oidcLoading} onClick={() => saveOidcSettings()} className="w-full sm:w-auto">
+                      Save OIDC settings
+                    </Button>
+                    <Button type="button" variant="outline" loading={oidcTestLoading} onClick={testOidcConnection} className="w-full sm:w-auto">
+                      Test connection
+                    </Button>
+                  </div>
+
+                  {oidcTestStatus.message && (
+                    <p className={`text-xs sm:text-sm ${oidcTestStatus.type === "error" ? "text-red-500" : "text-green-500"}`}>
+                      {oidcTestStatus.message}
+                    </p>
+                  )}
+
+                  {oidcStatus.message && (
+                    <p className={`text-xs sm:text-sm ${oidcStatus.type === "error" ? "text-red-500" : "text-green-500"}`}>
+                      {oidcStatus.message}
+                    </p>
+                  )}
+                </div>
+              )}
+
+              {settings.authMode === "oidc" || settings.authMode === "saml" || settings.authMode === "sso" ? (
+                <p className="text-xs sm:text-sm text-amber-600 dark:text-amber-400">
+                  SSO login ({settings.ssoType === "saml" ? "SAML 2.0" : "OIDC"}) is currently active. Password login is disabled until you switch back.
+                </p>
+              ) : null}
+
+              {settings.authMode === "both" && (
+                <p className="text-xs sm:text-sm text-amber-600 dark:text-amber-400">
+                  Password and SSO login ({settings.ssoType === "saml" ? "SAML 2.0" : "OIDC"}) are both active.
+                </p>
+              )}
             </div>
-
-            <div className="grid grid-cols-1 gap-4">
-              <div className="flex flex-col gap-2">
-                <label className="font-medium text-sm sm:text-base">Issuer URL</label>
-                <Input
-                  placeholder="https://auth.example.com/application/o/9router/"
-                  value={oidcForm.oidcIssuerUrl}
-                  onChange={(e) => updateOidcForm("oidcIssuerUrl", e.target.value)}
-                  disabled={loading || oidcLoading}
-                />
-              </div>
-
-              <div className="flex flex-col gap-2">
-                <label className="font-medium text-sm sm:text-base">Client ID</label>
-                <Input
-                  placeholder="9router-dashboard"
-                  value={oidcForm.oidcClientId}
-                  onChange={(e) => updateOidcForm("oidcClientId", e.target.value)}
-                  disabled={loading || oidcLoading}
-                />
-              </div>
-
-              <div className="flex flex-col gap-2">
-                <label className="font-medium text-sm sm:text-base">Client Secret</label>
-                <Input
-                  type="password"
-                  placeholder="Leave blank to keep existing secret"
-                  value={oidcClientSecret}
-                  onChange={(e) => setOidcClientSecret(e.target.value)}
-                  disabled={loading || oidcLoading}
-                />
-                <p className="text-xs sm:text-sm text-text-muted">This value is write-only after saving.</p>
-              </div>
-
-              <div className="flex flex-col gap-2">
-                <label className="font-medium text-sm sm:text-base">Scopes</label>
-                <Input
-                  placeholder="openid profile email"
-                  value={oidcForm.oidcScopes}
-                  onChange={(e) => updateOidcForm("oidcScopes", e.target.value)}
-                  disabled={loading || oidcLoading}
-                />
-              </div>
-
-              <div className="flex flex-col gap-2">
-                <label className="font-medium text-sm sm:text-base">Login Button Label</label>
-                <Input
-                  placeholder="Sign in with OIDC"
-                  value={oidcForm.oidcLoginLabel}
-                  onChange={(e) => updateOidcForm("oidcLoginLabel", e.target.value)}
-                  disabled={loading || oidcLoading}
-                />
-              </div>
-            </div>
-
-            <div className="rounded-lg border border-border bg-bg p-3 text-xs sm:text-sm text-text-muted">
-              <p className="font-medium text-text-main mb-1">Redirect URI</p>
-              <code className="block break-all font-mono">{oidcRedirectUri}</code>
-            </div>
-
-            <div className="flex flex-col sm:flex-row gap-2 pt-2 border-t border-border/50">
-              <Button type="button" variant="primary" loading={oidcLoading} onClick={() => saveOidcSettings()} className="w-full sm:w-auto">
-                Save auth mode
-              </Button>
-              <Button type="button" variant="outline" loading={oidcTestLoading} onClick={testOidcConnection} className="w-full sm:w-auto">
-                Test connection
-              </Button>
-            </div>
-
-            {oidcTestStatus.message && (
-              <p className={`text-xs sm:text-sm ${oidcTestStatus.type === "error" ? "text-red-500" : "text-green-500"}`}>
-                {oidcTestStatus.message}
-              </p>
-            )}
-
-            {oidcStatus.message && (
-              <p className={`text-xs sm:text-sm ${oidcStatus.type === "error" ? "text-red-500" : "text-green-500"}`}>
-                {oidcStatus.message}
-              </p>
-            )}
-
-            {settings.authMode === "oidc" && (
-              <p className="text-xs sm:text-sm text-amber-600 dark:text-amber-400">
-                OIDC login is currently active. Password login is disabled until you switch back.
-              </p>
-            )}
-
-            {settings.authMode === "both" && (
-              <p className="text-xs sm:text-sm text-amber-600 dark:text-amber-400">
-                Password and OIDC login are both active.
-              </p>
-            )}
-          </div>
           )}
         </Card>
 

src/app/(dashboard)/dashboard/token-saver/TokenSaverClient.js

@@ -529,8 +529,7 @@ export default function TokenSaverClient() {
             </p>
           </div>
           <Toggle
-            checked={headroomEnabled && headroomRunning}
-            disabled={!headroomRunning}
+            checked={headroomEnabled}
             onChange={() => handleHeadroomEnabled(!headroomEnabled)}
           />
         </div>

src/app/(dashboard)/dashboard/usage/components/ProviderLimits/index.js

@@ -216,7 +216,7 @@ export default function ProviderLimits() {
   );
 
   // Fetch quota for a specific connection
-  const fetchQuota = useCallback(async (connectionId, provider) => {
+  const fetchQuota = useCallback(async (connectionId, provider, { force = false } = {}) => {
     setLoading((prev) => ({ ...prev, [connectionId]: true }));
     setErrors((prev) => ({ ...prev, [connectionId]: null }));
 
@@ -224,7 +224,8 @@ export default function ProviderLimits() {
       console.log(
         `[ProviderLimits] Fetching quota for ${provider} (${connectionId})`,
       );
-      const response = await fetch(`/api/usage/${connectionId}`);
+      const url = `/api/usage/${connectionId}${force ? "?force=1" : ""}`;
+      const response = await fetch(url);
 
       if (!response.ok) {
         const errorData = await response.json().catch(() => ({}));
@@ -295,7 +296,7 @@ export default function ProviderLimits() {
   // Refresh quota for a specific provider
   const refreshProvider = useCallback(
     async (connectionId, provider) => {
-      await fetchQuota(connectionId, provider);
+      await fetchQuota(connectionId, provider, { force: true });
       setLastUpdated(new Date());
     },
     [fetchQuota],

src/app/(dashboard)/dashboard/usage/components/ProviderLimits/utils.js

@@ -4,7 +4,7 @@ import { getModelsByProviderId } from "open-sse/config/providerModels.js";
 export const QUOTA_CACHE_KEY = "quotaCacheData";
 export const REFRESH_INTERVAL_MS = 60000;
 // Claude usage/quota endpoint rate-limits; poll it less often than other providers
-export const CLAUDE_REFRESH_INTERVAL_MS = 180000;
+export const CLAUDE_REFRESH_INTERVAL_MS = 600000;
 export const DEPLETED_QUOTA_THRESHOLD = 5;
 export const AUTO_REFRESH_STORAGE_KEY = "quotaAutoRefresh";
 export const CONNECTIONS_PAGE_SIZE = 20;
@@ -36,6 +36,17 @@ export function getConnectionQuotaRemaining(connection, quotaData) {
   return Number.POSITIVE_INFINITY;
 }
 
+// Stable group-by-provider: first-seen provider order, original order within group.
+function groupByProviderStable(connections) {
+  const seen = new Map();
+  for (const conn of connections) {
+    const key = conn.provider || "";
+    if (!seen.has(key)) seen.set(key, []);
+    seen.get(key).push(conn);
+  }
+  return Array.from(seen.values()).flat();
+}
+
 export function sortVisibleConnections(
   connections,
   quotaData,
@@ -58,7 +69,7 @@ export function sortVisibleConnections(
     });
   }
 
-  if (!expiringFirst) return connections;
+  if (!expiringFirst) return groupByProviderStable(connections);
 
   const getEarliestResetTime = (connection) => {
     const resetTimes = (quotaData[connection.id]?.quotas || [])

src/app/api/auth/login/route.js

@@ -4,6 +4,7 @@ import bcrypt from "bcryptjs";
 import { cookies } from "next/headers";
 import { setDashboardAuthCookie } from "@/lib/auth/dashboardSession";
 import { isOidcConfigured } from "@/lib/auth/oidc";
+import { isSamlConfigured } from "@/lib/auth/saml.js";
 import { checkLock, recordFail, recordSuccess, getClientIp } from "@/lib/auth/loginLimiter";
 import { isLocalRequest } from "@/dashboardGuard";
 
@@ -39,8 +40,14 @@ export async function POST(request) {
     // Default password is '123456' if not set
     const storedHash = settings.password;
 
-    if (settings.authMode === "oidc" && isOidcConfigured(settings)) {
-      return NextResponse.json({ error: "Password login is disabled. Use OIDC sign in." }, { status: 403 });
+    if (settings.authMode === "sso" || settings.authMode === "saml" || settings.authMode === "oidc") {
+      const ssoType = settings.ssoType || (settings.authMode === "saml" ? "saml" : "oidc");
+      if (ssoType === "saml" && isSamlConfigured(settings)) {
+        return NextResponse.json({ error: "Password login is disabled. Use SAML SSO sign in." }, { status: 403 });
+      }
+      if (ssoType === "oidc" && isOidcConfigured(settings)) {
+        return NextResponse.json({ error: "Password login is disabled. Use OIDC sign in." }, { status: 403 });
+      }
     }
 
     let isValid = false;
@@ -54,15 +61,36 @@ export async function POST(request) {
 
     if (isValid) {
       recordSuccess(ip);
-      const cookieStore = await cookies();
-      await setDashboardAuthCookie(cookieStore, request);
 
       // Default password still in use on a remote client → force a password
       // change before the dashboard is exposed remotely (keeps local UX intact).
       const mustChangePassword =
         !storedHash && !process.env.INITIAL_PASSWORD && !isLocalRequest(request);
 
-      return NextResponse.json({ success: true, mustChangePassword }, { headers: NO_STORE_HEADERS });
+      if (mustChangePassword) {
+        // Do NOT issue a session token: a fresh install's default password is
+        // public knowledge ("123456"), so handing out a valid JWT would let any
+        // remote attacker authenticate and (e.g.) PATCH /api/settings to disable
+        // authentication entirely (CVE-2026-56679 class). Require the password
+        // to be changed first.
+        //
+        // NOTE: this intentionally leaves no remote self-service password-change
+        // path — the change-password flow (PATCH /api/settings) requires a JWT,
+        // which we deliberately withhold. A remote fresh-install user must either
+        // change the password from the local machine or set INITIAL_PASSWORD
+        // before first launch. This is a deliberate security trade-off, not an
+        // oversight: issuing any credential before the default password is
+        // rotated re-opens the exact attack chain this branch closes.
+        return NextResponse.json(
+          { success: false, error: "Default password must be changed before remote access. Change it from the local machine (or set INITIAL_PASSWORD).", mustChangePassword },
+          { status: 403, headers: NO_STORE_HEADERS }
+        );
+      }
+
+      const cookieStore = await cookies();
+      await setDashboardAuthCookie(cookieStore, request);
+
+      return NextResponse.json({ success: true, mustChangePassword: false }, { headers: NO_STORE_HEADERS });
     }
 
     const { remainingBeforeLock } = recordFail(ip);

src/app/api/auth/saml/acs/route.js

@@ -0,0 +1,69 @@
+import { NextResponse } from "next/server";
+import { cookies } from "next/headers";
+import { getSettings } from "@/lib/localDb";
+import {
+  getSamlBaseUrl,
+  isSamlConfigured,
+  pickSamlDisplayName,
+  pickSamlEmail,
+  validateSamlResponse,
+} from "@/lib/auth/saml.js";
+import { setDashboardAuthCookie } from "@/lib/auth/dashboardSession";
+import { checkLock, recordFail, recordSuccess, getClientIp } from "@/lib/auth/loginLimiter";
+
+export async function POST(request) {
+  const settings = await getSettings();
+  const origin = getSamlBaseUrl(request, settings);
+  const ip = getClientIp(request);
+
+  const lock = checkLock(ip);
+  if (lock.locked) {
+    return NextResponse.redirect(
+      new URL(
+        `/login?error=${encodeURIComponent(`Too many failed attempts. Try again in ${lock.retryAfter}s.`)}`,
+        origin
+      )
+    );
+  }
+
+  const cookieStore = await cookies();
+  const storedRequestId = cookieStore.get("saml_state")?.value || "";
+
+  // Always clear saml_state cookie after attempt
+  cookieStore.delete("saml_state");
+
+  try {
+    const formData = await request.formData();
+    const SAMLResponse = formData.get("SAMLResponse");
+
+    if (!SAMLResponse) {
+      recordFail(ip);
+      return NextResponse.redirect(new URL("/login?error=saml_missing_response", origin));
+    }
+
+    if (!isSamlConfigured(settings)) {
+      recordFail(ip);
+      return NextResponse.redirect(new URL("/login?error=saml_not_configured", origin));
+    }
+
+    const profile = await validateSamlResponse(request, { SAMLResponse }, storedRequestId, settings);
+
+    const samlEmail = pickSamlEmail(profile, settings) || null;
+    const samlName = pickSamlDisplayName(profile, settings) || "SAML user";
+
+    recordSuccess(ip);
+
+    await setDashboardAuthCookie(cookieStore, request, {
+      saml: true,
+      samlEmail,
+      samlName,
+    });
+
+    return NextResponse.redirect(new URL("/dashboard", origin));
+  } catch (error) {
+    recordFail(ip);
+    return NextResponse.redirect(
+      new URL(`/login?error=${encodeURIComponent(error.message || "saml_acs_failed")}`, origin)
+    );
+  }
+}

src/app/api/auth/saml/metadata/route.js

@@ -0,0 +1,25 @@
+import { getSettings } from "@/lib/localDb";
+import { generateSamlMetadata } from "@/lib/auth/saml";
+
+export async function GET(request) {
+  try {
+    const settings = await getSettings();
+    const origin = new URL(request.url).origin;
+    const metadataXml = generateSamlMetadata(origin, settings);
+
+    return new Response(metadataXml, {
+      status: 200,
+      headers: {
+        "Content-Type": "application/xml",
+        "Cache-Control": "no-cache",
+      },
+    });
+  } catch (error) {
+    return new Response(`<?xml version="1.0"?><Error>${error.message || "Failed to generate metadata"}</Error>`, {
+      status: 500,
+      headers: {
+        "Content-Type": "application/xml",
+      },
+    });
+  }
+}

src/app/api/auth/saml/start/route.js

@@ -0,0 +1,32 @@
+import { NextResponse } from "next/server";
+import { cookies } from "next/headers";
+import { getSettings } from "@/lib/localDb";
+import { buildSamlAuthorizeUrl, getSamlBaseUrl, isSamlConfigured } from "@/lib/auth/saml.js";
+import { shouldUseSecureCookie } from "@/lib/auth/dashboardSession";
+
+export async function GET(request) {
+  const settings = await getSettings();
+  const origin = getSamlBaseUrl(request, settings);
+  try {
+    if (!isSamlConfigured(settings)) {
+      return NextResponse.redirect(new URL("/login?error=saml_not_configured", origin));
+    }
+
+    const { authorizeUrl, requestId } = await buildSamlAuthorizeUrl(request, settings);
+
+    const cookieStore = await cookies();
+    cookieStore.set("saml_state", requestId, {
+      httpOnly: true,
+      secure: shouldUseSecureCookie(request),
+      sameSite: "lax",
+      path: "/",
+      maxAge: 10 * 60,
+    });
+
+    return NextResponse.redirect(authorizeUrl);
+  } catch (error) {
+    return NextResponse.redirect(
+      new URL(`/login?error=${encodeURIComponent(error.message || "saml_start_failed")}`, origin)
+    );
+  }
+}

src/app/api/auth/saml/test/route.js

@@ -0,0 +1,72 @@
+import { NextResponse } from "next/server";
+import { cookies } from "next/headers";
+import { getSettings } from "@/lib/localDb";
+import { formatX509Certificate } from "@/lib/auth/saml.js";
+import { verifyDashboardAuthToken } from "@/lib/auth/dashboardSession";
+
+async function canAccessTestRoute() {
+  const settings = await getSettings();
+  if (settings.requireLogin === false) return true;
+
+  const cookieStore = await cookies();
+  const token = cookieStore.get("auth_token")?.value;
+  return await verifyDashboardAuthToken(token);
+}
+
+export async function POST(request) {
+  try {
+    if (!(await canAccessTestRoute())) {
+      return NextResponse.json({ error: "Unauthorized" }, { status: 401 });
+    }
+
+    const body = await request.json().catch(() => ({}));
+    const settings = await getSettings();
+
+    const samlEntryPoint = String(body.samlEntryPoint || settings.samlEntryPoint || "").trim();
+    const samlIssuer = String(body.samlIssuer || settings.samlIssuer || "urn:9router:sp").trim();
+    const samlCert = String(
+      Object.prototype.hasOwnProperty.call(body, "samlCert")
+        ? body.samlCert
+        : settings.samlCert || ""
+    ).trim();
+
+    if (!samlEntryPoint) {
+      return NextResponse.json({ error: "Single Sign-On Service URL (samlEntryPoint) is required" }, { status: 400 });
+    }
+
+    try {
+      new URL(samlEntryPoint);
+    } catch {
+      return NextResponse.json({ error: "Single Sign-On Service URL must be a valid URL" }, { status: 400 });
+    }
+
+    if (!samlIssuer) {
+      return NextResponse.json({ error: "SP Entity ID / Issuer (samlIssuer) is required" }, { status: 400 });
+    }
+
+    if (!samlCert) {
+      return NextResponse.json({ error: "IdP X.509 Certificate (samlCert) is required" }, { status: 400 });
+    }
+
+    const formattedCert = formatX509Certificate(samlCert);
+    if (!formattedCert) {
+      return NextResponse.json({ error: "Invalid IdP X.509 Certificate format" }, { status: 400 });
+    }
+
+    const origin = new URL(request.url).origin;
+    const acsUrl = `${origin}/api/auth/saml/acs`;
+    const metadataUrl = `${origin}/api/auth/saml/metadata`;
+
+    return NextResponse.json({
+      ok: true,
+      samlEntryPoint,
+      samlIssuer,
+      certValid: true,
+      acsUrl,
+      metadataUrl,
+      message: "SAML 2.0 configuration verified successfully.",
+    });
+  } catch (error) {
+    return NextResponse.json({ error: error.message || "SAML test failed" }, { status: 500 });
+  }
+}

src/app/api/auth/status/route.js

@@ -2,6 +2,7 @@ import { NextResponse } from "next/server";
 import { cookies } from "next/headers";
 import { getSettings } from "@/lib/localDb";
 import { isOidcConfigured } from "@/lib/auth/oidc";
+import { isSamlConfigured } from "@/lib/auth/saml.js";
 import { getDashboardAuthSession } from "@/lib/auth/dashboardSession";
 
 export async function GET() {
@@ -11,16 +12,29 @@ export async function GET() {
     const session = await getDashboardAuthSession(cookieStore.get("auth_token")?.value);
     const requireLogin = settings.requireLogin !== false;
     const authMode = settings.authMode || "password";
+    const ssoType = settings.ssoType || "oidc";
     const oidcName = String(session?.oidcName || "").trim();
     const oidcEmail = String(session?.oidcEmail || "").trim();
-    const displayName = oidcName || oidcEmail || (session?.oidc ? "OIDC user" : "Password user");
-    const loginMethod = session?.oidc ? "OIDC" : "Password";
+    const samlName = String(session?.samlName || "").trim();
+    const samlEmail = String(session?.samlEmail || "").trim();
+
+    const displayName =
+      samlName ||
+      samlEmail ||
+      oidcName ||
+      oidcEmail ||
+      (session?.saml ? "SAML user" : session?.oidc ? "OIDC user" : "Password user");
+
+    const loginMethod = session?.saml ? "SAML" : session?.oidc ? "OIDC" : "Password";
 
     return NextResponse.json({
       requireLogin,
       authMode,
+      ssoType,
       oidcConfigured: isOidcConfigured(settings),
       oidcLoginLabel: (settings.oidcLoginLabel || "Sign in with OIDC").trim() || "Sign in with OIDC",
+      samlConfigured: isSamlConfigured(settings),
+      samlLoginLabel: (settings.samlLoginLabel || "Sign in with SAML SSO").trim() || "Sign in with SAML SSO",
       hasPassword: !!settings.password,
       displayName,
       loginMethod,
@@ -28,13 +42,19 @@ export async function GET() {
       oidcName: oidcName || null,
       oidcEmail: oidcEmail || null,
       oidcLogin: !!session?.oidc,
+      samlName: samlName || null,
+      samlEmail: samlEmail || null,
+      samlLogin: !!session?.saml,
     });
   } catch {
     return NextResponse.json({
       requireLogin: true,
       authMode: "password",
+      ssoType: "oidc",
       oidcConfigured: false,
       oidcLoginLabel: "Sign in with OIDC",
+      samlConfigured: false,
+      samlLoginLabel: "Sign in with SAML SSO",
       hasPassword: false,
       displayName: "Password user",
       loginMethod: "Password",
@@ -42,6 +62,9 @@ export async function GET() {
       oidcName: null,
       oidcEmail: null,
       oidcLogin: false,
+      samlName: null,
+      samlEmail: null,
+      samlLogin: false,
     });
   }
 }

src/app/api/cli-tools/hermes-settings/route.js

@@ -20,7 +20,7 @@ const getHermesEnvPath = () => path.join(getHermesDir(), ".env");
 const MODEL_BLOCK_RE = /^model:[ \t]*\r?\n((?:[ \t]+.*\r?\n?|[ \t]*\r?\n)*)/m;
 
 const buildModelBlock = (model, baseUrl) =>
-  `model:\n  default: "${model}"\n  provider: "custom"\n  base_url: "${baseUrl}"\n`;
+  `model:\n  default: "${model}"\n  provider: "custom"\n  base_url: "${baseUrl}"\n  api_key: \${OPENAI_API_KEY}\n`;
 
 // Parse current model block back to fields (best-effort, simple key:value)
 const parseModelBlock = (yaml) => {
@@ -35,6 +35,7 @@ const parseModelBlock = (yaml) => {
     default: get("default"),
     provider: get("provider"),
     base_url: get("base_url"),
+    api_key: get("api_key"),
   };
 };
 

src/app/api/models/test/ping.js

@@ -135,9 +135,11 @@ export async function pingModelByKind(model, kind, baseUrl = `http://127.0.0.1:$
     headers,
     body: JSON.stringify({
       model,
-      // Claude-on-Copilot returns empty choices at max_tokens:1 (budget is spent
-      // before a content token emits), so a 1-token probe yields a false negative.
-      max_tokens: 16,
+      // 1024 tokens: reasoning models (ClinePass/kimi-k3, deepseek-v4-pro, etc.) spend
+      // their budget on chain-of-thought before emitting an answer. A tiny probe like
+      // max_tokens:16 starves the answer and yields a false "no choices" failure.
+      // See issue #3010.
+      max_tokens: 1024,
       stream: false,
       messages: [{ role: "user", content: "hi" }],
     }),
@@ -180,6 +182,21 @@ export async function pingModelByKind(model, kind, baseUrl = `http://127.0.0.1:$
   }
 
   const hasChoices = Array.isArray(parsed?.choices) && parsed.choices.length > 0;
+
+  // Soft-pass (issue #3010): a reasoning model may burn its whole budget on
+  // chain-of-thought and return finish_reason:"length" with empty content but
+  // non-empty reasoning/thinking. That's a successful connection, not a failure.
+  const firstChoice = parsed?.choices?.[0] || {};
+  const hasReasoning =
+    firstChoice.message?.reasoning ||
+    firstChoice.message?.reasoning_content ||
+    firstChoice.message?.thinking ||
+    firstChoice.message?.thinking_content;
+  const contentEmpty = !String(firstChoice.message?.content || "").trim();
+  if (hasChoices && firstChoice.finish_reason === "length" && contentEmpty && hasReasoning) {
+    return { ok: true, latencyMs, error: null, status: res.status, note: "reasoning-only response (length-limited)" };
+  }
+
   if (!hasChoices) {
     return {
       ok: false,

src/app/api/providers/[id]/test/testUtils.js

@@ -788,6 +788,27 @@ async function testApiKeyConnection(connection, effectiveProxy = null) {
         );
         return { valid: exRes.ok, error: exRes.ok ? null : "Invalid Personal Access Token" };
       }
+case "llm7": {
+        const baseUrl = connection.providerSpecificData?.baseUrl || "https://api.llm7.io/v1";
+        const res = await fetchWithConnectionProxy(`${baseUrl.replace(/\/$/, "")}/models`, {
+          headers: { Authorization: `Bearer ${connection.apiKey}` },
+        }, effectiveProxy);
+        return { valid: res.ok, error: res.ok ? null : "Invalid API key or base URL" };
+      }
+      case "kimchi": {
+        // Dual-auth: same validation endpoint as the OAuth flow — the token (API key
+        // or OAuth access token) is sent as Authorization: Bearer.
+        const url = KIMCHI_CONFIG.validationUrl || "https://api.cast.ai/v1/llm/openai/supported-providers";
+        const res = await fetchWithConnectionProxy(url, {
+          method: "GET",
+          headers: {
+            Accept: "application/json",
+            Authorization: `Bearer ${connection.apiKey}`,
+            "User-Agent": "kimchi/0.1.40",
+          },
+        }, effectiveProxy);
+        return { valid: res.ok, error: res.ok ? null : "Invalid API key", refreshed: false };
+      }
       default:
         return { valid: false, error: "Provider test not supported" };
     }

src/app/api/usage/[connectionId]/route.js

@@ -123,6 +123,7 @@ export async function GET(request, { params }) {
   let connection;
   try {
     const { connectionId } = await params;
+    const force = new URL(request.url).searchParams.get("force") === "1";
 
 
     // Get connection from database
@@ -168,7 +169,7 @@ export async function GET(request, { params }) {
     }
 
     // Fetch usage from provider API
-    let usage = await getUsageForProvider(connection, proxyOptions);
+    let usage = await getUsageForProvider(connection, proxyOptions, { force });
 
     // If provider returned an auth-expired message instead of throwing,
     // force-refresh token and retry once (OAuth only)
@@ -176,7 +177,7 @@ export async function GET(request, { params }) {
       try {
         const retryResult = await refreshAndUpdateCredentials(connection, true, proxyOptions);
         connection = retryResult.connection;
-        usage = await getUsageForProvider(connection, proxyOptions);
+        usage = await getUsageForProvider(connection, proxyOptions, { force });
       } catch (retryError) {
         console.warn(`[Usage] ${connection.provider}: force refresh failed: ${retryError.message}`);
       }

src/app/api/usage/request-details/route.js

@@ -47,8 +47,23 @@ export async function GET(request) {
     if (endDate) filter.endDate = endDate;
     
     const result = await getRequestDetails(filter);
-    
-    return NextResponse.json(result);
+
+    // Redact conversation payloads: the stored details include full request
+    // bodies (user prompts, tool calls) and provider responses. Returning them
+    // wholesale lets any dashboard-authenticated user (or, if requireLogin is
+    // disabled, anyone) read every user's conversation history. Keep the
+    // metadata (model, tokens, latency, status) but drop message content.
+    const redactedDetails = (result.details || []).map((d) => {
+      const redacted = { ...d };
+      for (const key of ["request", "providerRequest", "providerResponse", "response"]) {
+        if (redacted[key] !== undefined) {
+          redacted[key] = { redacted: true };
+        }
+      }
+      return redacted;
+    });
+
+    return NextResponse.json({ ...result, details: redactedDetails });
   } catch (error) {
     console.error("[API] Failed to get request details:", error);
     return NextResponse.json(

src/app/api/v1/models/route.js

@@ -485,6 +485,27 @@ export async function buildModelsList(kindFilter, options = {}) {
           || capabilitiesFromServiceKind(customKind || liveKind)
           || (kind === LLM_KIND ? getCapabilitiesForModel(providerId, modelId) : null);
         if (caps) model.capabilities = caps;
+        // Token limits under the snake_case names the OpenAI/OpenRouter
+        // convention uses. `capabilities.contextWindow` is camelCase and nested,
+        // so clients matching context_length find nothing, fall back to guessing
+        // the window from the model name, and guess high — a 372k model read as
+        // 1.05M never reaches its compaction threshold and hard-fails upstream.
+        // Emitted at top level because not every client recurses into nested
+        // objects; the camelCase `capabilities` block stays for compatibility.
+        if (kind === LLM_KIND || allowAsLlm) {
+          let contextWindow = caps?.contextWindow;
+          let maxOutput = caps?.maxOutput;
+          // Live-catalog and service-kind capabilities are usually partial
+          // (often just { tools: true }), so fill the gaps from the static
+          // table rather than emitting null and leaving clients to guess.
+          if (!Number.isFinite(contextWindow) || !Number.isFinite(maxOutput)) {
+            const fallback = getCapabilitiesForModel(providerId, modelId);
+            if (!Number.isFinite(contextWindow)) contextWindow = fallback.contextWindow;
+            if (!Number.isFinite(maxOutput)) maxOutput = fallback.maxOutput;
+          }
+          if (Number.isFinite(contextWindow)) model.context_length = contextWindow;
+          if (Number.isFinite(maxOutput)) model.max_completion_tokens = maxOutput;
+        }
         models.push(model);
       }
 

src/app/login/page.js

@@ -11,8 +11,11 @@ export default function LoginPage() {
   const [loading, setLoading] = useState(false);
   const [hasPassword, setHasPassword] = useState(null);
   const [authMode, setAuthMode] = useState("password");
+  const [ssoType, setSsoType] = useState("oidc");
   const [oidcConfigured, setOidcConfigured] = useState(false);
   const [oidcLoginLabel, setOidcLoginLabel] = useState("Sign in with OIDC");
+  const [samlConfigured, setSamlConfigured] = useState(false);
+  const [samlLoginLabel, setSamlLoginLabel] = useState("Sign in with SAML SSO");
   const [mustChange, setMustChange] = useState(false);
   const [newPassword, setNewPassword] = useState("");
 
@@ -43,8 +46,11 @@ export default function LoginPage() {
           }
           setHasPassword(!!data.hasPassword);
           setAuthMode(data.authMode || "password");
+          setSsoType(data.ssoType || "oidc");
           setOidcConfigured(data.oidcConfigured === true);
           setOidcLoginLabel(data.oidcLoginLabel || "Sign in with OIDC");
+          setSamlConfigured(data.samlConfigured === true);
+          setSamlLoginLabel(data.samlLoginLabel || "Sign in with SAML SSO");
         } else {
           // Safe fallback on non-OK response to avoid infinite loading state.
           setHasPassword(true);
@@ -118,8 +124,18 @@ export default function LoginPage() {
     window.location.href = "/api/auth/oidc/start";
   };
 
-  const oidcAvailable = oidcConfigured && ["oidc", "both"].includes(authMode);
-  const passwordAvailable = authMode !== "oidc" || !oidcConfigured;
+  const handleSamlLogin = () => {
+    window.location.href = "/api/auth/saml/start";
+  };
+
+  const isSsoEnabled = ["sso", "oidc", "saml", "both"].includes(authMode);
+  const activeSsoType = ssoType || (authMode === "saml" ? "saml" : "oidc");
+
+  const samlAvailable = isSsoEnabled && activeSsoType === "saml" && samlConfigured;
+  const oidcAvailable = isSsoEnabled && activeSsoType === "oidc" && oidcConfigured;
+  const ssoAvailable = samlAvailable || oidcAvailable;
+
+  const passwordAvailable = authMode === "password" || authMode === "both" || !ssoAvailable;
 
   // Show loading state while checking password
   if (hasPassword === null) {
@@ -141,7 +157,9 @@ export default function LoginPage() {
         <div className="text-center mb-8">
           <h1 className="text-3xl font-bold text-primary mb-2">9Router</h1>
           <p className="text-text-muted">
-            {authMode === "oidc" && oidcConfigured
+            {samlAvailable
+              ? "Sign in with SAML 2.0 Single Sign-On"
+              : oidcAvailable
               ? "Sign in with your OIDC provider to access the dashboard"
               : "Enter your password to access the dashboard"}
           </p>
@@ -171,25 +189,31 @@ export default function LoginPage() {
             </form>
           ) : (
           <div className="flex flex-col gap-4">
+            {samlAvailable && (
+              <Button type="button" variant="primary" className="w-full" onClick={handleSamlLogin}>
+                {samlLoginLabel}
+              </Button>
+            )}
+
             {oidcAvailable && (
               <Button type="button" variant="primary" className="w-full" onClick={handleOidcLogin}>
                 {oidcLoginLabel}
               </Button>
             )}
 
-            {oidcAvailable && passwordAvailable && <div className="h-px bg-border/60" />}
+            {ssoAvailable && passwordAvailable && <div className="h-px bg-border/60" />}
 
             {passwordAvailable ? (
               <form onSubmit={handleLogin} className="flex flex-col gap-4">
-                {((authMode === "oidc" && !oidcConfigured) || (authMode === "both" && !oidcConfigured)) && (
+                {isSsoEnabled && !ssoAvailable && (
                   <p className="text-xs text-amber-600 dark:text-amber-400 text-center">
-                    OIDC login is enabled, but the issuer/client fields are not configured yet. Password login is still available for recovery.
+                    {activeSsoType === "saml" ? "SAML SSO" : "OIDC"} login is enabled, but configuration is incomplete. Password login is still available for recovery.
                   </p>
                 )}
 
-                {authMode === "both" && oidcConfigured && (
+                {authMode === "both" && ssoAvailable && (
                   <p className="text-xs text-text-muted text-center">
-                    Password and OIDC login are both enabled.
+                    Password and {activeSsoType === "saml" ? "SAML SSO" : "OIDC"} login are both enabled.
                   </p>
                 )}
 

src/dashboardGuard.js

@@ -2,6 +2,7 @@ import { NextResponse } from "next/server";
 import { getSettings, validateApiKey } from "@/lib/localDb";
 import { getConsistentMachineId } from "@/shared/utils/machineId";
 import { verifyDashboardAuthToken } from "@/lib/auth/dashboardSession";
+import { hasTrustedPeerHeaders } from "@/lib/auth/trustedPeer";
 
 const CLI_TOKEN_HEADER = "x-9r-cli-token";
 const CLI_TOKEN_SALT = "9r-cli-auth";
@@ -27,6 +28,7 @@ const PUBLIC_API_PATHS = [
   "/api/auth/logout",
   "/api/auth/status",
   "/api/auth/oidc",
+  "/api/auth/saml",
   "/api/version",
   "/api/settings/require-login",
 ];
@@ -86,24 +88,40 @@ const LOCAL_ONLY_PATHS = [
 
 const LOOPBACK_HOSTS = new Set(["localhost", "127.0.0.1", "::1"]);
 
+// Accepts a Host header, a URL hostname or a raw socket address. Splitting on the first
+// colon only works for IPv4 and would reduce every IPv6 form to "", so a dual-stack
+// listener handing back ::ffff:127.0.0.1 would not read as loopback.
 function isLoopbackHostname(h) {
   if (!h) return false;
-  const name = h.split(":")[0].replace(/^\[|\]$/g, "").toLowerCase();
+  let name = String(h).trim().toLowerCase();
+  if (name.startsWith("[")) {
+    const end = name.indexOf("]");
+    if (end === -1) return false;
+    name = name.slice(1, end);
+  } else if (name.indexOf(":") !== -1 && name.indexOf(":") === name.lastIndexOf(":")) {
+    name = name.slice(0, name.indexOf(":"));
+  }
+  if (name.startsWith("::ffff:")) name = name.slice(7);
   return LOOPBACK_HOSTS.has(name);
 }
 
+function isLoopbackPeer(request) {
+  if (hasTrustedPeerHeaders(request)) {
+    return isLoopbackHostname(request.headers.get("x-9r-real-ip"));
+  }
+  // Bare `next dev` forks its server, so the wrapper never loads and no peer address
+  // reaches us. Host is spoofable, so this stays confined to development.
+  if (process.env.NODE_ENV === "development") {
+    return isLoopbackHostname(request.headers.get("host"));
+  }
+  return false;
+}
+
 export function isLocalRequest(request) {
   // Stamped by custom-server.js when forwarding headers exist: request came through
   // a reverse proxy, so the loopback socket is the proxy hop, not the end-user.
   if (request.headers.get("x-9r-via-proxy")) return false;
-  // Trusted peer IP from TCP socket (custom-server.js); unspoofable. Primary anchor for "local".
-  const realIp = request.headers.get("x-9r-real-ip");
-  if (realIp) {
-    if (!isLoopbackHostname(realIp)) return false;
-  } else if (!isLoopbackHostname(request.headers.get("host"))) {
-    // Fallback for bare server.js (dev) without custom-server: legacy Host-based check.
-    return false;
-  }
+  if (!isLoopbackPeer(request)) return false;
   const origin = request.headers.get("origin");
   if (origin) {
     try {

src/lib/auth/loginLimiter.js

@@ -1,4 +1,5 @@
 // In-memory progressive lockout for dashboard login. Resets on process restart.
+import { hasTrustedPeerHeaders } from "./trustedPeer.js";
 
 const MAX_FAILS_BEFORE_LOCK = 5;
 const LOCK_STEPS_MS = [30_000, 120_000, 600_000, 1_800_000]; // 30s, 2m, 10m, 30m
@@ -46,9 +47,12 @@ export function recordSuccess(ip) {
 }
 
 export function getClientIp(request) {
-  // Trusted: set from TCP socket by custom-server.js (client cannot spoof).
-  const realIp = request.headers.get("x-9r-real-ip");
-  if (realIp) return realIp;
+  // Trusted only when custom-server.js proves it stamped the header from the TCP socket;
+  // otherwise a client could rotate the value to escape its own lockout bucket.
+  if (hasTrustedPeerHeaders(request)) {
+    const realIp = request.headers.get("x-9r-real-ip");
+    if (realIp) return realIp;
+  }
   // Behind a trusted reverse proxy that overwrites XFF with the real client IP.
   if (process.env.TRUST_PROXY === "true") {
     const xff = request.headers.get("x-forwarded-for");

src/lib/auth/saml.js

@@ -0,0 +1,268 @@
+import { SAML } from "@node-saml/node-saml";
+import { getSettings } from "../db/repos/settingsRepo.js";
+
+/**
+ * Formats a raw Base64 string or unformatted X.509 certificate into standard PEM format.
+ * @param {string} certStr
+ * @returns {string}
+ */
+export function formatX509Certificate(certStr) {
+  if (!certStr || typeof certStr !== "string") return "";
+  const clean = certStr
+    .replace(/-----BEGIN CERTIFICATE-----/gi, "")
+    .replace(/-----END CERTIFICATE-----/gi, "")
+    .replace(/[^A-Za-z0-9+/=]/g, "");
+
+  if (!clean) return "";
+
+  const lines = clean.match(/.{1,64}/g) || [];
+  return `-----BEGIN CERTIFICATE-----\n${lines.join("\n")}\n-----END CERTIFICATE-----`;
+}
+
+/**
+ * Checks whether SAML configuration has essential parameters (entryPoint & cert).
+ * @param {object} settings
+ * @returns {boolean}
+ */
+export function isSamlConfigured(settings) {
+  return Boolean(settings?.samlEntryPoint && settings?.samlCert);
+}
+
+/**
+ * Fetches settings and returns runtime status + settings.
+ * @returns {Promise<{ configured: boolean, settings: object }>}
+ */
+export async function getSamlRuntimeConfig() {
+  const settings = await getSettings();
+  return {
+    configured: isSamlConfigured(settings),
+    settings,
+  };
+}
+
+/**
+ * Creates a configured `@node-saml/node-saml` SAML instance with security defaults.
+ * @param {object} settings
+ * @param {string} origin
+ * @returns {SAML}
+ */
+const DUMMY_FALLBACK_CERT =
+  "-----BEGIN CERTIFICATE-----\nMIIC...DUMMY...\n-----END CERTIFICATE-----";
+
+function trimTrailingSlashes(str) {
+  return (str || "").replace(/\/+$/, "");
+}
+
+/**
+ * Resolves the public Base URL / Origin for SAML requests.
+ * Respects settings.baseUrl, process.env.BASE_URL, x-forwarded-proto, and x-forwarded-host.
+ * @param {Request} request
+ * @param {object} settings
+ * @returns {string}
+ */
+export function getSamlBaseUrl(request, settings) {
+  const configuredBaseUrl =
+    (settings?.baseUrl || "").trim() ||
+    process.env.BASE_URL ||
+    process.env.NEXT_PUBLIC_BASE_URL ||
+    "";
+
+  if (configuredBaseUrl) {
+    return trimTrailingSlashes(configuredBaseUrl);
+  }
+
+  if (request) {
+    const forwardedProto = request?.headers?.get?.("x-forwarded-proto") || "";
+    const forwardedHost = request?.headers?.get?.("x-forwarded-host") || "";
+    const host = forwardedHost || request?.headers?.get?.("host") || "";
+    if (host) {
+      const protocol = (forwardedProto || new URL(request.url).protocol || "http:").replace(/:$/, "");
+      return `${protocol}://${host}`.replace(/\/+$/, "");
+    }
+    if (request.url) {
+      return trimTrailingSlashes(new URL(request.url).origin);
+    }
+  }
+
+  return "http://localhost:20128";
+}
+
+export function createSamlInstance(settings, origin) {
+  const cert = formatX509Certificate(settings?.samlCert || "") || DUMMY_FALLBACK_CERT;
+  const callbackUrl = `${origin}/api/auth/saml/acs`;
+  return new SAML({
+    entryPoint: settings?.samlEntryPoint || "https://example.com/sso",
+    issuer: settings?.samlIssuer || "urn:9router:sp",
+    idpCert: cert,
+    cert: cert,
+    callbackUrl: callbackUrl,
+    acceptedClockSkewMs: 60000,
+    wantAssertionsSigned: true,
+    validateInResponseTo: "never",
+    requestIdExpirationMs: 28800000, // 8 hours
+  });
+}
+
+/**
+ * Builds SAML AuthnRequest redirect URL and returns { authorizeUrl, requestId }.
+ * @param {Request} request
+ * @param {object} settings
+ * @returns {Promise<{ authorizeUrl: string, requestId: string }>}
+ */
+export async function buildSamlAuthorizeUrl(request, settings) {
+  const origin = getSamlBaseUrl(request, settings);
+  const samlInstance = createSamlInstance(settings, origin);
+
+  const xml = await samlInstance.generateAuthorizeRequestAsync(false, false);
+  const match = xml.match(/ID="([^"]+)"/);
+  const requestId = match ? match[1] : "";
+
+  const authorizeUrl = await samlInstance._requestToUrlAsync(xml, null, "authorize", {});
+
+  return { authorizeUrl, requestId };
+}
+
+/**
+ * Validates SAML POST response from IdP ACS callback and returns user profile.
+ * @param {Request} request
+ * @param {object} body - Parsed form body or object containing SAMLResponse
+ * @param {string} expectedRequestId - Request ID stored in saml_state cookie
+ * @param {object} settings
+ * @returns {Promise<object>}
+ */
+export async function validateSamlResponse(request, body, expectedRequestId, settings) {
+  if (!settings?.samlCert) {
+    throw new Error("IdP X.509 Certificate (samlCert) is missing or not configured");
+  }
+
+  const origin = getSamlBaseUrl(request, settings);
+  const samlInstance = createSamlInstance(settings, origin);
+
+  const container = typeof body === "object" && body !== null ? body : { SAMLResponse: body };
+  const rawSamlResponse = container.SAMLResponse;
+
+  if (!rawSamlResponse) {
+    throw new Error("Missing SAMLResponse parameter in assertion POST body");
+  }
+
+  // Parse response XML to inspect InResponseTo for replay protection
+  if (expectedRequestId) {
+    const xml = Buffer.from(rawSamlResponse, "base64").toString("utf8");
+    const match = xml.match(/InResponseTo=["']([^"']+)["']/i);
+    const inResponseTo = match ? match[1] : null;
+
+    if (!inResponseTo || inResponseTo !== expectedRequestId) {
+      throw new Error(`InResponseTo mismatch: expected ${expectedRequestId}, received ${inResponseTo || "none"}`);
+    }
+  }
+
+  const result = await samlInstance.validatePostResponseAsync({ SAMLResponse: rawSamlResponse });
+  const profile = result?.profile || result;
+
+  return profile;
+}
+
+/**
+ * Generates standard SP XML Metadata.
+ * @param {string} origin
+ * @param {object} settings
+ * @returns {string}
+ */
+export function generateSamlMetadata(origin, settings) {
+  const samlInstance = createSamlInstance(settings, origin);
+  return samlInstance.generateServiceProviderMetadata();
+}
+
+/**
+ * Extracts email claim from SAML profile assertion.
+ * @param {object} profile
+ * @param {object} settings
+ * @returns {string}
+ */
+export function pickSamlEmail(profile = {}, settings = {}) {
+  if (!profile) return "";
+
+  // 1. Configured custom attribute
+  const customAttr = settings.samlAttributeEmail;
+  if (customAttr && profile[customAttr]) {
+    const val = profile[customAttr];
+    return Array.isArray(val) ? val[0] : String(val);
+  }
+
+  // 2. Common email claims
+  const emailKeys = [
+    "email",
+    "emailAddress",
+    "mail",
+    "nameID",
+    "nameId",
+    "upn",
+    "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/emailaddress",
+    "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/nameidentifier",
+    "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/upn",
+  ];
+
+  for (const key of emailKeys) {
+    if (profile[key]) {
+      const val = profile[key];
+      return Array.isArray(val) ? val[0] : String(val);
+    }
+  }
+
+  // 3. Fallback: check attributes object if present
+  if (profile.attributes) {
+    for (const key of emailKeys) {
+      if (profile.attributes[key]) {
+        const val = profile.attributes[key];
+        return Array.isArray(val) ? val[0] : String(val);
+      }
+    }
+  }
+
+  return "";
+}
+
+/**
+ * Extracts display name claim from SAML profile assertion.
+ * @param {object} profile
+ * @param {object} settings
+ * @returns {string}
+ */
+export function pickSamlDisplayName(profile = {}, settings = {}) {
+  if (!profile) return "";
+
+  // 1. Configured custom attribute
+  const customAttr = settings.samlAttributeName;
+  if (customAttr && profile[customAttr]) {
+    const val = profile[customAttr];
+    return Array.isArray(val) ? val[0] : String(val);
+  }
+
+  // 2. Common name claims
+  const nameKeys = [
+    "displayName",
+    "name",
+    "cn",
+    "commonName",
+    "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/name",
+    "http://schemas.xmlsoap.org/ws/2005/05/identity/claims/givenname",
+  ];
+
+  for (const key of nameKeys) {
+    if (profile[key]) {
+      const val = profile[key];
+      return Array.isArray(val) ? val[0] : String(val);
+    }
+  }
+
+  // 3. Combined givenName + surname
+  if (profile.givenName || profile.sn || profile.surname) {
+    const given = profile.givenName || "";
+    const surname = profile.sn || profile.surname || "";
+    const combined = `${given} ${surname}`.trim();
+    if (combined) return combined;
+  }
+
+  // 4. Fallback to email
+  return pickSamlEmail(profile, settings);
+}

src/lib/auth/trustedPeer.js

@@ -0,0 +1,7 @@
+// x-9r-real-ip is only trustworthy when custom-server.js stamped it from the TCP socket.
+// It proves that by echoing the per-process secret it generated at boot, which a client
+// cannot guess. Without the proof the header is just attacker-supplied input.
+export function hasTrustedPeerHeaders(request) {
+  const token = process.env.NINEROUTER_PEER_TOKEN;
+  return Boolean(token) && request.headers.get("x-9r-peer-token") === token;
+}

src/lib/db/repos/requestDetailsRepo.js

@@ -68,6 +68,8 @@ function sanitizeHeaders(headers) {
   return sanitized;
 }
 
+export const __test__ = { sanitizeHeaders };
+
 function generateDetailId(model) {
   const timestamp = new Date().toISOString();
   const random = Math.random().toString(36).substring(2, 8);

src/lib/db/repos/settingsRepo.js

@@ -27,11 +27,18 @@ const DEFAULT_SETTINGS = {
   requireApiKey: true,
   tunnelDashboardAccess: true,
   authMode: "password",
+  ssoType: "oidc",
   oidcIssuerUrl: "",
   oidcClientId: "",
   oidcClientSecret: "",
   oidcScopes: "openid profile email",
   oidcLoginLabel: "Sign in with OIDC",
+  samlEntryPoint: "",
+  samlIssuer: "urn:9router:sp",
+  samlCert: "",
+  samlLoginLabel: "Sign in with SAML SSO",
+  samlAttributeEmail: "email",
+  samlAttributeName: "name",
   enableObservability: false,
   observabilityMaxRecords: 1000,
   observabilityBatchSize: 20,
@@ -63,7 +70,7 @@ async function readRaw() {
 }
 
 // Merge raw settings with defaults; backward-compat for missing keys
-function mergeWithDefaults(raw) {
+export function mergeWithDefaults(raw) {
   const merged = { ...DEFAULT_SETTINGS, ...(raw || {}) };
   for (const [key, defVal] of Object.entries(DEFAULT_SETTINGS)) {
     if (merged[key] === undefined) {

src/mitm/config.js

@@ -26,10 +26,28 @@ const TARGET_HOSTS = [
 const URL_PATTERNS = {
   antigravity: [":generateContent", ":streamGenerateContent"],
   copilot: ["/chat/completions", "/v1/messages", "/responses"],
+  // Legacy path form. Kiro IDE 1.0.228+ posts to `/` with x-amz-target instead —
+  // see isChatRequest() for the header-based match.
   kiro: ["/generateAssistantResponse"],
   cursor: ["/BidiAppend", "/RunSSE", "/RunPoll", "/Run"],
 };
 
+/**
+ * Whether this request is a chat turn we should intercept (vs passthrough).
+ * Kiro Runtime moved GenerateAssistantResponse from path `/generateAssistantResponse`
+ * to `POST /` + `x-amz-target: KiroRuntimeService.GenerateAssistantResponse`
+ * (verified via live mitmproxy capture of Kiro IDE 1.0.228).
+ */
+function isChatRequest(tool, req) {
+  const patterns = URL_PATTERNS[tool] || [];
+  if (patterns.some((p) => (req.url || "").includes(p))) return true;
+  if (tool === "kiro") {
+    const target = String(req.headers?.["x-amz-target"] || "");
+    return target.includes("GenerateAssistantResponse");
+  }
+  return false;
+}
+
 // Synonym map: rawModel from request → canonical alias key in mitmAlias DB
 const MODEL_SYNONYMS = {
   antigravity: {
@@ -37,6 +55,9 @@ const MODEL_SYNONYMS = {
     "gemini-3.5-flash-high": "gemini-3-flash-agent",
     "gemini-3.5-flash-medium": "gemini-3.5-flash-low",
     "gemini-3.5-flash-extra-low": "gemini-3.5-flash-extra-low",
+     "gemini-3.7-flash-high": "gemini-3.7-flash-high",
+    "gemini-3.7-flash-medium": "gemini-3.7-flash-medium",
+    "gemini-3.7-flash-low": "gemini-3.7-flash-low",
     "gemini-3.1-pro-high": "gemini-pro-agent",
     "gemini-3-pro-high": "gemini-pro-agent",
     "gemini-3-pro-low": "gemini-3.1-pro-low",
@@ -113,13 +134,15 @@ function extractModel(url, body) {
       return parsed.conversationState.currentMessage?.userInputMessage?.modelId || null;
     }
     const model = urlModel || parsed.model || null;
-    if (String(model).replace(/^models\//, "") === "gemini-3.6-flash-tiered") {
+    const cleanModelName = String(model).replace(/^models\//, "");
+    if (cleanModelName === "gemini-3.6-flash-tiered" || cleanModelName === "gemini-3.7-flash-tiered") {
+      const ver = cleanModelName.includes("3.7") ? "3.7" : "3.6";
       const rawLevel = parsed.request?.generationConfig?.thinkingConfig?.thinkingLevel
         || parsed.generationConfig?.thinkingConfig?.thinkingLevel;
       const level = ["high", "medium", "low"].includes(String(rawLevel).toLowerCase())
         ? String(rawLevel).toLowerCase()
         : "medium";
-      return `gemini-3.6-flash-${level}`;
+      return `gemini-${ver}-flash-${level}`;
     }
     return model;
   } catch {
@@ -127,4 +150,4 @@ function extractModel(url, body) {
   }
 }
 
-module.exports = { IS_DEV, LSOF_BIN, TARGET_HOSTS, URL_PATTERNS, MODEL_SYNONYMS, MODEL_PATTERNS, MODEL_NO_MAP, LOG_BLACKLIST_URL_PARTS, getToolForHost, extractModel };
+module.exports = { IS_DEV, LSOF_BIN, TARGET_HOSTS, URL_PATTERNS, MODEL_SYNONYMS, MODEL_PATTERNS, MODEL_NO_MAP, LOG_BLACKLIST_URL_PARTS, getToolForHost, isChatRequest, extractModel };

src/mitm/handlers/kiro.js

@@ -43,7 +43,8 @@ function initKiroState(modelId) {
     finishSent: false,             // Whether termination has been emitted
     usage: null,                   // Accumulated usage from usage-only chunks
     inThink: false,                // Whether inside a <thinking> block
-    thinkBuf: ""                   // Buffer for partial thinking content
+    thinkBuf: "",                  // Buffer for partial thinking content
+    initialSent: false,            // Whether initial-response frame was emitted
   };
 }
 
@@ -130,9 +131,9 @@ function encodeHeader(name, value) {
  * The SmithyMessageDecoderStream layer requires three system headers on every frame:
  *   :message-type  = "event"             (or "exception" / "error")
  *   :event-type    = e.g. "assistantResponseEvent"
- *   :content-type  = "application/json"
+ *   :content-type  = "application/json"  (initial-response uses x-amz-json-1.0)
  */
-function buildEventStreamFrame(eventType, payload) {
+function buildEventStreamFrame(eventType, payload, contentType = "application/json") {
   const payloadBuf = Buffer.from(
     typeof payload === "string" ? payload : JSON.stringify(payload),
     "utf8"
@@ -142,7 +143,7 @@ function buildEventStreamFrame(eventType, payload) {
   const headersBuf = Buffer.concat([
     encodeHeader(":message-type", "event"),
     encodeHeader(":event-type", eventType),
-    encodeHeader(":content-type", "application/json"),
+    encodeHeader(":content-type", contentType),
   ]);
   const headersLen = headersBuf.length;
 
@@ -159,6 +160,24 @@ function buildEventStreamFrame(eventType, payload) {
   return frame;
 }
 
+/** Real Kiro Runtime always starts the stream with this frame (capture of IDE 1.0.228). */
+function buildInitialResponseFrame(conversationId = "") {
+  return buildEventStreamFrame(
+    "initial-response",
+    { conversationId: conversationId || "" },
+    "application/x-amz-json-1.0"
+  );
+}
+
+/** Prepend initial-response once per stream so Smithy decoder is happy. */
+function withInitialFrame(state, frames) {
+  const list = frames == null ? [] : Array.isArray(frames) ? frames : [frames];
+  if (state.initialSent) return list.length === 0 ? null : list.length === 1 ? list[0] : list;
+  state.initialSent = true;
+  const out = [buildInitialResponseFrame(""), ...list];
+  return out.length === 1 ? out[0] : out;
+}
+
 // ─── CodeWhisperer → OpenAI conversion ───────────────────────────────────────
 
 /**
@@ -321,12 +340,12 @@ function convertOpenAIToKiro(chunk, state) {
       state.inThink = false;
       const thinking = state.thinkBuf;
       state.thinkBuf = "";
-      return buildEventStreamFrame("reasoningContentEvent", {
+      return withInitialFrame(state, buildEventStreamFrame("reasoningContentEvent", {
         content: thinking,
         modelId: state.modelId || "kiro-unknown"
-      });
+      }));
     }
-    return buildEventStreamFrame("messageStopEvent", {});
+    return withInitialFrame(state, buildEventStreamFrame("messageStopEvent", {}));
   }
 
   const frames = [];
@@ -408,8 +427,12 @@ function convertOpenAIToKiro(chunk, state) {
     }
   }
 
-  if (frames.length === 0) return null;
-  return frames.length === 1 ? frames[0] : frames;
+  if (frames.length === 0) {
+    // اولین چانک ممکنه فقط role/empty باشه — initial رو همون‌جا بفرست
+    if (!state.initialSent) return withInitialFrame(state, null);
+    return null;
+  }
+  return withInitialFrame(state, frames.length === 1 ? frames[0] : frames);
 }
 
 /**

src/mitm/server.js

@@ -7,7 +7,7 @@ const dns = require("dns");
 const { promisify } = require("util");
 const { execSync } = require("child_process");
 const { log, err, dumpRequest, createResponseDumper, clearDumpDir } = require("./logger");
-const { IS_DEV, LSOF_BIN, TARGET_HOSTS, URL_PATTERNS, MODEL_SYNONYMS, MODEL_PATTERNS, MODEL_NO_MAP, getToolForHost, extractModel } = require("./config");
+const { IS_DEV, LSOF_BIN, TARGET_HOSTS, URL_PATTERNS, MODEL_SYNONYMS, MODEL_PATTERNS, MODEL_NO_MAP, getToolForHost, isChatRequest, extractModel } = require("./config");
 const { DATA_DIR, MITM_DIR } = require("./paths");
 const { generateCert, getCertForDomain } = require("./cert/generate");
 const { getMitmAlias } = require("./dbReader");
@@ -311,9 +311,8 @@ const server = https.createServer(sslOptions, async (req, res) => {
     const tool = getToolForHost(req.headers.host);
     if (!tool) return passthrough(req, res, bodyBuffer);
 
-    const patterns = URL_PATTERNS[tool] || [];
-    const isChat = patterns.some(p => req.url.includes(p));
-    if (!isChat) return passthrough(req, res, bodyBuffer);
+    // Kiro IDE posts chat to `/` with x-amz-target (not path /generateAssistantResponse)
+    if (!isChatRequest(tool, req)) return passthrough(req, res, bodyBuffer);
 
     // Cursor uses binary proto — model extraction not possible at this layer.
     // Delegate directly to handler which decodes proto internally.

src/shared/components/Header.js

@@ -198,7 +198,7 @@ export default function Header({ onMenuClick, showMenuButton = true }) {
         if (!res.ok) return;
         const data = await res.json();
         if (!cancelled) {
-          setDisplayName(data?.displayName || data?.oidcName || data?.oidcEmail || "");
+          setDisplayName(data?.displayName || data?.samlName || data?.samlEmail || data?.oidcName || data?.oidcEmail || "");
           setLoginMethod(data?.loginMethod || "");
         }
       } catch {
@@ -303,12 +303,15 @@ export default function Header({ onMenuClick, showMenuButton = true }) {
 
       {/* Right actions */}
       <div className="flex items-center gap-1 shrink-0">
-        {displayName && loginMethod === "OIDC" && (
-          <div className="hidden sm:flex items-center max-w-[220px] px-3 py-1.5 rounded-full border border-border bg-surface/70 text-xs text-text-muted truncate">
+        {displayName && (loginMethod === "OIDC" || loginMethod === "SAML") && (
+          <div
+            className="hidden sm:flex items-center max-w-[220px] px-3 py-1.5 rounded-full border border-border bg-surface/70 text-xs text-text-muted truncate"
+            title={displayName}
+          >
             <span className="material-symbols-outlined text-[14px] mr-1.5 text-primary">person</span>
             <span className="truncate">{displayName}</span>
             <span className="ml-2 shrink-0 rounded-full bg-primary/10 px-2 py-0.5 text-[10px] font-semibold uppercase tracking-wide text-primary">
-              OIDC
+              {loginMethod}
             </span>
           </div>
         )}

src/shared/constants/cliTools.js

@@ -8,7 +8,7 @@ export const MITM_TOOLS = {
     description: "Google Antigravity IDE with MITM",
     configType: "mitm",
     mitmDomain: "daily-cloudcode-pa.googleapis.com",
-    modelAliases: ["gemini-3.6-flash-high", "gemini-3.6-flash-medium", "gemini-3.6-flash-low", "gemini-3.5-flash-low", "gemini-3-flash-agent", "gemini-3.5-flash-extra-low", "gemini-3.1-pro-low", "gemini-pro-agent", "claude-sonnet-4-6", "claude-opus-4-6-thinking", "gpt-oss-120b-medium", "gemini-3-flash"],
+    modelAliases: ["gemini-3.7-flash-high", "gemini-3.7-flash-medium", "gemini-3.7-flash-low", "gemini-3.6-flash-high", "gemini-3.6-flash-medium", "gemini-3.6-flash-low", "gemini-3.5-flash-low", "gemini-3-flash-agent", "gemini-3.5-flash-extra-low", "gemini-3.1-pro-low", "gemini-pro-agent", "claude-sonnet-4-6", "claude-opus-4-6-thinking", "gpt-oss-120b-medium", "gemini-3-flash"],
     defaultModels: [
       { id: "gemini-3.6-flash-high", name: "Gemini 3.6 Flash (High)", alias: "gemini-3.6-flash-high" },
       { id: "gemini-3.6-flash-medium", name: "Gemini 3.6 Flash (Medium)", alias: "gemini-3.6-flash-medium" },
@@ -57,8 +57,13 @@ export const MITM_TOOLS = {
     color: "#FF6B00",
     description: "Kiro IDE with MITM",
     configType: "mitm",
-    mitmDomain: "q.us-east-1.amazonaws.com",
+    mitmDomain: "runtime.us-east-1.kiro.dev",
     defaultModels: [
+      // Kiro's agent/"vibe" mode sends modelId "auto" for the main turn and "simple-task"
+      // for background sub-tasks (verified via MITM request dump of generateAssistantResponse).
+      // Both need a mappable slot — otherwise getMappedModel returns null and the chat call
+      // is passed through to AWS instead of being routed to the chosen provider.
+      { id: "auto", name: "Auto (Kiro Agent)", alias: "auto" },
       { id: "claude-sonnet-5", name: "Claude Sonnet 5", alias: "claude-sonnet-5" },
       { id: "claude-sonnet-4.5", name: "Claude Sonnet 4.5", alias: "claude-sonnet-4.5" },
       { id: "claude-sonnet-4", name: "Claude Sonnet 4", alias: "claude-sonnet-4" },

tests/__baseline__/alias-baseline.json

@@ -122,6 +122,7 @@
     "alicode": "alicode",
     "alicode-intl": "alicode-intl",
     "alims-intl": "alims-intl",
+    "alitp-intl": "alitp-intl",
     "anthropic": "anthropic",
     "antigravity": "ag",
     "api-airforce": "af",
@@ -206,6 +207,7 @@
     "alicode",
     "alicode-intl",
     "alims-intl",
+    "alitp-intl",
     "anthropic",
     "assemblyai",
     "black-forest-labs",

tests/__baseline__/providers-baseline.json

@@ -708,7 +708,37 @@
   "opencode-go": {
     "baseUrl": "https://opencode.ai/zen/go/v1/chat/completions",
     "headers": {},
-    "format": "openai"
+    "format": "openai",
+    "transports": [
+      {
+        "format": "openai",
+        "baseUrl": "https://opencode.ai/zen/go/v1/chat/completions",
+        "auth": {
+          "combined": true,
+          "header": "Authorization",
+          "scheme": "bearer"
+        }
+      },
+      {
+        "format": "claude",
+        "baseUrl": "https://opencode.ai/zen/go/v1/messages",
+        "auth": {
+          "combined": true,
+          "header": "x-api-key",
+          "scheme": "raw",
+          "anthropicVersion": true
+        }
+      },
+      {
+        "format": "openai-responses",
+        "baseUrl": "https://opencode.ai/zen/go/v1/responses",
+        "auth": {
+          "combined": true,
+          "header": "Authorization",
+          "scheme": "bearer"
+        }
+      }
+    ]
   },
   "opencode": {
     "baseUrl": "https://opencode.ai",
@@ -968,7 +998,15 @@
   "tokenrouter": {
     "baseUrl": "https://api.tokenrouter.com/v1/chat/completions",
     "validateUrl": "https://api.tokenrouter.com/v1/models",
-    "thinkingFormat": "openai",
+    "thinkingFormat": "tokenrouter",
+    "format": "openai"
+  },
+  "alitp-intl": {
+    "baseUrl": "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions",
+    "headers": {},
+    "quirks": {
+      "preserveCacheControl": true
+    },
     "format": "openai"
   }
 }

tests/auth/saml.test.js

@@ -0,0 +1,40 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import {
+  formatX509Certificate,
+  isSamlConfigured,
+  generateSamlMetadata,
+  pickSamlEmail,
+  pickSamlDisplayName,
+} from "../../src/lib/auth/saml.js";
+
+test("formatX509Certificate normalizes Base64 strings into PEM blocks", () => {
+  const rawBase64 = "MIIC1234567890123456789012345678901234567890123456789012345678901234567890";
+  const formatted = formatX509Certificate(rawBase64);
+  assert.match(formatted, /-----BEGIN CERTIFICATE-----/);
+  assert.match(formatted, /-----END CERTIFICATE-----/);
+  assert.equal(formatX509Certificate(""), "");
+});
+
+test("isSamlConfigured checks required fields", () => {
+  assert.equal(isSamlConfigured({ samlEntryPoint: "https://idp.com/sso", samlCert: "cert" }), true);
+  assert.equal(isSamlConfigured({ samlEntryPoint: "https://idp.com/sso" }), false);
+  assert.equal(isSamlConfigured({}), false);
+});
+
+test("generateSamlMetadata produces valid SP XML", () => {
+  const settings = {
+    samlEntryPoint: "https://idp.example.com/sso",
+    samlIssuer: "urn:9router:sp",
+    samlCert: "MIIC123456789012345678901234567890123456789012345678901234567890",
+  };
+  const xml = generateSamlMetadata("https://localhost:20127", settings);
+  assert.match(xml, /entityID="urn:9router:sp"/);
+  assert.match(xml, /Location="https:\/\/localhost:20127\/api\/auth\/saml\/acs"/);
+});
+
+test("Claims Extraction pickSamlEmail & pickSamlDisplayName", () => {
+  const profile = { email: "test@example.com", name: "Test User" };
+  assert.equal(pickSamlEmail(profile, {}), "test@example.com");
+  assert.equal(pickSamlDisplayName(profile, {}), "Test User");
+});

tests/translator/__snapshots__/golden-url-header.test.js.snap

@@ -38,6 +38,25 @@ exports[`GOLDEN buildHeaders (default executor providers) > alicode-intl → hea
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > alims-intl → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > anthropic → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -66,6 +85,31 @@ exports[`GOLDEN buildHeaders (default executor providers) > anthropic → header
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > api-airforce → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "HTTP-Referer": "https://endpoint-proxy.local",
+    "X-Title": "Endpoint Proxy",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "HTTP-Referer": "https://endpoint-proxy.local",
+    "X-Title": "Endpoint Proxy",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "HTTP-Referer": "https://endpoint-proxy.local",
+    "X-Title": "Endpoint Proxy",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > assemblyai → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -85,6 +129,44 @@ exports[`GOLDEN buildHeaders (default executor providers) > assemblyai → heade
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > baidu → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
+exports[`GOLDEN buildHeaders (default executor providers) > bazaarlink → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > blackbox → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -104,6 +186,25 @@ exports[`GOLDEN buildHeaders (default executor providers) > blackbox → headers
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > bluesminds → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > byteplus → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -268,6 +369,52 @@ exports[`GOLDEN buildHeaders (default executor providers) > cline → headers (a
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > clinepass → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "HTTP-Referer": "https://cline.bot",
+    "User-Agent": "9Router/0.5.50",
+    "X-CLIENT-TYPE": "9router",
+    "X-CLIENT-VERSION": "0.5.50",
+    "X-CORE-VERSION": "0.5.50",
+    "X-IS-MULTIROOT": "false",
+    "X-PLATFORM": "linux",
+    "X-PLATFORM-VERSION": "v24.15.0",
+    "X-Title": "Cline",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "HTTP-Referer": "https://cline.bot",
+    "User-Agent": "9Router/0.5.50",
+    "X-CLIENT-TYPE": "9router",
+    "X-CLIENT-VERSION": "0.5.50",
+    "X-CORE-VERSION": "0.5.50",
+    "X-IS-MULTIROOT": "false",
+    "X-PLATFORM": "linux",
+    "X-PLATFORM-VERSION": "v24.15.0",
+    "X-Title": "Cline",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "HTTP-Referer": "https://cline.bot",
+    "User-Agent": "9Router/0.5.50",
+    "X-CLIENT-TYPE": "9router",
+    "X-CLIENT-VERSION": "0.5.50",
+    "X-CORE-VERSION": "0.5.50",
+    "X-IS-MULTIROOT": "false",
+    "X-PLATFORM": "linux",
+    "X-PLATFORM-VERSION": "v24.15.0",
+    "X-Title": "Cline",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > cloudflare-ai → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -324,6 +471,43 @@ exports[`GOLDEN buildHeaders (default executor providers) > codebuddy-cn → hea
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > codebuddy-intl → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "IDE/2.108.1 CodeBuddy/2.108.1",
+    "X-IDE-Name": "IDE",
+    "X-IDE-Type": "IDE",
+    "X-Product": "SaaS",
+    "x-codebuddy-request": "1",
+    "x-requested-with": "XMLHttpRequest",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "IDE/2.108.1 CodeBuddy/2.108.1",
+    "X-IDE-Name": "IDE",
+    "X-IDE-Type": "IDE",
+    "X-Product": "SaaS",
+    "x-codebuddy-request": "1",
+    "x-requested-with": "XMLHttpRequest",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "IDE/2.108.1 CodeBuddy/2.108.1",
+    "X-IDE-Name": "IDE",
+    "X-IDE-Type": "IDE",
+    "X-Product": "SaaS",
+    "x-codebuddy-request": "1",
+    "x-requested-with": "XMLHttpRequest",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > cohere → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -381,6 +565,25 @@ exports[`GOLDEN buildHeaders (default executor providers) > deepseek → headers
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > featherless → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > fireworks → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -482,6 +685,34 @@ exports[`GOLDEN buildHeaders (default executor providers) > glm-cn → headers (
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > grok-cli → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "grok-shell/0.2.99 (linux; x86_64)",
+    "x-grok-client-identifier": "grok-shell",
+    "x-grok-client-version": "0.2.99",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "grok-shell/0.2.99 (linux; x86_64)",
+    "x-grok-client-identifier": "grok-shell",
+    "x-grok-client-version": "0.2.99",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "grok-shell/0.2.99 (linux; x86_64)",
+    "x-grok-client-identifier": "grok-shell",
+    "x-grok-client-version": "0.2.99",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > groq → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -520,6 +751,25 @@ exports[`GOLDEN buildHeaders (default executor providers) > hyperbolic → heade
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > kilo-gateway → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > kilocode → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -539,6 +789,28 @@ exports[`GOLDEN buildHeaders (default executor providers) > kilocode → headers
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > kimchi → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "kimchi/0.1.50",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "kimchi/0.1.50",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+    "User-Agent": "kimchi/0.1.50",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > kimi → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -601,6 +873,25 @@ exports[`GOLDEN buildHeaders (default executor providers) > kimi-coding → head
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > llm7 → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > minimax → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -689,6 +980,25 @@ exports[`GOLDEN buildHeaders (default executor providers) > mmf → headers (api
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > morph → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > nanobanana → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -828,6 +1138,63 @@ exports[`GOLDEN buildHeaders (default executor providers) > perplexity → heade
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > perplexity-agent → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
+exports[`GOLDEN buildHeaders (default executor providers) > poolside → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
+exports[`GOLDEN buildHeaders (default executor providers) > sambanova → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > siliconflow → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -847,6 +1214,25 @@ exports[`GOLDEN buildHeaders (default executor providers) > siliconflow → head
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > tencent → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > together → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -866,6 +1252,44 @@ exports[`GOLDEN buildHeaders (default executor providers) > together → headers
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > tokenrouter → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
+exports[`GOLDEN buildHeaders (default executor providers) > venice → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "Bearer <TOK>",
+    "Content-Type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildHeaders (default executor providers) > vercel-ai-gateway → headers (apiKey / oauth) 1`] = `
 {
   "apiKey": {
@@ -942,6 +1366,28 @@ exports[`GOLDEN buildHeaders (default executor providers) > xiaomi-mimo → head
 }
 `;
 
+exports[`GOLDEN buildHeaders (default executor providers) > zed → headers (apiKey / oauth) 1`] = `
+{
+  "apiKey": {
+    "Accept": "text/event-stream",
+    "Authorization": "<CRED>",
+    "Content-Type": "application/json",
+    "content-type": "application/json",
+  },
+  "nonStream": {
+    "Authorization": "<CRED>",
+    "Content-Type": "application/json",
+    "content-type": "application/json",
+  },
+  "oauth": {
+    "Accept": "text/event-stream",
+    "Authorization": "<CRED>",
+    "Content-Type": "application/json",
+    "content-type": "application/json",
+  },
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > alicode → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://coding.dashscope.aliyuncs.com/v1/chat/completions",
@@ -956,6 +1402,13 @@ exports[`GOLDEN buildUrl (default executor providers) > alicode-intl → url (st
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > alims-intl → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions",
+  "stream": "https://dashscope-intl.aliyuncs.com/compatible-mode/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > anthropic → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.anthropic.com/v1/messages",
@@ -963,6 +1416,13 @@ exports[`GOLDEN buildUrl (default executor providers) > anthropic → url (strea
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > api-airforce → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.airforce/v1/chat/completions",
+  "stream": "https://api.airforce/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > assemblyai → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.assemblyai.com/v1/audio/transcriptions",
@@ -970,6 +1430,20 @@ exports[`GOLDEN buildUrl (default executor providers) > assemblyai → url (stre
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > baidu → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://qianfan.baidubce.com/v2/chat/completions",
+  "stream": "https://qianfan.baidubce.com/v2/chat/completions",
+}
+`;
+
+exports[`GOLDEN buildUrl (default executor providers) > bazaarlink → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://bazaarlink.ai/api/v1/chat/completions",
+  "stream": "https://bazaarlink.ai/api/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > blackbox → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.blackbox.ai/chat/completions",
@@ -977,6 +1451,13 @@ exports[`GOLDEN buildUrl (default executor providers) > blackbox → url (stream
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > bluesminds → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.bluesminds.com/v1/chat/completions",
+  "stream": "https://api.bluesminds.com/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > byteplus → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://ark.ap-southeast.bytepluses.com/api/coding/v3/chat/completions",
@@ -1012,6 +1493,13 @@ exports[`GOLDEN buildUrl (default executor providers) > cline → url (stream +
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > clinepass → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.cline.bot/api/v1/chat/completions",
+  "stream": "https://api.cline.bot/api/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > cloudflare-ai → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.cloudflare.com/client/v4/accounts/ACC123/ai/v1/chat/completions",
@@ -1026,6 +1514,13 @@ exports[`GOLDEN buildUrl (default executor providers) > codebuddy-cn → url (st
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > codebuddy-intl → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://www.codebuddy.ai/v2/chat/completions",
+  "stream": "https://www.codebuddy.ai/v2/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > cohere → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.cohere.ai/v1/chat/completions",
@@ -1047,6 +1542,13 @@ exports[`GOLDEN buildUrl (default executor providers) > deepseek → url (stream
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > featherless → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.featherless.ai/v1/chat/completions",
+  "stream": "https://api.featherless.ai/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > fireworks → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.fireworks.ai/inference/v1/chat/completions",
@@ -1082,6 +1584,13 @@ exports[`GOLDEN buildUrl (default executor providers) > glm-cn → url (stream +
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > grok-cli → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://cli-chat-proxy.grok.com/v1/responses",
+  "stream": "https://cli-chat-proxy.grok.com/v1/responses",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > groq → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.groq.com/openai/v1/chat/completions",
@@ -1096,6 +1605,13 @@ exports[`GOLDEN buildUrl (default executor providers) > hyperbolic → url (stre
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > kilo-gateway → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.kilo.ai/api/gateway/chat/completions",
+  "stream": "https://api.kilo.ai/api/gateway/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > kilocode → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.kilo.ai/api/openrouter/chat/completions",
@@ -1103,6 +1619,13 @@ exports[`GOLDEN buildUrl (default executor providers) > kilocode → url (stream
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > kimchi → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://llm.kimchi.dev/openai/v1/chat/completions",
+  "stream": "https://llm.kimchi.dev/openai/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > kimi → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.kimi.com/coding/v1/messages?beta=true",
@@ -1117,6 +1640,13 @@ exports[`GOLDEN buildUrl (default executor providers) > kimi-coding → url (str
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > llm7 → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.llm7.io/v1/chat/completions",
+  "stream": "https://api.llm7.io/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > minimax → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.minimax.io/anthropic/v1/messages?beta=true",
@@ -1145,6 +1675,13 @@ exports[`GOLDEN buildUrl (default executor providers) > mmf → url (stream + no
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > morph → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.morphllm.com/v1/chat/completions",
+  "stream": "https://api.morphllm.com/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > nanobanana → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.nanobananaapi.ai/v1/chat/completions",
@@ -1194,6 +1731,27 @@ exports[`GOLDEN buildUrl (default executor providers) > perplexity → url (stre
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > perplexity-agent → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.perplexity.ai/v1/responses",
+  "stream": "https://api.perplexity.ai/v1/responses",
+}
+`;
+
+exports[`GOLDEN buildUrl (default executor providers) > poolside → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://inference.poolside.ai/v1/chat/completions",
+  "stream": "https://inference.poolside.ai/v1/chat/completions",
+}
+`;
+
+exports[`GOLDEN buildUrl (default executor providers) > sambanova → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.sambanova.ai/v1/chat/completions",
+  "stream": "https://api.sambanova.ai/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > siliconflow → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.siliconflow.com/v1/chat/completions",
@@ -1201,6 +1759,13 @@ exports[`GOLDEN buildUrl (default executor providers) > siliconflow → url (str
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > tencent → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.hunyuan.cloud.tencent.com/v1/chat/completions",
+  "stream": "https://api.hunyuan.cloud.tencent.com/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > together → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://api.together.xyz/v1/chat/completions",
@@ -1208,6 +1773,20 @@ exports[`GOLDEN buildUrl (default executor providers) > together → url (stream
 }
 `;
 
+exports[`GOLDEN buildUrl (default executor providers) > tokenrouter → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.tokenrouter.com/v1/chat/completions",
+  "stream": "https://api.tokenrouter.com/v1/chat/completions",
+}
+`;
+
+exports[`GOLDEN buildUrl (default executor providers) > venice → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://api.venice.ai/api/v1/chat/completions",
+  "stream": "https://api.venice.ai/api/v1/chat/completions",
+}
+`;
+
 exports[`GOLDEN buildUrl (default executor providers) > vercel-ai-gateway → url (stream + non-stream) 1`] = `
 {
   "nonStream": "https://ai-gateway.vercel.sh/v1/chat/completions",
@@ -1235,3 +1814,10 @@ exports[`GOLDEN buildUrl (default executor providers) > xiaomi-mimo → url (str
   "stream": "https://api.xiaomimimo.com/v1/chat/completions",
 }
 `;
+
+exports[`GOLDEN buildUrl (default executor providers) > zed → url (stream + non-stream) 1`] = `
+{
+  "nonStream": "https://cloud.zed.dev/completions",
+  "stream": "https://cloud.zed.dev/completions",
+}
+`;

tests/unit/alibaba-token-plan-provider.test.js

@@ -0,0 +1,45 @@
+import { describe, expect, it } from "vitest";
+
+import REGISTRY from "../../open-sse/providers/registry/index.js";
+import { PROVIDERS, PROVIDER_MODELS } from "../../open-sse/providers/index.js";
+
+describe("Alibaba Token Plan provider", () => {
+  const entry = REGISTRY.find((e) => e.id === "alitp-intl");
+
+  it("is registered as an OpenAI-compatible apikey provider", () => {
+    expect(entry).toBeDefined();
+    expect(entry.category).toBe("apikey");
+    expect(PROVIDERS["alitp-intl"]).toBeDefined();
+    expect(PROVIDERS["alitp-intl"].format).toBe("openai");
+  });
+
+  it("targets the Singapore Token Plan host in compatible mode", () => {
+    // eu-central-1 answers IllegalEndpoint; the plan is Singapore-only.
+    expect(PROVIDERS["alitp-intl"].baseUrl).toBe(
+      "https://token-plan.ap-southeast-1.maas.aliyuncs.com/compatible-mode/v1/chat/completions",
+    );
+  });
+
+  it("does not collide with the other three Alibaba key types", () => {
+    const hosts = ["alicode", "alicode-intl", "alims-intl", "alitp-intl"]
+      .map((id) => new URL(PROVIDERS[id].baseUrl).host);
+    expect(new Set(hosts).size).toBe(hosts.length);
+  });
+
+  it("exposes the models the plan actually serves", () => {
+    const ids = (PROVIDER_MODELS["alitp-intl"] || []).map((m) => m.id);
+    expect(ids).toEqual(expect.arrayContaining([
+      "qwen3.8-max-preview",
+      "qwen3.7-max",
+      "qwen3.7-plus",
+      "qwen3.6-flash",
+      "glm-5.2",
+      "deepseek-v4-pro",
+    ]));
+  });
+
+  it("keeps every registry id unique after adding the provider", () => {
+    const ids = REGISTRY.map((e) => e.id);
+    expect(new Set(ids).size).toBe(ids.length);
+  });
+});

tests/unit/antigravity-nonstream-usage-3260.test.js

@@ -0,0 +1,54 @@
+import { describe, expect, it, vi } from "vitest";
+
+vi.mock("@/lib/usageDb.js", () => ({
+  appendRequestLog: vi.fn(async () => {}),
+  saveRequestDetail: vi.fn(async () => {}),
+  saveRequestUsage: vi.fn(async () => {})
+}));
+
+const { extractUsageFromResponse } = await import("../../open-sse/handlers/chatCore/requestDetail.js");
+
+const USAGE_METADATA = {
+  promptTokenCount: 1234,
+  candidatesTokenCount: 56,
+  cachedContentTokenCount: 78,
+  thoughtsTokenCount: 90,
+};
+
+const EXPECTED = {
+  prompt_tokens: 1234,
+  completion_tokens: 56,
+  cached_tokens: 78,
+  reasoning_tokens: 90,
+};
+
+describe("#3260 non-streaming usage extraction for enveloped Gemini responses", () => {
+  it("reads usageMetadata out of the antigravity { response } envelope", () => {
+    expect(extractUsageFromResponse({ response: { usageMetadata: USAGE_METADATA } })).toEqual(EXPECTED);
+  });
+
+  it("still reads a top-level usageMetadata", () => {
+    expect(extractUsageFromResponse({ usageMetadata: USAGE_METADATA })).toEqual(EXPECTED);
+  });
+
+  it("prefers the top-level metadata when both are present", () => {
+    const enveloped = { ...USAGE_METADATA, promptTokenCount: 1 };
+    const out = extractUsageFromResponse({
+      usageMetadata: USAGE_METADATA,
+      response: { usageMetadata: enveloped },
+    });
+    expect(out.prompt_tokens).toBe(1234);
+  });
+
+  it("leaves the OpenAI and Claude shapes alone", () => {
+    expect(extractUsageFromResponse({ usage: { prompt_tokens: 10, completion_tokens: 2 } }))
+      .toMatchObject({ prompt_tokens: 10, completion_tokens: 2 });
+    expect(extractUsageFromResponse({ usage: { input_tokens: 10, output_tokens: 2 } }))
+      .toMatchObject({ prompt_tokens: 10, completion_tokens: 2 });
+  });
+
+  it("returns null when there is no usage anywhere", () => {
+    expect(extractUsageFromResponse({ response: { candidates: [] } })).toBeNull();
+    expect(extractUsageFromResponse(null)).toBeNull();
+  });
+});

tests/unit/antigravity-quota-gemini-3.7.test.js

@@ -0,0 +1,61 @@
+import { describe, expect, it, vi, beforeEach } from "vitest";
+
+const proxyAwareFetch = vi.fn(async (url) => ({
+  ok: true,
+  status: 200,
+  json: async () => url.includes(":loadCodeAssist")
+    ? { cloudaicompanionProject: "project-1", currentTier: { name: "Pro" } }
+    : {
+        models: {
+          "gemini-3.7-flash-high": {
+            displayName: "Gemini 3.7 Flash (High)",
+            quotaInfo: { remainingFraction: 0.85, resetTime: "2026-08-25T12:00:00Z" },
+          },
+          "gemini-3.7-flash-medium": {
+            displayName: "Gemini 3.7 Flash (Medium)",
+            quotaInfo: { remainingFraction: 0.6, resetTime: "2026-08-25T12:00:00Z" },
+          },
+          "gemini-3.7-flash-low": {
+            displayName: "Gemini 3.7 Flash (Low)",
+            quotaInfo: { remainingFraction: 0.35, resetTime: "2026-08-25T12:00:00Z" },
+          },
+          "internal-model": {
+            displayName: "Internal",
+            isInternal: true,
+            quotaInfo: { remainingFraction: 0.5 },
+          },
+        },
+      },
+  text: async () => "{}",
+}));
+
+vi.mock("../../open-sse/utils/proxyFetch.js", () => ({
+  proxyAwareFetch,
+}));
+
+describe("Antigravity quota tracker: Gemini 3.7 Flash usage bars", () => {
+  beforeEach(() => proxyAwareFetch.mockClear());
+
+  it("returns Gemini 3.7 Flash tier quotas so the dashboard can render usage bars", async () => {
+    const { getAntigravityUsage } = await import("../../open-sse/services/usage/google.js");
+
+    const usage = await getAntigravityUsage("access-token", {});
+
+    expect(usage.quotas["gemini-3.7-flash-high"]).toMatchObject({
+      used: 150,
+      total: 1000,
+      remainingPercentage: 85,
+      displayName: "Gemini 3.7 Flash (High)",
+    });
+    expect(usage.quotas["gemini-3.7-flash-medium"]).toMatchObject({
+      used: 400,
+      total: 1000,
+      remainingPercentage: 60,
+    });
+    expect(usage.quotas["gemini-3.7-flash-low"]).toMatchObject({
+      used: 650,
+      total: 1000,
+      remainingPercentage: 35,
+    });
+  });
+});

tests/unit/custom-server-peer-headers.test.js

@@ -0,0 +1,86 @@
+// custom-server.js is the only thing that makes x-9r-real-ip trustworthy. Boot a real
+// HTTP server through it and confirm a client cannot smuggle its own peer headers in.
+import { describe, it, expect, beforeAll, afterAll } from "vitest";
+import { createRequire } from "node:module";
+import http from "node:http";
+import { __test__ as requestDetails } from "@/lib/db/repos/requestDetailsRepo.js";
+
+const require = createRequire(import.meta.url);
+
+let server;
+let baseUrl;
+let seenHeaders;
+
+beforeAll(async () => {
+  require("../../custom-server.js");
+  server = http.createServer((req, res) => {
+    seenHeaders = req.headers;
+    res.end("ok");
+  });
+  await new Promise((resolve) => server.listen(0, "127.0.0.1", resolve));
+  baseUrl = `http://127.0.0.1:${server.address().port}`;
+});
+
+afterAll(async () => {
+  await new Promise((resolve) => server.close(resolve));
+});
+
+async function get(headers = {}) {
+  await fetch(baseUrl, { headers });
+  return seenHeaders;
+}
+
+describe("custom-server peer header sanitizing", () => {
+  it("generates a peer trust token at boot", () => {
+    expect(process.env.NINEROUTER_PEER_TOKEN).toMatch(/^[0-9a-f]{48}$/);
+  });
+
+  it("replaces a client-supplied x-9r-real-ip with the socket address", async () => {
+    const headers = await get({ "x-9r-real-ip": "203.0.113.55" });
+
+    expect(headers["x-9r-real-ip"]).toMatch(/^(::ffff:)?127\.0\.0\.1$/);
+  });
+
+  it("stamps the trust token so downstream can tell the wrapper ran", async () => {
+    const headers = await get();
+
+    expect(headers["x-9r-peer-token"]).toBe(process.env.NINEROUTER_PEER_TOKEN);
+  });
+
+  it("drops a client-supplied peer trust token", async () => {
+    const headers = await get({ "x-9r-peer-token": "forged-token" });
+
+    expect(headers["x-9r-peer-token"]).toBe(process.env.NINEROUTER_PEER_TOKEN);
+    expect(headers["x-9r-peer-token"]).not.toBe("forged-token");
+  });
+
+  it("drops a client-supplied x-9r-via-proxy marker", async () => {
+    const headers = await get({ "x-9r-via-proxy": "1" });
+
+    expect(headers["x-9r-via-proxy"]).toBeUndefined();
+  });
+
+  it("marks via-proxy and adopts the forwarded IP for a loopback proxy hop", async () => {
+    const headers = await get({ "x-forwarded-for": "203.0.113.9, 10.0.0.1" });
+
+    expect(headers["x-9r-via-proxy"]).toBe("1");
+    expect(headers["x-9r-real-ip"]).toBe("203.0.113.9");
+    expect(headers["x-forwarded-for"]).toBeUndefined();
+  });
+
+  // chat.js snapshots every client header into the request detail. Anything that grants
+  // access must not survive into a record the dashboard renders and cloud sync uploads.
+  it("keeps the peer token out of persisted request details", () => {
+    const sanitized = requestDetails.sanitizeHeaders({
+      "x-9r-peer-token": "secret",
+      "x-9r-cli-token": "secret",
+      "authorization": "Bearer sk-x",
+      "x-9r-real-ip": "127.0.0.1",
+    });
+
+    expect(sanitized["x-9r-peer-token"]).toBeUndefined();
+    expect(sanitized["x-9r-cli-token"]).toBeUndefined();
+    expect(sanitized["authorization"]).toBeUndefined();
+    expect(sanitized["x-9r-real-ip"]).toBe("127.0.0.1");
+  });
+});

tests/unit/dashboard-guard.test.js

@@ -35,6 +35,8 @@ vi.mock("@/lib/auth/dashboardSession", () => ({
 
 const { proxy, __test__ } = await import("../../src/dashboardGuard.js");
 
+const PEER_TOKEN = "peer-token-fixture";
+
 function request(pathname, headers = {}) {
   const normalizedHeaders = new Headers(headers);
   return {
@@ -45,9 +47,16 @@ function request(pathname, headers = {}) {
   };
 }
 
+// A request that actually came through custom-server.js: peer IP stamped from the TCP
+// socket and proven by the per-process secret.
+function localRequest(pathname, headers = {}) {
+  return request(pathname, { "x-9r-peer-token": PEER_TOKEN, "x-9r-real-ip": "127.0.0.1", ...headers });
+}
+
 describe("dashboard guard public LLM API access", () => {
   beforeEach(() => {
     vi.clearAllMocks();
+    process.env.NINEROUTER_PEER_TOKEN = PEER_TOKEN;
     mocks.getSettings.mockResolvedValue({ requireLogin: true });
     mocks.validateApiKey.mockResolvedValue(false);
     mocks.getConsistentMachineId.mockResolvedValue("cli-token");
@@ -55,14 +64,14 @@ describe("dashboard guard public LLM API access", () => {
   });
 
   it("allows loopback public LLM API without API key", async () => {
-    const response = await proxy(request("/v1/chat/completions", { host: "localhost:20128" }));
+    const response = await proxy(localRequest("/v1/chat/completions", { host: "localhost:20128" }));
 
     expect(response).toBe(mocks.nextResponse);
     expect(mocks.validateApiKey).not.toHaveBeenCalled();
   });
 
   it("rejects remote Host-spoof when real peer IP is non-loopback", async () => {
-    const response = await proxy(request("/v1/chat/completions", {
+    const response = await proxy(localRequest("/v1/chat/completions", {
       host: "localhost",
       "x-9r-real-ip": "10.204.111.34",
     }));
@@ -72,7 +81,7 @@ describe("dashboard guard public LLM API access", () => {
   });
 
   it("allows loopback peer IP regardless of Host", async () => {
-    const response = await proxy(request("/v1/chat/completions", {
+    const response = await proxy(localRequest("/v1/chat/completions", {
       host: "localhost:20128",
       "x-9r-real-ip": "127.0.0.1",
     }));
@@ -89,7 +98,7 @@ describe("dashboard guard public LLM API access", () => {
   });
 
   it("allows loopback rewritten public LLM API without API key", async () => {
-    const response = await proxy(request("/api/v1/chat/completions", { host: "localhost:20128" }));
+    const response = await proxy(localRequest("/api/v1/chat/completions", { host: "localhost:20128" }));
 
     expect(response).toBe(mocks.nextResponse);
     expect(mocks.validateApiKey).not.toHaveBeenCalled();
@@ -191,6 +200,7 @@ describe("dashboard guard public LLM API access", () => {
 describe("dashboard guard local-only access", () => {
   beforeEach(() => {
     vi.clearAllMocks();
+    process.env.NINEROUTER_PEER_TOKEN = PEER_TOKEN;
     mocks.getSettings.mockResolvedValue({ requireLogin: true });
     mocks.validateApiKey.mockResolvedValue(false);
     mocks.getConsistentMachineId.mockResolvedValue("cli-token");
@@ -207,7 +217,7 @@ describe("dashboard guard local-only access", () => {
   });
 
   it("rejects local-only route on loopback when requireLogin=true and no JWT", async () => {
-    const response = await proxy(request("/api/mcp/filesystem/sse", {
+    const response = await proxy(localRequest("/api/mcp/filesystem/sse", {
       host: "localhost:20128",
       origin: "http://localhost:20128",
     }));
@@ -219,7 +229,7 @@ describe("dashboard guard local-only access", () => {
   it("allows local-only route on loopback when requireLogin=false", async () => {
     mocks.getSettings.mockResolvedValue({ requireLogin: false });
 
-    const response = await proxy(request("/api/cli-tools/antigravity-mitm", {
+    const response = await proxy(localRequest("/api/cli-tools/antigravity-mitm", {
       host: "localhost:20128",
       origin: "http://localhost:20128",
     }));
@@ -240,7 +250,7 @@ describe("dashboard guard local-only access", () => {
   it("rejects local-only route when Origin is non-loopback (CSRF block)", async () => {
     mocks.getSettings.mockResolvedValue({ requireLogin: false });
 
-    const response = await proxy(request("/api/cli-tools/antigravity-mitm", {
+    const response = await proxy(localRequest("/api/cli-tools/antigravity-mitm", {
       host: "localhost:20128",
       origin: "http://evil.example.com",
     }));

tests/unit/fish-audio-tts.test.js

@@ -0,0 +1,109 @@
+import { beforeEach, describe, expect, it, vi } from "vitest";
+
+import REGISTRY from "../../open-sse/providers/registry/index.js";
+import { PROVIDER_MEDIA } from "../../open-sse/providers/index.js";
+import { FORMAT_HANDLERS } from "../../open-sse/handlers/ttsProviders/genericFormats.js";
+import { AI_PROVIDERS } from "@/shared/constants/providers";
+
+const AUDIO = new Uint8Array(256).fill(7);
+
+function okResponse() {
+  return {
+    ok: true,
+    status: 200,
+    headers: new Headers({ "content-type": "audio/mpeg" }),
+    arrayBuffer: async () => AUDIO.buffer,
+  };
+}
+
+describe("Fish Audio TTS provider", () => {
+  const entry = REGISTRY.find((e) => e.id === "fish-audio");
+
+  it("is registered as a TTS-only apikey provider", () => {
+    expect(entry).toBeDefined();
+    expect(entry.category).toBe("apikey");
+    expect(entry.serviceKinds).toEqual(["tts"]);
+    expect(PROVIDER_MEDIA["fish-audio"]?.ttsConfig?.baseUrl).toBe("https://api.fish.audio/v1/tts");
+  });
+
+  it("is visible to the generic dispatcher, which reads AI_PROVIDERS", () => {
+    // synthesizeViaConfig() looks the provider up here, not in PROVIDER_MEDIA.
+    expect(AI_PROVIDERS["fish-audio"]?.ttsConfig?.format).toBe("fish-audio");
+    expect(typeof FORMAT_HANDLERS["fish-audio"]).toBe("function");
+  });
+
+  it("exposes the four documented models", () => {
+    const ids = (entry.ttsConfig.models || []).map((m) => m.id);
+    expect(ids).toEqual(["s2.1-pro-free", "s2.1-pro", "s2-pro", "s1"]);
+  });
+
+  it("keeps registry ids and aliases unique", () => {
+    const ids = REGISTRY.map((e) => e.id);
+    expect(new Set(ids).size).toBe(ids.length);
+    const aliases = REGISTRY.map((e) => e.alias).filter(Boolean);
+    expect(new Set(aliases).size).toBe(aliases.length);
+  });
+});
+
+describe("Fish Audio TTS request shape", () => {
+  const handler = FORMAT_HANDLERS["fish-audio"];
+  let fetchMock;
+
+  beforeEach(() => {
+    fetchMock = vi.fn(async () => okResponse());
+    global.fetch = fetchMock;
+  });
+
+  const callArgs = () => {
+    const [url, init] = fetchMock.mock.calls.at(-1);
+    return { url, init, body: JSON.parse(init.body) };
+  };
+
+  it("sends the model as an HTTP header, not in the body", async () => {
+    await handler({
+      baseUrl: "https://api.fish.audio/v1/tts",
+      apiKey: "sk-test",
+      text: "xin chào",
+      modelId: "s1",
+      voiceId: "",
+    });
+
+    const { url, init, body } = callArgs();
+    expect(url).toBe("https://api.fish.audio/v1/tts");
+    expect(init.headers.model).toBe("s1");
+    expect(init.headers.Authorization).toBe("Bearer sk-test");
+    expect(body).toEqual({ text: "xin chào", format: "mp3" });
+  });
+
+  it("maps the voice onto reference_id, and omits it when unset", async () => {
+    await handler({ baseUrl: "u", apiKey: "k", text: "t", modelId: "s1", voiceId: "voice-abc" });
+    expect(callArgs().body.reference_id).toBe("voice-abc");
+
+    await handler({ baseUrl: "u", apiKey: "k", text: "t", modelId: "s1", voiceId: "" });
+    expect(callArgs().body).not.toHaveProperty("reference_id");
+  });
+
+  it("defaults to the free model when none is given", async () => {
+    await handler({ baseUrl: "u", apiKey: "k", text: "t", modelId: "", voiceId: "" });
+    expect(callArgs().init.headers.model).toBe("s2.1-pro-free");
+  });
+
+  it("returns base64 audio with its format", async () => {
+    const out = await handler({ baseUrl: "u", apiKey: "k", text: "t", modelId: "s1", voiceId: "" });
+    expect(out.format).toBe("mp3");
+    expect(typeof out.base64).toBe("string");
+    expect(out.base64.length).toBeGreaterThan(0);
+  });
+
+  it("surfaces the upstream error message", async () => {
+    global.fetch = vi.fn(async () => ({
+      ok: false,
+      status: 402,
+      text: async () => JSON.stringify({ message: "Insufficient credit" }),
+    }));
+
+    await expect(
+      handler({ baseUrl: "u", apiKey: "k", text: "t", modelId: "s1", voiceId: "" }),
+    ).rejects.toThrow("Insufficient credit");
+  });
+});

tests/unit/fusion-strip-stream-options-3024.test.js

@@ -0,0 +1,83 @@
+// Issue #3024 — Fusion combo must strip `stream_options` from panel requests
+// when running non-streaming, or DeepSeek rejects with
+// "stream_options should be set along with stream = true".
+
+import { describe, it, expect, vi } from "vitest";
+import { handleFusionChat } from "../../open-sse/services/combo.js";
+
+// Minimal logger stub (combo.js calls log.info/warn).
+const log = { info: () => {}, warn: () => {}, error: () => {} };
+
+function makeBody(extra = {}) {
+  return {
+    model: "combo/gemseek",
+    stream: true,
+    stream_options: { include_usage: true },
+    messages: [{ role: "user", content: "hi" }],
+    ...extra,
+  };
+}
+
+describe("Fusion strips stream_options (#3024)", () => {
+  it("removes stream_options before fanning out to panel models", async () => {
+    let capturedPanelBody = null;
+    const handleSingleModel = vi.fn(async (panelBody, model, isPanel) => {
+      if (isPanel) capturedPanelBody = panelBody;
+      // Simulate a successful non-stream JSON answer for the panel.
+      if (isPanel) {
+        return new Response(JSON.stringify({ choices: [{ message: { content: `ans-${model}` } }] }), {
+          status: 200,
+          headers: { "Content-Type": "application/json" },
+        });
+      }
+      // Judge leg: return a final answer.
+      return new Response(JSON.stringify({ choices: [{ message: { content: "final" } }] }), {
+        status: 200,
+        headers: { "Content-Type": "application/json" },
+      });
+    });
+
+    const res = await handleFusionChat({
+      body: makeBody(),
+      models: ["ds/deepseek-v4-flash", "gemini/gemini-3.5-flash-lite"],
+      handleSingleModel,
+      log,
+      comboName: "GemSeek",
+      judgeModel: "gemini/gemini-3.5-flash-lite",
+    });
+
+    expect(res).toBeInstanceOf(Response);
+    expect(capturedPanelBody).not.toBeNull();
+    // Critical assertion: stream_options must NOT leak into panel requests.
+    expect(capturedPanelBody.stream_options).toBeUndefined();
+    expect(capturedPanelBody.stream).toBe(false);
+    // Ensure the original client body still had it (proves we stripped deliberately).
+    expect(makeBody().stream_options).toBeDefined();
+  });
+
+  it("does not throw for a 2-model fusion with stream_options present", async () => {
+    const handleSingleModel = vi.fn(async (panelBody, model, isPanel) => {
+      if (isPanel) {
+        return new Response(JSON.stringify({ choices: [{ message: { content: "ok" } }] }), {
+          status: 200,
+          headers: { "Content-Type": "application/json" },
+        });
+      }
+      return new Response(JSON.stringify({ choices: [{ message: { content: "final" } }] }), {
+        status: 200,
+        headers: { "Content-Type": "application/json" },
+      });
+    });
+
+    const res = await handleFusionChat({
+      body: makeBody({ stream_options: { include_usage: true } }),
+      models: ["ds/deepseek-v4-pro", "ds/deepseek-v4-flash"],
+      handleSingleModel,
+      log,
+      comboName: "GemSeek",
+      judgeModel: "ds/deepseek-v4-pro",
+    });
+
+    expect(res.status).toBe(200);
+  });
+});

tests/unit/gemini-3.7-antigravity.test.js

@@ -0,0 +1,36 @@
+import { describe, it, expect } from "vitest";
+import { getCapabilitiesForModel } from "../../open-sse/providers/capabilities.js";
+import antigravityRegistry from "../../open-sse/providers/registry/antigravity.js";
+import geminiRegistry from "../../open-sse/providers/registry/gemini.js";
+import { MODEL_PRICING } from "../../open-sse/providers/pricing.js";
+
+describe("Gemini 3.7 Flash Support & Config (#3286, #3281)", () => {
+  it("registers gemini-3.7-flash tiered models in antigravity provider registry", () => {
+    const agIds = antigravityRegistry.models.map(m => m.id);
+    expect(agIds).toContain("gemini-3.7-flash-high");
+    expect(agIds).toContain("gemini-3.7-flash-medium");
+    expect(agIds).toContain("gemini-3.7-flash-low");
+    expect(agIds).not.toContain("gemini-3.7-flash");
+  });
+
+  it("registers gemini-3.7-flash in gemini provider registry", () => {
+    const geminiIds = geminiRegistry.models.map(m => m.id);
+    expect(geminiIds).toContain("gemini-3.7-flash");
+  });
+
+  it("resolves capabilities correctly for gemini-3.7 models with official limits", () => {
+    const caps = getCapabilitiesForModel("antigravity", "gemini-3.7-flash-high");
+    expect(caps.vision).toBe(true);
+    expect(caps.reasoning).toBe(true);
+    expect(caps.thinkingFormat).toBe("gemini-level");
+    expect(caps.contextWindow).toBe(1048576);
+    expect(caps.maxOutput).toBe(65536);
+  });
+
+  it("defines pricing matching gemini-3.6-flash baseline", () => {
+    expect(MODEL_PRICING["gemini-3.7-flash"]).toEqual(MODEL_PRICING["gemini-3.6-flash"]);
+    expect(MODEL_PRICING["gemini-3.7-flash-high"]).toEqual(MODEL_PRICING["gemini-3.6-flash-high"]);
+    expect(MODEL_PRICING["gemini-3.7-flash-medium"]).toEqual(MODEL_PRICING["gemini-3.6-flash-medium"]);
+    expect(MODEL_PRICING["gemini-3.7-flash-low"]).toEqual(MODEL_PRICING["gemini-3.6-flash-low"]);
+  });
+});

tests/unit/hermes-vision-detection.test.js

@@ -0,0 +1,115 @@
+import { describe, it, expect } from "vitest";
+import { detectRequiredCapabilities } from "../../open-sse/services/combo.js";
+import { augmentModelsWithCapacityAdapter } from "../../open-sse/services/capacityAdapter.js";
+import { stripUnsupportedModalities } from "../../open-sse/translator/concerns/modality.js";
+import { FORMATS } from "../../open-sse/translator/formats.js";
+
+describe("Hermes Vision Image Detection", () => {
+  it("detects vision from Ollama / Hermes images array", () => {
+    const body = {
+      messages: [
+        {
+          role: "user",
+          content: "Please analyze this image from Hermes",
+          images: ["iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg=="],
+        },
+      ],
+    };
+    const caps = detectRequiredCapabilities(body);
+    expect(caps.has("vision")).toBe(true);
+  });
+
+  it("detects vision from Vercel AI SDK / Hermes experimental_attachments", () => {
+    const body = {
+      messages: [
+        {
+          role: "user",
+          content: "Describe this attachment",
+          experimental_attachments: [
+            {
+              contentType: "image/png",
+              url: "data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg==",
+            },
+          ],
+        },
+      ],
+    };
+    const caps = detectRequiredCapabilities(body);
+    expect(caps.has("vision")).toBe(true);
+  });
+
+  it("detects vision from Hermes attachments array", () => {
+    const body = {
+      messages: [
+        {
+          role: "user",
+          content: "Look at this photo",
+          attachments: [
+            {
+              mediaType: "image/jpeg",
+              url: "https://example.com/photo.jpg",
+            },
+          ],
+        },
+      ],
+    };
+    const caps = detectRequiredCapabilities(body);
+    expect(caps.has("vision")).toBe(true);
+  });
+
+  it("detects vision from embedded data:image URI in string content", () => {
+    const body = {
+      messages: [
+        {
+          role: "user",
+          content: "Here is an inline image: data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAYAAAAfFcSJAAAADUlEQVR42mNk+M9QDwADhgGAWjR9awAAAABJRU5ErkJggg==",
+        },
+      ],
+    };
+    const caps = detectRequiredCapabilities(body);
+    expect(caps.has("vision")).toBe(true);
+  });
+
+  it("auto-switches non-vision model (deepseek-v4-pro) to Vision Adapter model (Kimi-K3)", () => {
+    const body = {
+      messages: [
+        {
+          role: "user",
+          content: "Analyze image",
+          images: ["base64data..."],
+        },
+      ],
+    };
+    const reqCaps = detectRequiredCapabilities(body);
+    const settings = {
+      capacityAdapter: {
+        vision: {
+          enabled: true,
+          models: ["cmc/moonshotai/Kimi-K3"],
+        },
+      },
+    };
+
+    const augmented = augmentModelsWithCapacityAdapter(["cmc/deepseek/deepseek-v4-pro"], reqCaps, settings);
+    expect(augmented).toEqual(["cmc/moonshotai/Kimi-K3", "cmc/deepseek/deepseek-v4-pro"]);
+  });
+
+  it("strips msg.images and attachments when model does not support vision", () => {
+    const body = {
+      messages: [
+        {
+          role: "user",
+          content: "Test text",
+          images: ["base64..."],
+          experimental_attachments: [{ contentType: "image/png", url: "data:image/png;base64,..." }],
+        },
+      ],
+    };
+    const noVisionCaps = { vision: false, pdf: false, audioInput: false };
+
+    stripUnsupportedModalities(body, FORMATS.OPENAI, noVisionCaps);
+
+    expect(body.messages[0].images).toBeUndefined();
+    expect(body.messages[0].experimental_attachments).toHaveLength(0);
+  });
+});

tests/unit/kiro-model-slots.test.js

@@ -14,6 +14,13 @@ describe("Kiro MITM model slots", () => {
     expect(Array.isArray(kiro.defaultModels)).toBe(true);
   });
 
+  it("offers a mappable slot for the agent default model id 'auto'", () => {
+    // اسلات auto برای vibe mode لازمه — وگرنه درخواست میره AWS
+    const auto = kiro.defaultModels.find((m) => m.id === "auto");
+    expect(auto).toBeTruthy();
+    expect(auto.alias).toBe("auto");
+  });
+
   it("offers a mappable slot for Claude Sonnet 5", () => {
     const sonnet5 = kiro.defaultModels.find((m) => m.id === "claude-sonnet-5");
     expect(sonnet5).toBeTruthy();

tests/unit/kiro-usage-and-tool-integrity.test.js

@@ -0,0 +1,369 @@
+/**
+ * Five Kiro defects, all on the OAuth/social Kiro route (kr/claude-sonnet-4.5
+ * with a >100k context). Production shape: 402 of 2156 usageHistory rows for
+ * Kiro recorded completionTokens 0, and the 25 newest rows all sat pinned at
+ * exactly 1 output token against prompts of 80k-103k -- i.e. the router was not
+ * measuring the answer, it was measuring nothing and rounding up.
+ *
+ * A. OUT 0 / OUT 1. finish() estimates completion tokens as
+ *    totalContentLength / 4, but tool-call bytes were never added to
+ *    totalContentLength. A turn whose entire answer is a tool call therefore
+ *    measured as an empty answer (Math.max(1, ...) is where the 1 comes from).
+ *
+ * B. Truncation threw away a complete-enough answer. stopDisposition() maps
+ *    model_context_window_exceeded (and max_tokens alongside tool calls) to
+ *    terminal_incomplete, which hard-fails the turn -- even when the model had
+ *    already streamed text. A truncated turn is what finish_reason "length" is
+ *    for. Both disposition gates needed the bypass: the declared-stop-reason
+ *    gate runs first and returns, so patching only finish() would be dead code.
+ *
+ * C. One bad tool fragment killed every good one. Three separate latches:
+ *    emitTools() validated per turn and threw out of the loop; the frame-loop
+ *    catch cleared state.tools wholesale; and the toolUseEvent branch returned
+ *    early forever once toolValidationError was set. Net effect for the client:
+ *    a turn that answered nothing.
+ *
+ * D. Cache tokens dropped on the kiro:claude route. kiro-to-claude built usage
+ *    from prompt_tokens/completion_tokens only, so Claude clients lost
+ *    cache_read_input_tokens / cache_creation_input_tokens and could neither
+ *    price the turn nor size their prompt cache.
+ *
+ * E. Defence-in-depth only: both request translators discarded canonical.valid.
+ *    canonicalizeKiroConversation() self-heals every failure mode it can detect
+ *    (see the test below), so the guard is unreachable by construction today --
+ *    it exists so a future validator rule cannot ship an unusable body silently.
+ *
+ * Kiro also answers an unusable conversation with 400 {"message":"Improperly
+ * formed request.","reason":"REQUEST_BODY_INVALID"}, which cools down every
+ * account that reports it. That is a property of ERROR_RULES rather than of this
+ * executor -- the same body fails identically on any account -- so it belongs to
+ * the request-scoped `fallback: false` rule kind, not here.
+ */
+import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
+
+const fetchMock = vi.fn();
+
+vi.mock("../../open-sse/utils/proxyFetch.js", () => ({
+  proxyAwareFetch: (...args) => fetchMock(...args)
+}));
+
+const { KiroExecutor } = await import("../../open-sse/executors/kiro.js");
+const { kiroToClaudeResponse, kiroToClaudeNonStreaming } = await import(
+  "../../open-sse/translator/response/kiro-to-claude.js"
+);
+const { validateKiroConversation, canonicalizeKiroConversation } = await import(
+  "../../open-sse/translator/concerns/kiroConversation.js"
+);
+
+const encoder = new TextEncoder();
+const credentials = {
+  accessToken: "test-token",
+  providerSpecificData: { kiroToolCallRepair: true }
+};
+
+function crc32(bytes) {
+  let crc = 0xffffffff;
+  for (const byte of bytes) {
+    crc ^= byte;
+    for (let bit = 0; bit < 8; bit++) {
+      crc = (crc >>> 1) ^ ((crc & 1) ? 0xedb88320 : 0);
+    }
+  }
+  return (crc ^ 0xffffffff) >>> 0;
+}
+
+function encodeHeader(name, value) {
+  const nameBytes = encoder.encode(name);
+  const valueBytes = encoder.encode(value);
+  const bytes = new Uint8Array(1 + nameBytes.length + 3 + valueBytes.length);
+  let offset = 0;
+  bytes[offset++] = nameBytes.length;
+  bytes.set(nameBytes, offset);
+  offset += nameBytes.length;
+  bytes[offset++] = 7;
+  new DataView(bytes.buffer).setUint16(offset, valueBytes.length, false);
+  offset += 2;
+  bytes.set(valueBytes, offset);
+  return bytes;
+}
+
+function concat(chunks) {
+  const output = new Uint8Array(chunks.reduce((size, chunk) => size + chunk.byteLength, 0));
+  let offset = 0;
+  for (const chunk of chunks) {
+    output.set(chunk, offset);
+    offset += chunk.byteLength;
+  }
+  return output;
+}
+
+function checksum(bytes) {
+  const view = new DataView(bytes.buffer, bytes.byteOffset, bytes.byteLength);
+  view.setUint32(8, crc32(bytes.subarray(0, 8)), false);
+  view.setUint32(bytes.byteLength - 4, crc32(bytes.subarray(0, bytes.byteLength - 4)), false);
+  return bytes;
+}
+
+function frameFromEntries(entries, payload) {
+  const headers = concat(entries.map(([name, value]) => encodeHeader(name, value)));
+  const payloadBytes = encoder.encode(JSON.stringify(payload));
+  const totalLength = 12 + headers.byteLength + payloadBytes.byteLength + 4;
+  const frame = new Uint8Array(totalLength);
+  const view = new DataView(frame.buffer);
+  view.setUint32(0, totalLength, false);
+  view.setUint32(4, headers.byteLength, false);
+  frame.set(headers, 12);
+  frame.set(payloadBytes, 12 + headers.byteLength);
+  return checksum(frame);
+}
+
+function frame(eventType, payload) {
+  return frameFromEntries([[":event-type", eventType]], payload);
+}
+
+function response(frames, status = 200) {
+  return new Response(new ReadableStream({
+    start(controller) {
+      for (const value of frames) controller.enqueue(value);
+      controller.close();
+    }
+  }), { status, statusText: status === 200 ? "OK" : "Upstream Error" });
+}
+
+async function execute(executor = new KiroExecutor(), overrides = {}) {
+  return executor.execute({
+    model: "kr/claude-opus-4.8",
+    body: { systemPrompt: "base", conversationState: {} },
+    stream: true,
+    credentials,
+    ...overrides
+  });
+}
+
+// Output is held behind the ": kiro-validation" heartbeat until clean EOF, so
+// every executor assertion has to drain the whole response.
+async function run(frames) {
+  fetchMock.mockResolvedValueOnce(response(frames));
+  return await (await execute()).response.text();
+}
+
+// Same as run(), but with the bounded tool-call repair retry disabled, so a
+// hard failure is surfaced from the first attempt instead of triggering a
+// second upstream fetch.
+async function runNoRepair(frames) {
+  fetchMock.mockResolvedValueOnce(response(frames));
+  const result = await execute(new KiroExecutor(), {
+    credentials: { accessToken: "test-token", providerSpecificData: { kiroToolCallRepair: false } }
+  });
+  return await result.response.text();
+}
+
+// The estimator only runs when metering and context usage both arrived and the
+// upstream reported no token totals of its own -- the exact production shape.
+const METERED = [
+  frame("meteringEvent", { usage: 2, unit: "credit" }),
+  frame("contextUsageEvent", { contextUsagePercentage: 10 })
+];
+
+function usageFrom(body) {
+  const usages = body
+    .split("\n")
+    .filter(line => line.startsWith("data: ") && line.includes('"usage"'))
+    .map(line => JSON.parse(line.slice(6)).usage)
+    .filter(Boolean);
+  return usages[usages.length - 1];
+}
+
+beforeEach(() => {
+  fetchMock.mockReset();
+  vi.spyOn(console, "error").mockImplementation(() => {});
+});
+afterEach(() => {
+  vi.restoreAllMocks();
+});
+
+describe("A: tool-call bytes count as output tokens", () => {
+  it("does not report a tool-only turn as OUT 1", async () => {
+    const body = await run([
+      frame("toolUseEvent", {
+        toolUseId: "call_a",
+        name: "tool_call",
+        input: { name: "mcp_search", arguments: { query: "why is the router reporting zero output" } }
+      }),
+      frame("metadataEvent", { stopReason: "tool_use" }),
+      ...METERED
+    ]);
+
+    const usage = usageFrom(body);
+    expect(usage).toBeDefined();
+    // Was 1: the Math.max floor over a totalContentLength of 0.
+    expect(usage.completion_tokens).toBeGreaterThan(10);
+    expect(usage.total_tokens).toBe(usage.prompt_tokens + usage.completion_tokens);
+  });
+
+  it("still counts plain text output", async () => {
+    const body = await run([
+      frame("assistantResponseEvent", { content: "x".repeat(400) }),
+      frame("metadataEvent", { stopReason: "end_turn" }),
+      ...METERED
+    ]);
+    expect(usageFrom(body).completion_tokens).toBe(100);
+  });
+});
+
+describe("C: one unusable tool fragment does not take the whole turn with it", () => {
+  it("ships the valid call and drops only the invalid one", async () => {
+    const body = await run([
+      frame("toolUseEvent", {
+        toolUseId: "good",
+        name: "tool_call",
+        input: { name: "mcp_search", arguments: { q: "router" } }
+      }),
+      // No nested MCP name -> unusable, cannot be forwarded to the client.
+      frame("toolUseEvent", { toolUseId: "bad", name: "tool_call", input: { arguments: { q: "router" } } }),
+      frame("metadataEvent", { stopReason: "tool_use" }),
+      ...METERED
+    ]);
+
+    expect(body).toContain('\\"name\\":\\"mcp_search\\"');
+    expect(body).not.toContain('"id":"bad"');
+    expect(body).toContain('"finish_reason":"tool_calls"');
+  });
+
+  it("keeps streamed text when the only tool call is unusable", async () => {
+    const body = await run([
+      frame("assistantResponseEvent", { content: "Here is what I found." }),
+      frame("toolUseEvent", { toolUseId: "bad", name: "tool_call", input: { arguments: {} } }),
+      frame("metadataEvent", { stopReason: "tool_use" }),
+      ...METERED
+    ]);
+
+    expect(body).toContain("Here is what I found.");
+    expect(body).not.toContain("invalid_kiro_tool_call");
+  });
+
+  it("still hard-fails when nothing usable was produced at all", async () => {
+    const body = await runNoRepair([
+      frame("toolUseEvent", { toolUseId: "bad", name: "tool_call", input: { arguments: {} } }),
+      frame("metadataEvent", { stopReason: "tool_use" })
+    ]);
+    expect(body).toContain("invalid_kiro_tool_call");
+    expect(fetchMock).toHaveBeenCalledTimes(1);
+  });
+});
+
+describe("B: truncation after output closes as length, not as a failure", () => {
+  it("keeps the text and reports finish_reason length", async () => {
+    const body = await run([
+      frame("assistantResponseEvent", { content: "Partial but usable answer." }),
+      frame("metadataEvent", { stopReason: "model_context_window_exceeded" }),
+      ...METERED
+    ]);
+
+    expect(body).toContain("Partial but usable answer.");
+    expect(body).toContain('"finish_reason":"length"');
+    expect(body).not.toContain("kiro_terminal_incomplete");
+  });
+
+  it("still fails a truncation that produced nothing", async () => {
+    const body = await run([
+      frame("metadataEvent", { stopReason: "model_context_window_exceeded" })
+    ]);
+    expect(body).toContain("kiro_terminal_incomplete");
+  });
+});
+
+describe("D: cache tokens survive the kiro -> claude translation", () => {
+  const finishChunk = (usage) => ({
+    choices: [{ index: 0, delta: {}, finish_reason: "stop" }],
+    usage
+  });
+
+  function finalUsage(usage) {
+    const state = {};
+    // Usage rides an earlier chunk in the real stream; feed it the same way.
+    kiroToClaudeResponse({ choices: [{ index: 0, delta: { content: "hi" } }], usage }, state);
+    const events = kiroToClaudeResponse(finishChunk(usage), state) || [];
+    return events.find(e => e.type === "message_delta")?.usage;
+  }
+
+  it("forwards the flat Chat spelling the executor emits", () => {
+    expect(finalUsage({
+      prompt_tokens: 103000,
+      completion_tokens: 640,
+      cache_read_input_tokens: 98000,
+      cache_creation_input_tokens: 1912
+    })).toEqual({
+      input_tokens: 103000,
+      output_tokens: 640,
+      cache_read_input_tokens: 98000,
+      cache_creation_input_tokens: 1912
+    });
+  });
+
+  it("also accepts the nested details spelling used on passthrough", () => {
+    expect(finalUsage({
+      prompt_tokens: 500,
+      completion_tokens: 20,
+      prompt_tokens_details: { cached_tokens: 480, cache_creation_tokens: 20 }
+    })).toEqual({
+      input_tokens: 500,
+      output_tokens: 20,
+      cache_read_input_tokens: 480,
+      cache_creation_input_tokens: 20
+    });
+  });
+
+  it("omits the cache keys when the upstream reported none", () => {
+    expect(finalUsage({ prompt_tokens: 500, completion_tokens: 20 }))
+      .toEqual({ input_tokens: 500, output_tokens: 20 });
+  });
+
+  it("preserves cache on the non-streaming path too", () => {
+    const message = kiroToClaudeNonStreaming({
+      choices: [{ message: { content: "hi" } }],
+      usage: { prompt_tokens: 90, completion_tokens: 4, cache_read_input_tokens: 80 }
+    });
+    expect(message.usage).toMatchObject({
+      input_tokens: 90,
+      output_tokens: 4,
+      cache_read_input_tokens: 80
+    });
+    expect(message.usage).not.toHaveProperty("cache_creation_input_tokens");
+  });
+});
+
+describe("E: the translator valid-guard is defence-in-depth", () => {
+  const SPECS = [{ toolSpecification: { name: "read_file", inputSchema: { json: { type: "object" } } } }];
+
+  it("validateKiroConversation names the offending turn", () => {
+    // Hand-built, NOT normalized: assistant first, then a tool call with no
+    // matching result and a name no spec declares.
+    const result = validateKiroConversation(
+      [{ assistantResponseMessage: { content: "hi", toolUses: [{ toolUseId: "t1", name: "ghost" }] } }],
+      { userInputMessage: { content: "go" } },
+      SPECS
+    );
+    expect(result.valid).toBe(false);
+    expect(result.errors).toContain("role:0");
+    expect(result.errors).toContain("pair:0");
+    expect(result.errors).toContain("spec:0");
+  });
+
+  it("canonicalizeKiroConversation heals that same conversation", () => {
+    // This is why the guard cannot fire today: normalizeTurns() forces
+    // user-first/user-last alternation and non-empty content, and the
+    // second-chance pass flattens every structured tool turn to text.
+    const out = canonicalizeKiroConversation({
+      history: [{ assistantResponseMessage: { content: "hi", toolUses: [{ toolUseId: "t1", name: "ghost" }] } }],
+      currentMessage: { userInputMessage: { content: "" } },
+      modelId: "claude-sonnet-4-5",
+      toolSpecs: SPECS
+    });
+
+    expect(out.valid).toBe(true);
+    expect(out.errors).toEqual([]);
+    expect(out.currentMessage.userInputMessage.content).toBe("continue");
+    expect(out.history[0].userInputMessage).toBeDefined();
+  });
+});

tests/unit/local-request-peer-trust-3294.test.js

@@ -0,0 +1,211 @@
+// GHSA-pjm4-8fpg-f9p6 (#3294): `next start` leaves custom-server.js out of the request
+// path, so x-9r-real-ip arrives straight from the client and a remote caller can claim to
+// be loopback. Host is spoofable the same way, so it cannot be the production fallback.
+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
+
+const mocks = vi.hoisted(() => ({
+  nextResponse: Symbol("next"),
+  jsonResponse: vi.fn((body, init) => ({ status: init?.status || 200, body })),
+  getSettings: vi.fn(),
+  validateApiKey: vi.fn(),
+  getConsistentMachineId: vi.fn(),
+  verifyDashboardAuthToken: vi.fn(),
+}));
+
+vi.mock("next/server", () => ({
+  NextResponse: {
+    next: vi.fn(() => mocks.nextResponse),
+    json: mocks.jsonResponse,
+    redirect: vi.fn((url) => ({ status: 307, url })),
+  },
+}));
+
+vi.mock("@/lib/localDb", () => ({
+  getSettings: mocks.getSettings,
+  validateApiKey: mocks.validateApiKey,
+}));
+
+vi.mock("@/shared/utils/machineId", () => ({
+  getConsistentMachineId: mocks.getConsistentMachineId,
+}));
+
+vi.mock("@/lib/auth/dashboardSession", () => ({
+  verifyDashboardAuthToken: mocks.verifyDashboardAuthToken,
+}));
+
+const { proxy } = await import("../../src/dashboardGuard.js");
+const { getClientIp } = await import("../../src/lib/auth/loginLimiter.js");
+
+const PEER_TOKEN = "peer-token-fixture";
+
+function request(pathname, headers = {}) {
+  return {
+    nextUrl: { pathname, searchParams: new URL(`http://localhost${pathname}`).searchParams },
+    headers: new Headers(headers),
+    cookies: { get: vi.fn(() => undefined) },
+    url: `http://localhost${pathname}`,
+  };
+}
+
+const originalNodeEnv = process.env.NODE_ENV;
+
+describe("peer header trust", () => {
+  beforeEach(() => {
+    vi.clearAllMocks();
+    process.env.NINEROUTER_PEER_TOKEN = PEER_TOKEN;
+    process.env.NODE_ENV = "production";
+    mocks.getSettings.mockResolvedValue({ requireLogin: true });
+    mocks.validateApiKey.mockResolvedValue(false);
+    mocks.getConsistentMachineId.mockResolvedValue("cli-token");
+    mocks.verifyDashboardAuthToken.mockResolvedValue(false);
+  });
+
+  afterEach(() => {
+    process.env.NODE_ENV = originalNodeEnv;
+    delete process.env.NINEROUTER_PEER_TOKEN;
+  });
+
+  it("rejects a spoofed loopback peer IP that carries no trust proof", async () => {
+    const response = await proxy(request("/api/v1/models", {
+      host: "172.18.192.1:20140",
+      "x-9r-real-ip": "127.0.0.1",
+    }));
+
+    expect(response.status).toBe(401);
+    expect(response.body.error).toBe("API key required for remote API access");
+  });
+
+  it("rejects a spoofed loopback peer IP carrying a wrong trust token", async () => {
+    const response = await proxy(request("/api/v1/models", {
+      host: "172.18.192.1:20140",
+      "x-9r-real-ip": "127.0.0.1",
+      "x-9r-peer-token": "guessed-token",
+    }));
+
+    expect(response.status).toBe(401);
+  });
+
+  it("rejects a spoofed loopback Host in production", async () => {
+    const response = await proxy(request("/api/v1/models", { host: "localhost" }));
+
+    expect(response.status).toBe(401);
+  });
+
+  it("rejects a spoofed loopback peer IP when the wrapper never booted", async () => {
+    delete process.env.NINEROUTER_PEER_TOKEN;
+
+    const response = await proxy(request("/api/v1/models", {
+      host: "172.18.192.1:20140",
+      "x-9r-real-ip": "127.0.0.1",
+      "x-9r-peer-token": "any-token",
+    }));
+
+    expect(response.status).toBe(401);
+  });
+
+  it("keeps serving a genuinely local request stamped by the wrapper", async () => {
+    const response = await proxy(request("/api/v1/models", {
+      host: "localhost:20128",
+      "x-9r-real-ip": "127.0.0.1",
+      "x-9r-peer-token": PEER_TOKEN,
+    }));
+
+    expect(response).toBe(mocks.nextResponse);
+    expect(mocks.validateApiKey).not.toHaveBeenCalled();
+  });
+
+  // A dual-stack listener reports loopback as ::ffff:127.0.0.1, which the old
+  // split-on-first-colon check reduced to "".
+  it.each(["::ffff:127.0.0.1", "::1", "[::1]", "127.0.0.1", "::FFFF:127.0.0.1"])(
+    "treats %s as a loopback peer",
+    async (peerIp) => {
+      const response = await proxy(request("/api/v1/models", {
+        host: "localhost:20128",
+        "x-9r-real-ip": peerIp,
+        "x-9r-peer-token": PEER_TOKEN,
+      }));
+
+      expect(response).toBe(mocks.nextResponse);
+    }
+  );
+
+  it.each(["::ffff:10.204.111.34", "2001:db8::1", "[2001:db8::1]", "10.204.111.34"])(
+    "refuses %s as a peer",
+    async (peerIp) => {
+      const response = await proxy(request("/api/v1/models", {
+        host: "localhost:20128",
+        "x-9r-real-ip": peerIp,
+        "x-9r-peer-token": PEER_TOKEN,
+      }));
+
+      expect(response.status).toBe(401);
+    }
+  );
+
+  it("still refuses a stamped non-loopback peer IP", async () => {
+    const response = await proxy(request("/api/v1/models", {
+      host: "localhost:20128",
+      "x-9r-real-ip": "10.204.111.34",
+      "x-9r-peer-token": PEER_TOKEN,
+    }));
+
+    expect(response.status).toBe(401);
+  });
+
+  it("blocks spoofed local-only routes that would otherwise spawn processes", async () => {
+    mocks.getSettings.mockResolvedValue({ requireLogin: false });
+
+    const response = await proxy(request("/api/mcp/filesystem/sse", {
+      host: "172.18.192.1:20140",
+      "x-9r-real-ip": "127.0.0.1",
+    }));
+
+    expect(response.status).toBe(403);
+    expect(response.body.error).toBe("Local only: CLI token required");
+  });
+
+  it("accepts the legacy Host fallback only in development", async () => {
+    process.env.NODE_ENV = "development";
+
+    const response = await proxy(request("/api/v1/models", { host: "localhost:20127" }));
+
+    expect(response).toBe(mocks.nextResponse);
+  });
+});
+
+describe("login limiter client IP", () => {
+  beforeEach(() => {
+    process.env.NINEROUTER_PEER_TOKEN = PEER_TOKEN;
+    delete process.env.TRUST_PROXY;
+  });
+
+  afterEach(() => {
+    delete process.env.NINEROUTER_PEER_TOKEN;
+    delete process.env.TRUST_PROXY;
+  });
+
+  it("buckets spoofed peer IPs together so lockout cannot be rotated away", () => {
+    const first = getClientIp(request("/api/auth/login", { "x-9r-real-ip": "1.1.1.1" }));
+    const second = getClientIp(request("/api/auth/login", { "x-9r-real-ip": "2.2.2.2" }));
+
+    expect(first).toBe("unknown");
+    expect(second).toBe("unknown");
+  });
+
+  it("keys on the stamped peer IP when the wrapper proved it", () => {
+    const ip = getClientIp(request("/api/auth/login", {
+      "x-9r-real-ip": "203.0.113.9",
+      "x-9r-peer-token": PEER_TOKEN,
+    }));
+
+    expect(ip).toBe("203.0.113.9");
+  });
+
+  it("still honours TRUST_PROXY for operators fronting 9router with a reverse proxy", () => {
+    process.env.TRUST_PROXY = "true";
+
+    const ip = getClientIp(request("/api/auth/login", { "x-forwarded-for": "198.51.100.7, 10.0.0.1" }));
+
+    expect(ip).toBe("198.51.100.7");
+  });
+});

tests/unit/openai-responses-empty-toolcalls.test.js

@@ -0,0 +1,52 @@
+/**
+ * Some providers (e.g. codebuddy / cbcn) attach `tool_calls: []` to every
+ * streaming chunk. An empty array is truthy in JS, so the guard
+ * `if (delta.tool_calls)` closed the message on the first content token,
+ * emitting `output_text.done` early and truncating the answer. This mirrors
+ * the real repro: `codex exec -m cbcn/kimi-k3` answered only "cod" instead
+ * of "codex-ok".
+ */
+import { describe, it, expect } from "vitest";
+import { openaiToOpenAIResponsesResponse } from "../../open-sse/translator/response/openai-responses.js";
+import { initState } from "../../open-sse/translator/index.js";
+import { FORMATS } from "../../open-sse/translator/formats.js";
+
+describe("OpenAI Chat stream → Responses: empty tool_calls arrays", () => {
+  it("does not emit output_text.done early when every chunk carries tool_calls: []", () => {
+    const state = initState(FORMATS.OPENAI_RESPONSES);
+    const chunks = [
+      { id: "cmb-test", choices: [{ index: 0, delta: { role: "assistant", content: "", reasoning_content: "", tool_calls: [] }, finish_reason: null }] },
+      { id: "cmb-test", choices: [{ index: 0, delta: { content: "", reasoning_content: "thinking", tool_calls: [] }, finish_reason: null }] },
+      { id: "cmb-test", choices: [{ index: 0, delta: { content: "cod", reasoning_content: "", tool_calls: [] }, finish_reason: null }] },
+      { id: "cmb-test", choices: [{ index: 0, delta: { content: "ex", reasoning_content: "", tool_calls: [] }, finish_reason: null }] },
+      { id: "cmb-test", choices: [{ index: 0, delta: { content: "-ok", reasoning_content: "", tool_calls: [] }, finish_reason: null }] },
+      { id: "cmb-test", choices: [{ index: 0, delta: { content: "", reasoning_content: "", tool_calls: [] }, finish_reason: "stop" }] },
+    ];
+
+    const events = chunks.flatMap((chunk) => openaiToOpenAIResponsesResponse(chunk, state));
+    const textDone = events.filter((e) => e.event === "response.output_text.done");
+    const textDeltas = events.filter((e) => e.event === "response.output_text.delta");
+
+    expect(textDone).toHaveLength(1);
+    expect(textDone[0].data.text).toBe("codex-ok");
+    expect(textDeltas.map((e) => e.data.delta).join("")).toBe("codex-ok");
+    // done must come after every delta
+    expect(events.indexOf(textDone[0])).toBe(events.indexOf(textDeltas[textDeltas.length - 1]) + 1);
+  });
+
+  it("still closes the message before a real tool call", () => {
+    const state = initState(FORMATS.OPENAI_RESPONSES);
+    const chunks = [
+      { id: "cmb-test", choices: [{ index: 0, delta: { content: "Let me run that.", tool_calls: [] }, finish_reason: null }] },
+      { id: "cmb-test", choices: [{ index: 0, delta: { tool_calls: [{ index: 0, id: "call_1", type: "function", function: { name: "exec", arguments: "" } }] }, finish_reason: null }] },
+      { id: "cmb-test", choices: [{ index: 0, delta: {}, finish_reason: "tool_calls" }] },
+    ];
+
+    const events = chunks.flatMap((chunk) => openaiToOpenAIResponsesResponse(chunk, state));
+    const added = events.find((e) => e.event === "response.output_item.added" && e.data.item?.type === "function_call");
+    const textDone = events.find((e) => e.event === "response.output_text.done");
+
+    expect(added).toBeTruthy();
+    expect(textDone.data.text).toBe("Let me run that.");
+  });
+});

tests/unit/opencode-go-models.test.js

@@ -1,71 +1,97 @@
 import { describe, expect, it } from "vitest";
-import { PROVIDER_MODELS, getModelTargetFormat } from "../../open-sse/config/providerModels.js";
-import { OpenCodeGoExecutor } from "../../open-sse/executors/opencode-go.js";
+import { PROVIDER_MODELS, getModelSupportedFormats } from "../../open-sse/config/providerModels.js";
+import { PROVIDERS } from "../../open-sse/config/providers.js";
+import { resolveTransport } from "../../open-sse/services/provider.js";
 
-const CHAT_MODELS = [
-  "glm-5.2",
-  "glm-5.1",
-  // OpenCode Go docs' endpoint table currently says kimi-k2.7, but its
-  // config example and the live API use kimi-k2.7-code.
-  "kimi-k2.7-code",
-  "kimi-k2.6",
-  "deepseek-v4-pro",
-  "deepseek-v4-flash",
-  "mimo-v2.5",
-  "mimo-v2.5-pro",
-];
+// Chat-only models (no /messages, no /responses support on opencode-go)
+const CHAT_ONLY = ["glm-5.2", "glm-5.1", "kimi-k2.7-code", "kimi-k2.6", "mimo-v2.5", "mimo-v2.5-pro"];
+// Models that also expose the Anthropic /messages endpoint
+const CLAUDE_CAPABLE = ["minimax-m3", "minimax-m2.7", "minimax-m2.5", "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus"];
+// Models that also expose the OpenAI /responses endpoint
+const RESPONSES_CAPABLE = ["deepseek-v4-pro", "deepseek-v4-flash"];
 
-const MESSAGES_MODELS = [
-  "minimax-m3",
-  "minimax-m2.7",
-  "minimax-m2.5",
-  "qwen3.7-max",
-  "qwen3.7-plus",
-  "qwen3.6-plus",
-];
+// Mirror of chatCore's per-model transport guard: use the sourceFormat-matched
+// transport only when the model declares support for that sourceFormat.
+function pickTransport(provider, sourceFormat, alias, model) {
+  const supported = getModelSupportedFormats(alias, model);
+  const rt = resolveTransport(provider, sourceFormat);
+  return supported?.includes(sourceFormat) ? rt : null;
+}
 
-describe("OpenCode Go official model catalog", () => {
-  it("matches the documented OpenCode Go model IDs", () => {
-    const ids = (PROVIDER_MODELS["opencode-go"] || []).map((model) => model.id);
-
-    expect(ids).toEqual([...CHAT_MODELS, ...MESSAGES_MODELS]);
+describe("OpenCode Go model catalog", () => {
+  it("matches the documented model IDs", () => {
+    const ids = (PROVIDER_MODELS["opencode-go"] || []).map((m) => m.id);
+    expect(ids).toEqual([
+      "glm-5.2", "glm-5.1", "kimi-k2.7-code", "kimi-k2.6",
+      "deepseek-v4-pro", "deepseek-v4-flash",
+      "mimo-v2.5", "mimo-v2.5-pro",
+      "minimax-m3", "minimax-m2.7", "minimax-m2.5",
+      "qwen3.7-max", "qwen3.7-plus", "qwen3.6-plus",
+    ]);
   });
+});
 
-  it("marks documented Qwen and MiniMax models as Anthropic messages format", () => {
-    for (const model of MESSAGES_MODELS) {
-      expect(getModelTargetFormat("opencode-go", model)).toBe("claude");
+describe("OpenCode Go per-model supportedFormats", () => {
+  it("declares [openai, claude] for MiniMax + Qwen models", () => {
+    for (const m of CLAUDE_CAPABLE) {
+      expect(getModelSupportedFormats("opencode-go", m)).toEqual(["openai", "claude"]);
     }
   });
 
-  it("keeps GLM, Kimi, DeepSeek, and MiMo on OpenAI-compatible chat format", () => {
-    for (const model of CHAT_MODELS) {
-      expect(getModelTargetFormat("opencode-go", model)).toBeNull();
+  it("declares [openai, claude, openai-responses] for DeepSeek models", () => {
+    for (const m of RESPONSES_CAPABLE) {
+      expect(getModelSupportedFormats("opencode-go", m)).toEqual(["openai", "claude", "openai-responses"]);
+    }
+  });
+
+  it("declares [openai] only for chat-only models (GLM/Kimi/MiMo) → guards /messages routing", () => {
+    for (const m of CHAT_ONLY) {
+      expect(getModelSupportedFormats("opencode-go", m)).toEqual(["openai"]);
     }
   });
 });
 
-describe("OpenCode Go endpoint routing", () => {
-  it("routes Qwen and MiniMax models to the messages endpoint with x-api-key auth", () => {
-    const executor = new OpenCodeGoExecutor();
+describe("OpenCode Go multi-endpoint transports", () => {
+  it("declares openai / claude / openai-responses transports", () => {
+    const formats = (PROVIDERS["opencode-go"].transports || []).map((t) => t.format);
+    expect(formats).toEqual(["openai", "claude", "openai-responses"]);
+  });
 
-    for (const model of MESSAGES_MODELS) {
-      expect(executor.buildUrl(model)).toBe("https://opencode.ai/zen/go/v1/messages");
-      const headers = executor.buildHeaders({ apiKey: "sk-test" }, false);
-      expect(headers["x-api-key"]).toBe("sk-test");
-      expect(headers["anthropic-version"]).toBeDefined();
-      expect(headers.Authorization).toBeUndefined();
+  it("resolveTransport picks the endpoint matching the client sourceFormat", () => {
+    expect(resolveTransport("opencode-go", "claude").baseUrl).toBe("https://opencode.ai/zen/go/v1/messages");
+    expect(resolveTransport("opencode-go", "openai-responses").baseUrl).toBe("https://opencode.ai/zen/go/v1/responses");
+    expect(resolveTransport("opencode-go", "openai").baseUrl).toBe("https://opencode.ai/zen/go/v1/chat/completions");
+  });
+
+  it("uses x-api-key + anthropicVersion on the claude transport", () => {
+    const t = resolveTransport("opencode-go", "claude");
+    expect(t.auth.header).toBe("x-api-key");
+    expect(t.auth.anthropicVersion).toBe(true);
+  });
+});
+
+describe("OpenCode Go per-model transport guard (chatCore logic)", () => {
+  it("routes MiniMax/Qwen + claude-format client to /messages", () => {
+    for (const m of CLAUDE_CAPABLE) {
+      expect(pickTransport("opencode-go", "claude", "opencode-go", m)?.baseUrl).toBe("https://opencode.ai/zen/go/v1/messages");
     }
   });
 
-  it("routes GLM, Kimi, DeepSeek, and MiMo models to chat/completions with bearer auth", () => {
-    const executor = new OpenCodeGoExecutor();
+  it("does NOT route chat-only models to /messages on a claude-format request", () => {
+    for (const m of CHAT_ONLY) {
+      expect(pickTransport("opencode-go", "claude", "opencode-go", m)).toBeNull();
+    }
+  });
 
-    for (const model of CHAT_MODELS) {
-      expect(executor.buildUrl(model)).toBe("https://opencode.ai/zen/go/v1/chat/completions");
-      const headers = executor.buildHeaders({ apiKey: "sk-test" }, false);
-      expect(headers.Authorization).toBe("Bearer sk-test");
-      expect(headers["x-api-key"]).toBeUndefined();
-      expect(headers["anthropic-version"]).toBeUndefined();
+  it("routes DeepSeek + responses-format client to /responses", () => {
+    for (const m of RESPONSES_CAPABLE) {
+      expect(pickTransport("opencode-go", "openai-responses", "opencode-go", m)?.baseUrl).toBe("https://opencode.ai/zen/go/v1/responses");
+    }
+  });
+
+  it("does NOT route MiniMax (no responses support) to /responses", () => {
+    for (const m of CLAUDE_CAPABLE) {
+      expect(pickTransport("opencode-go", "openai-responses", "opencode-go", m)).toBeNull();
     }
   });
 });

tests/unit/ping-reasoning-models-3010.test.js

@@ -0,0 +1,77 @@
+// Issue #3010 — Dashboard "Test" button fails for reasoning models because of a
+// tiny max_tokens probe. pingModelByKind must use a sane budget (1024) and treat a
+// reasoning-only (length-limited) response as a successful connection.
+//
+// The route module pulls in Next.js-only deps (@/lib/localDb, etc.) that don't
+// resolve under raw vitest, so we mock them and exercise the exported function.
+
+import { describe, it, expect, vi, beforeEach, afterEach } from "vitest";
+
+// Mock the heavy Next.js-dependent imports BEFORE importing ping.js.
+vi.mock("@/lib/localDb", () => ({ getApiKeys: vi.fn(async () => [{ key: "test-key", isActive: true }]) }));
+vi.mock("@/shared/constants/config", () => ({ UPDATER_CONFIG: { appPort: 20127 } }));
+vi.mock("@/shared/utils/machineId", () => ({ getConsistentMachineId: vi.fn(async () => "cli-token") }));
+
+const { pingModelByKind } = await import("../../src/app/api/models/test/ping.js");
+
+describe("pingModelByKind reasoning models (#3010)", () => {
+  let fetchMock;
+
+  beforeEach(() => {
+    fetchMock = vi.fn();
+    vi.stubGlobal("fetch", fetchMock);
+  });
+
+  afterEach(() => {
+    vi.unstubAllGlobals();
+  });
+
+  function jsonResponse(obj) {
+    return {
+      ok: true,
+      status: 200,
+      text: async () => JSON.stringify(obj),
+      json: async () => obj,
+    };
+  }
+
+  it("uses a 1024-token budget for the chat completions probe", async () => {
+    fetchMock.mockResolvedValue(jsonResponse({ choices: [{ message: { content: "Hi there!" } }] }));
+
+    await pingModelByKind("cline-pass/kimi-k3", "llm", "http://127.0.0.1:20127");
+
+    expect(fetchMock).toHaveBeenCalledTimes(1);
+    const body = JSON.parse(fetchMock.mock.calls[0][1].body);
+    expect(body.max_tokens).toBe(1024);
+  });
+
+  it("treats a reasoning-only (length-limited) response as ok:true", async () => {
+    fetchMock.mockResolvedValue(
+      jsonResponse({
+        choices: [
+          {
+            finish_reason: "length",
+            message: { content: "", reasoning: "The user said hi — a simple greeting..." },
+          },
+        ],
+      })
+    );
+
+    const result = await pingModelByKind("cline-pass/kimi-k3", "llm", "http://127.0.0.1:20127");
+    expect(result.ok).toBe(true);
+    expect(result.note).toMatch(/reasoning-only/);
+  });
+
+  it("still fails when there are no choices and no reasoning", async () => {
+    fetchMock.mockResolvedValue(jsonResponse({ choices: [] }));
+    const result = await pingModelByKind("some/model", "llm", "http://127.0.0.1:20127");
+    expect(result.ok).toBe(false);
+    expect(result.error).toMatch(/no completion choices/);
+  });
+
+  it("passes a normal answer with the larger budget", async () => {
+    fetchMock.mockResolvedValue(jsonResponse({ choices: [{ message: { content: "Hello!" } }] }));
+    const result = await pingModelByKind("openai/gpt-4o", "llm", "http://127.0.0.1:20127");
+    expect(result.ok).toBe(true);
+  });
+});

tests/unit/qoder-billing.test.js

@@ -0,0 +1,147 @@
+/**
+ * Unit tests for qoder billing error detection.
+ *
+ * Ensures that billing blocks (code 112, 10605, pricingUrl) are detected
+ * on the first SSE frame and returned as 403 responses so chatCore can
+ * mark the connection unavailable and trigger combo failover.
+ */
+
+import { describe, it, expect } from "vitest";
+import { __test__ as qoderExecutorInternals } from "../../open-sse/executors/qoder.js";
+
+describe("isBillingBlock", () => {
+  const { isBillingBlock } = qoderExecutorInternals;
+
+  it("detects code 112 (quota exhausted)", () => {
+    const msg = '{"code":"112","message":"Quota exhausted","pricingUrl":"..."}';
+    expect(isBillingBlock(msg)).toBe(true);
+  });
+
+  it("detects code 10605 (queue throttle)", () => {
+    const msg = '{"code":"10605","message":"Queue limit"}';
+    expect(isBillingBlock(msg)).toBe(true);
+  });
+
+  it("detects pricingUrl field", () => {
+    const msg = '{"message":"Upgrade required","pricingUrl":"https://..."}';
+    expect(isBillingBlock(msg)).toBe(true);
+  });
+
+  it("returns false for normal errors without billing markers", () => {
+    const msg = '{"code":"500","message":"Internal error"}';
+    expect(isBillingBlock(msg)).toBe(false);
+  });
+
+  it("returns false for empty or non-string input", () => {
+    expect(isBillingBlock("")).toBe(false);
+    expect(isBillingBlock(null)).toBe(false);
+    expect(isBillingBlock(undefined)).toBe(false);
+  });
+});
+
+describe("wrapQoderSSE billing detection", () => {
+  const { wrapQoderSSE } = qoderExecutorInternals;
+
+  function makeResponse(lines, { status = 200 } = {}) {
+    const body = new ReadableStream({
+      start(controller) {
+        const encoder = new TextEncoder();
+        for (const line of lines) controller.enqueue(encoder.encode(line));
+        controller.close();
+      },
+    });
+    return new Response(body, { status });
+  }
+
+  it("returns 403 response when first frame is billing block (code 112)", async () => {
+    const billingEnv = JSON.stringify({
+      statusCodeValue: 403,
+      body: '{"code":"112","message":"Quota exhausted","pricingUrl":"https://qoder.sh/pricing"}',
+    });
+    const upstream = `data: ${billingEnv}\n\n`;
+
+    const wrapped = await wrapQoderSSE(makeResponse([upstream]), "qoder/ultimate");
+
+    expect(wrapped.status).toBe(403);
+    expect(wrapped.ok).toBe(false);
+    const json = await wrapped.json();
+    expect(json.error).toBeDefined();
+    expect(json.error.message).toContain("112");
+  });
+
+  it("returns 403 response when first frame is billing block (code 10605)", async () => {
+    const billingEnv = JSON.stringify({
+      statusCodeValue: 429,
+      body: '{"code":"10605","message":"Queue limit"}',
+    });
+    const upstream = `data: ${billingEnv}\n\n`;
+
+    const wrapped = await wrapQoderSSE(makeResponse([upstream]), "qoder/ultimate");
+
+    expect(wrapped.status).toBe(403);
+    expect(wrapped.ok).toBe(false);
+  });
+
+  it("returns 403 response when first frame has pricingUrl", async () => {
+    const billingEnv = JSON.stringify({
+      statusCodeValue: 402,
+      body: '{"message":"Payment required","pricingUrl":"https://..."}',
+    });
+    const upstream = `data: ${billingEnv}\n\n`;
+
+    const wrapped = await wrapQoderSSE(makeResponse([upstream]), "qoder/ultimate");
+
+    expect(wrapped.status).toBe(403);
+  });
+
+  it("passes through normal errors (non-billing) as wrapped SSE", async () => {
+    const errorEnv = JSON.stringify({
+      statusCodeValue: 500,
+      body: "Internal server error",
+    });
+    const upstream = `data: ${errorEnv}\n\n`;
+
+    const wrapped = await wrapQoderSSE(makeResponse([upstream]), "qoder/ultimate");
+
+    // Normal error: still 200 response, error text in SSE body
+    expect(wrapped.status).toBe(200);
+    expect(wrapped.ok).toBe(true);
+
+    const reader = wrapped.body.getReader();
+    const decoder = new TextDecoder();
+    let buf = "";
+    while (true) {
+      const { done, value } = await reader.read();
+      if (done) break;
+      buf += decoder.decode(value, { stream: true });
+    }
+    buf += decoder.decode();
+
+    expect(buf).toContain("[qoder error 500");
+    expect(buf).toContain("data: [DONE]");
+  });
+
+  it("passes through successful responses unchanged", async () => {
+    const inner = JSON.stringify({ choices: [{ delta: { content: "hello" } }] });
+    const successEnv = JSON.stringify({ statusCodeValue: 200, body: inner });
+    const upstream = `data: ${successEnv}\n\n`;
+
+    const wrapped = await wrapQoderSSE(makeResponse([upstream]), "qoder/ultimate");
+
+    expect(wrapped.status).toBe(200);
+    expect(wrapped.ok).toBe(true);
+
+    const reader = wrapped.body.getReader();
+    const decoder = new TextDecoder();
+    let buf = "";
+    while (true) {
+      const { done, value } = await reader.read();
+      if (done) break;
+      buf += decoder.decode(value, { stream: true });
+    }
+    buf += decoder.decode();
+
+    expect(buf).toContain(`data: ${inner}`);
+    expect(buf).toContain("data: [DONE]");
+  });
+});

tests/unit/qoder.test.js

@@ -401,7 +401,7 @@ describe("wrapQoderSSE", () => {
   it("forwards an OpenAI envelope chunk and emits [DONE] in flush", async () => {
     const inner = JSON.stringify({ choices: [{ delta: { content: "hi" } }] });
     const upstream = `data: ${JSON.stringify({ statusCodeValue: 200, body: inner })}\n\n`;
-    const wrapped = wrapQoderSSE(makeResponse([upstream]), "qoder/auto");
+    const wrapped = await wrapQoderSSE(makeResponse([upstream]), "qoder/auto");
     const out = await drain(wrapped);
     expect(out).toContain(`data: ${inner}\n\n`);
     expect(out).toContain("data: [DONE]\n\n");
@@ -413,7 +413,7 @@ describe("wrapQoderSSE", () => {
     const inner = JSON.stringify({ choices: [{ delta: { content: "tail" } }], finish_reason: "stop" });
     // Note: NO trailing \n on the final line.
     const upstream = `data: ${JSON.stringify({ statusCodeValue: 200, body: inner })}`;
-    const wrapped = wrapQoderSSE(makeResponse([upstream]), "qoder/auto");
+    const wrapped = await wrapQoderSSE(makeResponse([upstream]), "qoder/auto");
     const out = await drain(wrapped);
     expect(out).toContain(`data: ${inner}\n\n`);
   });
@@ -426,7 +426,7 @@ describe("wrapQoderSSE", () => {
     const errorEnv = JSON.stringify({ statusCodeValue: 500, body: "boom" });
     const validInner = JSON.stringify({ choices: [{ delta: { content: "leak" } }] });
     const validEnv = JSON.stringify({ statusCodeValue: 200, body: validInner });
-    const wrapped = wrapQoderSSE(
+    const wrapped = await wrapQoderSSE(
       makeResponse([`data: ${errorEnv}\n\ndata: ${validEnv}\n\n`]),
       "qoder/auto",
     );
@@ -443,7 +443,7 @@ describe("wrapQoderSSE", () => {
   it("strips embedded newlines from inner body before forwarding", async () => {
     const innerWithNewlines = '{"choices":[{"delta":{"content":"a\nb"}}]}';
     const env = JSON.stringify({ statusCodeValue: 200, body: innerWithNewlines });
-    const wrapped = wrapQoderSSE(makeResponse([`data: ${env}\n\n`]), "qoder/auto");
+    const wrapped = await wrapQoderSSE(makeResponse([`data: ${env}\n\n`]), "qoder/auto");
     const out = await drain(wrapped);
     // The forwarded data: line should be a single event terminated by \n\n
     // and contain no internal \n other than the trailing pair.
@@ -455,15 +455,15 @@ describe("wrapQoderSSE", () => {
 
   it("upstream error envelope produces an error chunk + [DONE]", async () => {
     const env = JSON.stringify({ statusCodeValue: 503, body: "service unavailable" });
-    const wrapped = wrapQoderSSE(makeResponse([`data: ${env}\n\n`]), "qoder/lite");
+    const wrapped = await wrapQoderSSE(makeResponse([`data: ${env}\n\n`]), "qoder/lite");
     const out = await drain(wrapped);
     expect(out).toContain("[qoder error 503");
     expect(out).toContain("data: [DONE]\n\n");
   });
 
-  it("non-ok responses are returned unchanged (no transform)", () => {
+  it("non-ok responses are returned unchanged (no transform)", async () => {
     const r = new Response("not ok", { status: 500 });
-    const wrapped = wrapQoderSSE(r, "qoder/auto");
+    const wrapped = await wrapQoderSSE(r, "qoder/auto");
     expect(wrapped).toBe(r);
   });
 });