Merge upstream/main (08779cc) - sync with original repo updates

internal/runtime/executor/aistudio_executor.go

@@ -393,7 +393,7 @@ func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts c
 	}
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, stream)
 	payload := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), stream)
-	payload, err := thinking.ApplyThinking(payload, req.Model, "gemini")
+	payload, err := thinking.ApplyThinking(payload, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, translatedPayload{}, err
 	}

internal/runtime/executor/antigravity_executor.go

@@ -137,7 +137,7 @@ func (e *AntigravityExecutor) Execute(ctx context.Context, auth *cliproxyauth.Au
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, false)
 	translated := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	translated, err = thinking.ApplyThinking(translated, req.Model, "antigravity")
+	translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -256,7 +256,7 @@ func (e *AntigravityExecutor) executeClaudeNonStream(ctx context.Context, auth *
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
 	translated := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 
-	translated, err = thinking.ApplyThinking(translated, req.Model, "antigravity")
+	translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -622,7 +622,7 @@ func (e *AntigravityExecutor) ExecuteStream(ctx context.Context, auth *cliproxya
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
 	translated := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 
-	translated, err = thinking.ApplyThinking(translated, req.Model, "antigravity")
+	translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -802,7 +802,7 @@ func (e *AntigravityExecutor) CountTokens(ctx context.Context, auth *cliproxyaut
 	// Prepare payload once (doesn't depend on baseURL)
 	payload := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	payload, err := thinking.ApplyThinking(payload, req.Model, "antigravity")
+	payload, err := thinking.ApplyThinking(payload, req.Model, from.String(), to.String())
 	if err != nil {
 		return cliproxyexecutor.Response{}, err
 	}
@@ -1005,9 +1005,6 @@ func FetchAntigravityModels(ctx context.Context, auth *cliproxyauth.Auth, cfg *c
 			}
 			modelCfg := modelConfig[modelID]
 			modelName := modelID
-			if modelCfg != nil && modelCfg.Name != "" {
-				modelName = modelCfg.Name
-			}
 			modelInfo := &registry.ModelInfo{
 				ID:          modelID,
 				Name:        modelName,
@@ -1410,13 +1407,6 @@ func geminiToAntigravity(modelName string, payload []byte, projectID string) []b
 	template, _ = sjson.Delete(template, "request.safetySettings")
 	template, _ = sjson.Set(template, "request.toolConfig.functionCallingConfig.mode", "VALIDATED")
 
-	if !strings.HasPrefix(modelName, "gemini-3-") {
-		if thinkingLevel := gjson.Get(template, "request.generationConfig.thinkingConfig.thinkingLevel"); thinkingLevel.Exists() {
-			template, _ = sjson.Delete(template, "request.generationConfig.thinkingConfig.thinkingLevel")
-			template, _ = sjson.Set(template, "request.generationConfig.thinkingConfig.thinkingBudget", -1)
-		}
-	}
-
 	if strings.Contains(modelName, "claude") {
 		gjson.Get(template, "request.tools").ForEach(func(key, tool gjson.Result) bool {
 			tool.Get("functionDeclarations").ForEach(func(funKey, funcDecl gjson.Result) bool {

internal/runtime/executor/claude_executor.go

@@ -17,7 +17,6 @@ import (
 	claudeauth "github.com/router-for-me/CLIProxyAPI/v6/internal/auth/claude"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/misc"
-	"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/util"
 	cliproxyauth "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/auth"
@@ -106,7 +105,7 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), stream)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "claude")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -119,9 +118,6 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	// Disable thinking if tool_choice forces tool use (Anthropic API constraint)
 	body = disableThinkingIfToolChoiceForced(body)
 
-	// Ensure max_tokens > thinking.budget_tokens when thinking is enabled
-	body = ensureMaxTokensForThinking(baseModel, body)
-
 	// Extract betas from body and convert to header
 	var extraBetas []string
 	extraBetas, body = extractAndRemoveBetas(body)
@@ -239,7 +235,7 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "claude")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -250,9 +246,6 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	// Disable thinking if tool_choice forces tool use (Anthropic API constraint)
 	body = disableThinkingIfToolChoiceForced(body)
 
-	// Ensure max_tokens > thinking.budget_tokens when thinking is enabled
-	body = ensureMaxTokensForThinking(baseModel, body)
-
 	// Extract betas from body and convert to header
 	var extraBetas []string
 	extraBetas, body = extractAndRemoveBetas(body)
@@ -541,81 +534,6 @@ func disableThinkingIfToolChoiceForced(body []byte) []byte {
 	return body
 }
 
-// ensureMaxTokensForThinking ensures max_tokens > thinking.budget_tokens when thinking is enabled.
-// Anthropic API requires this constraint; violating it returns a 400 error.
-// This function should be called after all thinking configuration is finalized.
-// It looks up the model's MaxCompletionTokens from the registry to use as the cap.
-func ensureMaxTokensForThinking(modelName string, body []byte) []byte {
-	thinkingType := gjson.GetBytes(body, "thinking.type").String()
-	if thinkingType != "enabled" {
-		return body
-	}
-
-	budgetTokens := gjson.GetBytes(body, "thinking.budget_tokens").Int()
-	if budgetTokens <= 0 {
-		return body
-	}
-
-	maxTokens := gjson.GetBytes(body, "max_tokens").Int()
-
-	// Look up the model's max completion tokens from the registry
-	maxCompletionTokens := 0
-	if modelInfo := registry.LookupModelInfo(modelName); modelInfo != nil {
-		maxCompletionTokens = modelInfo.MaxCompletionTokens
-	}
-
-	// Fall back to budget + buffer if registry lookup fails or returns 0
-	const fallbackBuffer = 4000
-	requiredMaxTokens := budgetTokens + fallbackBuffer
-	if maxCompletionTokens > 0 {
-		requiredMaxTokens = int64(maxCompletionTokens)
-	}
-
-	if maxTokens < requiredMaxTokens {
-		body, _ = sjson.SetBytes(body, "max_tokens", requiredMaxTokens)
-	}
-	return body
-}
-
-func (e *ClaudeExecutor) resolveClaudeConfig(auth *cliproxyauth.Auth) *config.ClaudeKey {
-	if auth == nil || e.cfg == nil {
-		return nil
-	}
-	var attrKey, attrBase string
-	if auth.Attributes != nil {
-		attrKey = strings.TrimSpace(auth.Attributes["api_key"])
-		attrBase = strings.TrimSpace(auth.Attributes["base_url"])
-	}
-	for i := range e.cfg.ClaudeKey {
-		entry := &e.cfg.ClaudeKey[i]
-		cfgKey := strings.TrimSpace(entry.APIKey)
-		cfgBase := strings.TrimSpace(entry.BaseURL)
-		if attrKey != "" && attrBase != "" {
-			if strings.EqualFold(cfgKey, attrKey) && strings.EqualFold(cfgBase, attrBase) {
-				return entry
-			}
-			continue
-		}
-		if attrKey != "" && strings.EqualFold(cfgKey, attrKey) {
-			if cfgBase == "" || strings.EqualFold(cfgBase, attrBase) {
-				return entry
-			}
-		}
-		if attrKey == "" && attrBase != "" && strings.EqualFold(cfgBase, attrBase) {
-			return entry
-		}
-	}
-	if attrKey != "" {
-		for i := range e.cfg.ClaudeKey {
-			entry := &e.cfg.ClaudeKey[i]
-			if strings.EqualFold(strings.TrimSpace(entry.APIKey), attrKey) {
-				return entry
-			}
-		}
-	}
-	return nil
-}
-
 type compositeReadCloser struct {
 	io.Reader
 	closers []func() error

internal/runtime/executor/codex_executor.go

@@ -96,7 +96,7 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	body = sdktranslator.TranslateRequest(from, to, baseModel, body, false)
 	body = misc.StripCodexUserAgent(body)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "codex")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -208,7 +208,7 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	body = sdktranslator.TranslateRequest(from, to, baseModel, body, true)
 	body = misc.StripCodexUserAgent(body)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "codex")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -316,7 +316,7 @@ func (e *CodexExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Auth
 	body = sdktranslator.TranslateRequest(from, to, baseModel, body, false)
 	body = misc.StripCodexUserAgent(body)
 
-	body, err := thinking.ApplyThinking(body, req.Model, "codex")
+	body, err := thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return cliproxyexecutor.Response{}, err
 	}

internal/runtime/executor/gemini_cli_executor.go

@@ -123,7 +123,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, false)
 	basePayload := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	basePayload, err = thinking.ApplyThinking(basePayload, req.Model, "gemini-cli")
+	basePayload, err = thinking.ApplyThinking(basePayload, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -272,7 +272,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
 	basePayload := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 
-	basePayload, err = thinking.ApplyThinking(basePayload, req.Model, "gemini-cli")
+	basePayload, err = thinking.ApplyThinking(basePayload, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -479,7 +479,7 @@ func (e *GeminiCLIExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.
 	for range models {
 		payload := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-		payload, err = thinking.ApplyThinking(payload, req.Model, "gemini-cli")
+		payload, err = thinking.ApplyThinking(payload, req.Model, from.String(), to.String())
 		if err != nil {
 			return cliproxyexecutor.Response{}, err
 		}

internal/runtime/executor/gemini_executor.go

@@ -120,7 +120,7 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, false)
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "gemini")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -222,7 +222,7 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "gemini")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -338,7 +338,7 @@ func (e *GeminiExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Aut
 	to := sdktranslator.FromString("gemini")
 	translatedReq := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	translatedReq, err := thinking.ApplyThinking(translatedReq, req.Model, "gemini")
+	translatedReq, err := thinking.ApplyThinking(translatedReq, req.Model, from.String(), to.String())
 	if err != nil {
 		return cliproxyexecutor.Response{}, err
 	}

internal/runtime/executor/gemini_vertex_executor.go

@@ -170,7 +170,7 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, false)
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "gemini")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -272,7 +272,7 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, false)
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "gemini")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -375,7 +375,7 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "gemini")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -494,7 +494,7 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "gemini")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -605,7 +605,7 @@ func (e *GeminiVertexExecutor) countTokensWithServiceAccount(ctx context.Context
 
 	translatedReq := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	translatedReq, err := thinking.ApplyThinking(translatedReq, req.Model, "gemini")
+	translatedReq, err := thinking.ApplyThinking(translatedReq, req.Model, from.String(), to.String())
 	if err != nil {
 		return cliproxyexecutor.Response{}, err
 	}
@@ -689,7 +689,7 @@ func (e *GeminiVertexExecutor) countTokensWithAPIKey(ctx context.Context, auth *
 
 	translatedReq := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 
-	translatedReq, err := thinking.ApplyThinking(translatedReq, req.Model, "gemini")
+	translatedReq, err := thinking.ApplyThinking(translatedReq, req.Model, from.String(), to.String())
 	if err != nil {
 		return cliproxyexecutor.Response{}, err
 	}

internal/runtime/executor/iflow_executor.go

@@ -92,7 +92,7 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "iflow")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), "iflow")
 	if err != nil {
 		return resp, err
 	}
@@ -190,7 +190,7 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "iflow")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), "iflow")
 	if err != nil {
 		return nil, err
 	}

internal/runtime/executor/openai_compat_executor.go

@@ -92,7 +92,7 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 	translated := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), opts.Stream)
 	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated)
 
-	translated, err = thinking.ApplyThinking(translated, req.Model, "openai")
+	translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -187,7 +187,7 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 	translated := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated)
 
-	translated, err = thinking.ApplyThinking(translated, req.Model, "openai")
+	translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}
@@ -297,7 +297,7 @@ func (e *OpenAICompatExecutor) CountTokens(ctx context.Context, auth *cliproxyau
 
 	modelForCounting := baseModel
 
-	translated, err := thinking.ApplyThinking(translated, req.Model, "openai")
+	translated, err := thinking.ApplyThinking(translated, req.Model, from.String(), to.String())
 	if err != nil {
 		return cliproxyexecutor.Response{}, err
 	}

internal/runtime/executor/qwen_executor.go

@@ -86,7 +86,7 @@ func (e *QwenExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), false)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "openai")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return resp, err
 	}
@@ -172,7 +172,7 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 	body := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
-	body, err = thinking.ApplyThinking(body, req.Model, "openai")
+	body, err = thinking.ApplyThinking(body, req.Model, from.String(), to.String())
 	if err != nil {
 		return nil, err
 	}