Add config for adaptive thinking Fixes #5236

3070cf26cd9c1bc10e47f32163afdc804a9e1add

Cohee <18619528+Cohee1207@users.noreply.github.com>

default/config.yaml+5 -1 Viewed

   # Otherwise, you'll just waste money on cache misses.
   enableSystemPromptCache: false
   # Enables caching of the message history at depth (if supported).
   # https://docsplatform.anthropicclaude.com/en/docs/en/build-with-claude/prompt-caching
   # -- IMPORTANT! --
   # Use with caution. Behavior may be unpredictable and no guarantees can or will be made.
   # Set to an integer to specify the desired depth. 0 (which does NOT include the prefill)

   ## 5m: base price x 1.25
   ## 1h: base price x 2
   extendedTTL: false
 +  # Enables adaptive thinking for supported models (Opus 4.6+).
 +  # Disable to enforce legacy thinking mode (with thinking budget).
 +  # https://platform.claude.com/docs/en/build-with-claude/adaptive-thinking
 +  enableAdaptiveThinking: true
 # -- GOOGLE GEMINI API CONFIGURATION --
 gemini:
   # API endpoint version ("v1beta" or "v1alpha")

src/endpoints/backends/chat-completions.js+2 -1 Viewed

     const value = getConfigValue('claude.cachingAtDepth', -1, 'number');
     return Number.isInteger(value) && value >= 0 ? value : -1;
 })();
 +const enableAdaptiveThinking = getConfigValue('claude.enableAdaptiveThinking', true, 'boolean');
 /**
  * Cache for cacheable (writing) OpenRouter model IDs.

         const isLimitedSampling = /^claude-(opus-4-1|sonnet-4-5|haiku-4-5|opus-4-5|opus-4-6|sonnet-4-6)/.test(request.body.model);
         const useVerbosity = /^claude-(opus-4-5|opus-4-6|sonnet-4-6)/.test(request.body.model);
         const noPrefillModel = /^claude-(opus-4-6|sonnet-4-6)/.test(request.body.model);
         const isAdaptiveModel = enableAdaptiveThinking && /^claude-(opus-4-6|sonnet-4-6)/.test(request.body.model);
         let fixThinkingPrefill = false;
         // Add custom stop sequences
         const stopSequences = [];