fix: granular reasoning_effort handling for Claude models (#791)

- github.js: split thinking vs reasoning_effort stripping
  - thinking (Claude-native format) still stripped for all Claude on Copilot
  - reasoning_effort now passed through for Opus 4.6 and Sonnet 4.6
  - still stripped for Haiku 4.5 and Opus 4.7 (rejected upstream)
  - reasoning_effort "none" stripped for all models (not all support it)
- openai-to-claude.js: map reasoning_effort → thinking.budget_tokens
  for direct Anthropic backend (none→skip, low→4096, medium→8192,
  high→16384, xhigh→32768)

Previously reasoning_effort was stripped for ALL Claude models,
meaning Opus 4.6 via Copilot never received thinking configuration.

AI-generated commit by Claude Opus 4.6 (Anthropic)

open-sse/executors/github.js

@@ -121,6 +121,19 @@ export class GithubExecutor extends BaseExecutor {
     return !/claude/i.test(model);
   }
 
+  // reasoning_effort works for GPT-5 family AND Claude Opus 4.6 / Sonnet 4.6
+  // on GitHub Copilot. Only strip for models that don't support it:
+  // Claude Haiku 4.5, Claude Opus 4.7 (rejected upstream).
+  supportsReasoningEffort(model) {
+    const m = model.toLowerCase();
+    // Claude models that DO support reasoning_effort
+    if (/claude.*opus.*4\.6/i.test(m) || /claude.*sonnet.*4\.6/i.test(m)) return true;
+    // All other Claude models: strip
+    if (/claude/i.test(model)) return false;
+    // GPT-5 family, Gemini, etc.: keep
+    return true;
+  }
+
   transformRequest(model, body, stream, credentials) {
     const transformed = { ...body };
     if (this.requiresMaxCompletionTokens(model) && transformed.max_tokens !== undefined) {
@@ -131,9 +144,16 @@ export class GithubExecutor extends BaseExecutor {
     if (!this.supportsTemperature(model) && transformed.temperature !== undefined) {
       delete transformed.temperature;
     }
-    // Strip thinking/reasoning_effort — unsupported on /chat/completions
+    // Always strip Claude-style thinking payload (Copilot doesn't understand it)
     if (!this.supportsThinking(model)) {
       delete transformed.thinking;
+    }
+    // "none" means no thinking — strip it so models that don't support "none" don't 400
+    if (transformed.reasoning_effort === "none") {
+      delete transformed.reasoning_effort;
+    }
+    // Strip reasoning_effort only for models that reject it
+    if (!this.supportsReasoningEffort(model) && transformed.reasoning_effort !== undefined) {
       delete transformed.reasoning_effort;
     }
     return transformed;

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

@@ -179,6 +179,25 @@ Respond ONLY with the JSON object, no other text.`);
     };
   }
 
+  // Map OpenAI reasoning_effort → Claude thinking.budget_tokens
+  // When client sends reasoning_effort (OpenAI format) but no explicit thinking block,
+  // translate to Claude's native format.
+  if (body.reasoning_effort && !result.thinking) {
+    const effortToBudget = {
+      none:   0,
+      low:    4096,
+      medium: 8192,
+      high:   16384,
+      xhigh:  32768,
+    };
+    const budget = effortToBudget[body.reasoning_effort.toLowerCase()];
+    if (budget === 0) {
+      // none → no thinking
+    } else if (budget) {
+      result.thinking = { type: "enabled", budget_tokens: budget };
+    }
+  }
+
   // Attach toolNameMap to result for response translation
   if (toolNameMap.size > 0) {
     result._toolNameMap = toolNameMap;