fix(thinking): map reasoning_effort to thinkingConfig

2026-02-18 20:30:51 +08:00 · 2026-01-14 22:45:07 +08:00
parent 2262479365
commit 6e4a602c60
13 changed files with 107 additions and 909 deletions
--- a/internal/translator/antigravity/claude/antigravity_claude_request.go
+++ b/internal/translator/antigravity/claude/antigravity_claude_request.go
@@ -13,6 +13,7 @@ import (

 	"github.com/router-for-me/CLIProxyAPI/v6/internal/cache"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/thinking"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/translator/gemini/common"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
 	"github.com/tidwall/gjson"
@@ -123,7 +124,7 @@ func ConvertClaudeRequestToAntigravity(modelName string, inputRawJSON []byte, _
 					contentTypeResult := contentResult.Get("type")
 					if contentTypeResult.Type == gjson.String && contentTypeResult.String() == "thinking" {
 						// Use GetThinkingText to handle wrapped thinking objects
-						thinkingText := util.GetThinkingText(contentResult)
+						thinkingText := thinking.GetThinkingText(contentResult)
 						signatureResult := contentResult.Get("signature")
 						clientSignature := ""
 						if signatureResult.Exists() && signatureResult.String() != "" {
--- a/internal/translator/antigravity/openai/chat-completions/antigravity_openai_request.go
+++ b/internal/translator/antigravity/openai/chat-completions/antigravity_openai_request.go
@@ -36,33 +36,27 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 	// Model
 	out, _ = sjson.SetBytes(out, "model", modelName)

-	// Reasoning effort -> thinkingBudget/include_thoughts
-	// Note: OpenAI official fields take precedence over extra_body.google.thinking_config
+	// Apply thinking configuration: convert OpenAI reasoning_effort to Gemini CLI thinkingConfig.
+	// Inline translation-only mapping; capability checks happen later in ApplyThinking.
+	modelInfo := registry.LookupModelInfo(modelName)
 	re := gjson.GetBytes(rawJSON, "reasoning_effort")
-	hasOfficialThinking := re.Exists()
-	modelInfo := registry.GetGlobalRegistry().GetModelInfo(modelName)
-	if hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil {
+	if re.Exists() {
 		effort := strings.ToLower(strings.TrimSpace(re.String()))
-		if util.IsGemini3Model(modelName) {
-			switch effort {
-			case "none":
-				out, _ = sjson.DeleteBytes(out, "request.generationConfig.thinkingConfig")
-			case "auto":
-				includeThoughts := true
-				out = util.ApplyGeminiCLIThinkingLevel(out, "", &includeThoughts)
-			default:
-				if level, ok := util.ValidateGemini3ThinkingLevel(modelName, effort); ok {
-					out = util.ApplyGeminiCLIThinkingLevel(out, level, nil)
-				}
+		if effort != "" {
+			thinkingPath := "request.generationConfig.thinkingConfig"
+			if effort == "auto" {
+				out, _ = sjson.SetBytes(out, thinkingPath+".thinkingBudget", -1)
+				out, _ = sjson.SetBytes(out, thinkingPath+".includeThoughts", true)
+			} else {
+				out, _ = sjson.SetBytes(out, thinkingPath+".thinkingLevel", effort)
+				out, _ = sjson.SetBytes(out, thinkingPath+".includeThoughts", effort != "none")
 			}
-		} else if len(modelInfo.Thinking.Levels) == 0 {
-			out = util.ApplyReasoningEffortToGeminiCLI(out, effort)
 		}
 	}

 	// Cherry Studio extension extra_body.google.thinking_config (effective only when official fields are absent)
 	// Only apply for models that use numeric budgets, not discrete levels.
-	if !hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
+	if !re.Exists() && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
 			var budget int
--- a/internal/translator/gemini-cli/openai/chat-completions/gemini-cli_openai_request.go
+++ b/internal/translator/gemini-cli/openai/chat-completions/gemini-cli_openai_request.go
@@ -36,18 +36,27 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 	// Model
 	out, _ = sjson.SetBytes(out, "model", modelName)

-	// Reasoning effort -> thinkingBudget/include_thoughts
-	// Note: OpenAI official fields take precedence over extra_body.google.thinking_config
+	// Apply thinking configuration: convert OpenAI reasoning_effort to Gemini CLI thinkingConfig.
+	// Inline translation-only mapping; capability checks happen later in ApplyThinking.
 	re := gjson.GetBytes(rawJSON, "reasoning_effort")
-	hasOfficialThinking := re.Exists()
-	modelInfo := registry.GetGlobalRegistry().GetModelInfo(modelName)
-	if hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
-		out = util.ApplyReasoningEffortToGeminiCLI(out, re.String())
+	modelInfo := registry.LookupModelInfo(modelName)
+	if re.Exists() {
+		effort := strings.ToLower(strings.TrimSpace(re.String()))
+		if effort != "" {
+			thinkingPath := "request.generationConfig.thinkingConfig"
+			if effort == "auto" {
+				out, _ = sjson.SetBytes(out, thinkingPath+".thinkingBudget", -1)
+				out, _ = sjson.SetBytes(out, thinkingPath+".includeThoughts", true)
+			} else {
+				out, _ = sjson.SetBytes(out, thinkingPath+".thinkingLevel", effort)
+				out, _ = sjson.SetBytes(out, thinkingPath+".includeThoughts", effort != "none")
+			}
+		}
 	}

 	// Cherry Studio extension extra_body.google.thinking_config (effective only when official fields are absent)
 	// Only apply for models that use numeric budgets, not discrete levels.
-	if !hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
+	if !re.Exists() && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
 			var budget int
--- a/internal/translator/gemini/openai/chat-completions/gemini_openai_request.go
+++ b/internal/translator/gemini/openai/chat-completions/gemini_openai_request.go
@@ -36,36 +36,27 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 	// Model
 	out, _ = sjson.SetBytes(out, "model", modelName)

-	// Reasoning effort -> thinkingBudget/include_thoughts
-	// Note: OpenAI official fields take precedence over extra_body.google.thinking_config
-	// Only apply numeric budgets for models that use budgets (not discrete levels) to avoid
-	// incorrectly applying thinkingBudget for level-based models like gpt-5. Gemini 3 models
-	// use thinkingLevel/includeThoughts instead.
+	// Apply thinking configuration: convert OpenAI reasoning_effort to Gemini thinkingConfig.
+	// Inline translation-only mapping; capability checks happen later in ApplyThinking.
+	modelInfo := registry.LookupModelInfo(modelName)
 	re := gjson.GetBytes(rawJSON, "reasoning_effort")
-	hasOfficialThinking := re.Exists()
-	modelInfo := registry.GetGlobalRegistry().GetModelInfo(modelName)
-	if hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil {
+	if re.Exists() {
 		effort := strings.ToLower(strings.TrimSpace(re.String()))
-		if util.IsGemini3Model(modelName) {
-			switch effort {
-			case "none":
-				out, _ = sjson.DeleteBytes(out, "generationConfig.thinkingConfig")
-			case "auto":
-				includeThoughts := true
-				out = util.ApplyGeminiThinkingLevel(out, "", &includeThoughts)
-			default:
-				if level, ok := util.ValidateGemini3ThinkingLevel(modelName, effort); ok {
-					out = util.ApplyGeminiThinkingLevel(out, level, nil)
-				}
+		if effort != "" {
+			thinkingPath := "generationConfig.thinkingConfig"
+			if effort == "auto" {
+				out, _ = sjson.SetBytes(out, thinkingPath+".thinkingBudget", -1)
+				out, _ = sjson.SetBytes(out, thinkingPath+".includeThoughts", true)
+			} else {
+				out, _ = sjson.SetBytes(out, thinkingPath+".thinkingLevel", effort)
+				out, _ = sjson.SetBytes(out, thinkingPath+".includeThoughts", effort != "none")
 			}
-		} else if len(modelInfo.Thinking.Levels) == 0 {
-			out = util.ApplyReasoningEffortToGemini(out, effort)
 		}
 	}

 	// Cherry Studio extension extra_body.google.thinking_config (effective only when official fields are absent)
 	// Only apply for models that use numeric budgets, not discrete levels.
-	if !hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
+	if !re.Exists() && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
 		if tc := gjson.GetBytes(rawJSON, "extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
 			var budget int
--- a/internal/translator/gemini/openai/responses/gemini_openai-responses_request.go
+++ b/internal/translator/gemini/openai/responses/gemini_openai-responses_request.go
@@ -6,7 +6,6 @@ import (

 	"github.com/router-for-me/CLIProxyAPI/v6/internal/registry"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/translator/gemini/common"
-	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
 	"github.com/tidwall/gjson"
 	"github.com/tidwall/sjson"
 )
@@ -389,18 +388,27 @@ func ConvertOpenAIResponsesRequestToGemini(modelName string, inputRawJSON []byte
 		out, _ = sjson.Set(out, "generationConfig.stopSequences", sequences)
 	}

-	// OpenAI official reasoning fields take precedence
-	// Only convert for models that use numeric budgets (not discrete levels).
-	hasOfficialThinking := root.Get("reasoning.effort").Exists()
-	modelInfo := registry.GetGlobalRegistry().GetModelInfo(modelName)
-	if hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
-		reasoningEffort := root.Get("reasoning.effort")
-		out = string(util.ApplyReasoningEffortToGemini([]byte(out), reasoningEffort.String()))
+	// Apply thinking configuration: convert OpenAI Responses API reasoning.effort to Gemini thinkingConfig.
+	// Inline translation-only mapping; capability checks happen later in ApplyThinking.
+	modelInfo := registry.LookupModelInfo(modelName)
+	re := root.Get("reasoning.effort")
+	if re.Exists() {
+		effort := strings.ToLower(strings.TrimSpace(re.String()))
+		if effort != "" {
+			thinkingPath := "generationConfig.thinkingConfig"
+			if effort == "auto" {
+				out, _ = sjson.Set(out, thinkingPath+".thinkingBudget", -1)
+				out, _ = sjson.Set(out, thinkingPath+".includeThoughts", true)
+			} else {
+				out, _ = sjson.Set(out, thinkingPath+".thinkingLevel", effort)
+				out, _ = sjson.Set(out, thinkingPath+".includeThoughts", effort != "none")
+			}
+		}
 	}

 	// Cherry Studio extension (applies only when official fields are missing)
 	// Only apply for models that use numeric budgets, not discrete levels.
-	if !hasOfficialThinking && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
+	if !re.Exists() && modelInfo != nil && modelInfo.Thinking != nil && len(modelInfo.Thinking.Levels) == 0 {
 		if tc := root.Get("extra_body.google.thinking_config"); tc.Exists() && tc.IsObject() {
 			var setBudget bool
 			var budget int
--- a/internal/translator/openai/claude/openai_claude_request.go
+++ b/internal/translator/openai/claude/openai_claude_request.go
@@ -10,7 +10,6 @@ import (
 	"strings"

 	"github.com/router-for-me/CLIProxyAPI/v6/internal/thinking"
-	"github.com/router-for-me/CLIProxyAPI/v6/internal/util"
 	"github.com/tidwall/gjson"
 	"github.com/tidwall/sjson"
 )
@@ -130,7 +129,7 @@ func ConvertClaudeRequestToOpenAI(modelName string, inputRawJSON []byte, stream
 					case "thinking":
 						// Only map thinking to reasoning_content for assistant messages (security: prevent injection)
 						if role == "assistant" {
-							thinkingText := util.GetThinkingText(part)
+							thinkingText := thinking.GetThinkingText(part)
 							// Skip empty or whitespace-only thinking
 							if strings.TrimSpace(thinkingText) != "" {
 								reasoningParts = append(reasoningParts, thinkingText)