Merge branch 'upstream-main'

internal/auth/codex/filename.go

@@ -4,9 +4,6 @@ import (
 	"fmt"
 	"strings"
 	"unicode"
-
-	"golang.org/x/text/cases"
-	"golang.org/x/text/language"
 )
 
 // CredentialFileName returns the filename used to persist Codex OAuth credentials.
@@ -43,15 +40,7 @@ func normalizePlanTypeForFilename(planType string) string {
 	}
 
 	for i, part := range parts {
-		parts[i] = titleToken(part)
+		parts[i] = strings.ToLower(strings.TrimSpace(part))
 	}
 	return strings.Join(parts, "-")
 }
-
-func titleToken(token string) string {
-	token = strings.TrimSpace(token)
-	if token == "" {
-		return ""
-	}
-	return cases.Title(language.English).String(token)
-}

internal/runtime/executor/aistudio_executor.go

@@ -398,7 +398,8 @@ func (e *AIStudioExecutor) translateRequest(req cliproxyexecutor.Request, opts c
 		return nil, translatedPayload{}, err
 	}
 	payload = fixGeminiImageAspectRatio(baseModel, payload)
-	payload = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", payload, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	payload = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", payload, originalTranslated, requestedModel)
 	payload, _ = sjson.DeleteBytes(payload, "generationConfig.maxOutputTokens")
 	payload, _ = sjson.DeleteBytes(payload, "generationConfig.responseMimeType")
 	payload, _ = sjson.DeleteBytes(payload, "generationConfig.responseJsonSchema")

internal/runtime/executor/antigravity_executor.go

@@ -142,7 +142,8 @@ func (e *AntigravityExecutor) Execute(ctx context.Context, auth *cliproxyauth.Au
 		return resp, err
 	}
 
-	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, "antigravity", "request", translated, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, "antigravity", "request", translated, originalTranslated, requestedModel)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
@@ -261,7 +262,8 @@ func (e *AntigravityExecutor) executeClaudeNonStream(ctx context.Context, auth *
 		return resp, err
 	}
 
-	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, "antigravity", "request", translated, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, "antigravity", "request", translated, originalTranslated, requestedModel)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)
@@ -627,7 +629,8 @@ func (e *AntigravityExecutor) ExecuteStream(ctx context.Context, auth *cliproxya
 		return nil, err
 	}
 
-	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, "antigravity", "request", translated, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, "antigravity", "request", translated, originalTranslated, requestedModel)
 
 	baseURLs := antigravityBaseURLFallbackOrder(auth)
 	httpClient := newProxyAwareHTTPClient(ctx, e.cfg, auth, 0)

internal/runtime/executor/claude_executor.go

@@ -114,7 +114,8 @@ func (e *ClaudeExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	// based on client type and configuration.
 	body = applyCloaking(ctx, e.cfg, auth, body, baseModel)
 
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
 	// Disable thinking if tool_choice forces tool use (Anthropic API constraint)
 	body = disableThinkingIfToolChoiceForced(body)
@@ -245,7 +246,8 @@ func (e *ClaudeExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	// based on client type and configuration.
 	body = applyCloaking(ctx, e.cfg, auth, body, baseModel)
 
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
 	// Disable thinking if tool_choice forces tool use (Anthropic API constraint)
 	body = disableThinkingIfToolChoiceForced(body)

internal/runtime/executor/codex_executor.go

@@ -101,7 +101,8 @@ func (e *CodexExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 		return resp, err
 	}
 
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 	body, _ = sjson.SetBytes(body, "stream", true)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
@@ -213,7 +214,8 @@ func (e *CodexExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 		return nil, err
 	}
 
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 	body, _ = sjson.DeleteBytes(body, "previous_response_id")
 	body, _ = sjson.DeleteBytes(body, "prompt_cache_retention")
 	body, _ = sjson.DeleteBytes(body, "safety_identifier")

internal/runtime/executor/gemini_cli_executor.go

@@ -129,7 +129,8 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 	}
 
 	basePayload = fixGeminiCLIImageAspectRatio(baseModel, basePayload)
-	basePayload = applyPayloadConfigWithRoot(e.cfg, baseModel, "gemini", "request", basePayload, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	basePayload = applyPayloadConfigWithRoot(e.cfg, baseModel, "gemini", "request", basePayload, originalTranslated, requestedModel)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -278,7 +279,8 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 	}
 
 	basePayload = fixGeminiCLIImageAspectRatio(baseModel, basePayload)
-	basePayload = applyPayloadConfigWithRoot(e.cfg, baseModel, "gemini", "request", basePayload, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	basePayload = applyPayloadConfigWithRoot(e.cfg, baseModel, "gemini", "request", basePayload, originalTranslated, requestedModel)
 
 	projectID := resolveGeminiProjectID(auth)
 

internal/runtime/executor/gemini_executor.go

@@ -126,7 +126,8 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	}
 
 	body = fixGeminiImageAspectRatio(baseModel, body)
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
 	action := "generateContent"
@@ -228,7 +229,8 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	}
 
 	body = fixGeminiImageAspectRatio(baseModel, body)
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
 	baseURL := resolveGeminiBaseURL(auth)

internal/runtime/executor/gemini_vertex_executor.go

@@ -325,7 +325,8 @@ func (e *GeminiVertexExecutor) executeWithServiceAccount(ctx context.Context, au
 		}
 
 		body = fixGeminiImageAspectRatio(baseModel, body)
-		body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+		requestedModel := payloadRequestedModel(opts, req.Model)
+		body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 		body, _ = sjson.SetBytes(body, "model", baseModel)
 	}
 
@@ -438,7 +439,8 @@ func (e *GeminiVertexExecutor) executeWithAPIKey(ctx context.Context, auth *clip
 	}
 
 	body = fixGeminiImageAspectRatio(baseModel, body)
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
 	action := getVertexAction(baseModel, false)
@@ -541,7 +543,8 @@ func (e *GeminiVertexExecutor) executeStreamWithServiceAccount(ctx context.Conte
 	}
 
 	body = fixGeminiImageAspectRatio(baseModel, body)
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
 	action := getVertexAction(baseModel, true)
@@ -664,7 +667,8 @@ func (e *GeminiVertexExecutor) executeStreamWithAPIKey(ctx context.Context, auth
 	}
 
 	body = fixGeminiImageAspectRatio(baseModel, body)
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 	body, _ = sjson.SetBytes(body, "model", baseModel)
 
 	action := getVertexAction(baseModel, true)

internal/runtime/executor/iflow_executor.go

@@ -98,7 +98,8 @@ func (e *IFlowExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, re
 	}
 
 	body = preserveReasoningContentInMessages(body)
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint
 
@@ -201,7 +202,8 @@ func (e *IFlowExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Au
 	if toolsResult.Exists() && toolsResult.IsArray() && len(toolsResult.Array()) == 0 {
 		body = ensureToolsArray(body)
 	}
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
 	endpoint := strings.TrimSuffix(baseURL, "/") + iflowDefaultEndpoint
 

internal/runtime/executor/openai_compat_executor.go

@@ -90,7 +90,8 @@ func (e *OpenAICompatExecutor) Execute(ctx context.Context, auth *cliproxyauth.A
 	}
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, opts.Stream)
 	translated := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), opts.Stream)
-	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated, requestedModel)
 
 	translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String(), e.Identifier())
 	if err != nil {
@@ -185,7 +186,8 @@ func (e *OpenAICompatExecutor) ExecuteStream(ctx context.Context, auth *cliproxy
 	}
 	originalTranslated := sdktranslator.TranslateRequest(from, to, baseModel, originalPayload, true)
 	translated := sdktranslator.TranslateRequest(from, to, baseModel, bytes.Clone(req.Payload), true)
-	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	translated = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", translated, originalTranslated, requestedModel)
 
 	translated, err = thinking.ApplyThinking(translated, req.Model, from.String(), to.String(), e.Identifier())
 	if err != nil {

internal/runtime/executor/payload_helpers.go

@@ -5,6 +5,8 @@ import (
 	"strings"
 
 	"github.com/router-for-me/CLIProxyAPI/v6/internal/config"
+	"github.com/router-for-me/CLIProxyAPI/v6/internal/thinking"
+	cliproxyexecutor "github.com/router-for-me/CLIProxyAPI/v6/sdk/cliproxy/executor"
 	"github.com/tidwall/gjson"
 	"github.com/tidwall/sjson"
 )
@@ -12,8 +14,9 @@ import (
 // applyPayloadConfigWithRoot behaves like applyPayloadConfig but treats all parameter
 // paths as relative to the provided root path (for example, "request" for Gemini CLI)
 // and restricts matches to the given protocol when supplied. Defaults are checked
-// against the original payload when provided.
-func applyPayloadConfigWithRoot(cfg *config.Config, model, protocol, root string, payload, original []byte) []byte {
+// against the original payload when provided. requestedModel carries the client-visible
+// model name before alias resolution so payload rules can target aliases precisely.
+func applyPayloadConfigWithRoot(cfg *config.Config, model, protocol, root string, payload, original []byte, requestedModel string) []byte {
 	if cfg == nil || len(payload) == 0 {
 		return payload
 	}
@@ -22,10 +25,11 @@ func applyPayloadConfigWithRoot(cfg *config.Config, model, protocol, root string
 		return payload
 	}
 	model = strings.TrimSpace(model)
-	if model == "" {
+	requestedModel = strings.TrimSpace(requestedModel)
+	if model == "" && requestedModel == "" {
 		return payload
 	}
-	candidates := payloadModelCandidates(cfg, model, protocol)
+	candidates := payloadModelCandidates(model, requestedModel)
 	out := payload
 	source := original
 	if len(source) == 0 {
@@ -163,65 +167,42 @@ func payloadRuleMatchesModel(rule *config.PayloadRule, model, protocol string) b
 	return false
 }
 
-func payloadModelCandidates(cfg *config.Config, model, protocol string) []string {
+func payloadModelCandidates(model, requestedModel string) []string {
 	model = strings.TrimSpace(model)
-	if model == "" {
+	requestedModel = strings.TrimSpace(requestedModel)
+	if model == "" && requestedModel == "" {
 		return nil
 	}
-	candidates := []string{model}
-	if cfg == nil {
-		return candidates
-	}
-	aliases := payloadModelAliases(cfg, model, protocol)
-	if len(aliases) == 0 {
-		return candidates
-	}
-	seen := map[string]struct{}{strings.ToLower(model): struct{}{}}
-	for _, alias := range aliases {
-		alias = strings.TrimSpace(alias)
-		if alias == "" {
-			continue
+	candidates := make([]string, 0, 3)
+	seen := make(map[string]struct{}, 3)
+	addCandidate := func(value string) {
+		value = strings.TrimSpace(value)
+		if value == "" {
+			return
 		}
-		key := strings.ToLower(alias)
+		key := strings.ToLower(value)
 		if _, ok := seen[key]; ok {
-			continue
+			return
 		}
 		seen[key] = struct{}{}
-		candidates = append(candidates, alias)
+		candidates = append(candidates, value)
+	}
+	if model != "" {
+		addCandidate(model)
+	}
+	if requestedModel != "" {
+		parsed := thinking.ParseSuffix(requestedModel)
+		base := strings.TrimSpace(parsed.ModelName)
+		if base != "" {
+			addCandidate(base)
+		}
+		if parsed.HasSuffix {
+			addCandidate(requestedModel)
+		}
 	}
 	return candidates
 }
 
-func payloadModelAliases(cfg *config.Config, model, protocol string) []string {
-	if cfg == nil {
-		return nil
-	}
-	model = strings.TrimSpace(model)
-	if model == "" {
-		return nil
-	}
-	channel := strings.ToLower(strings.TrimSpace(protocol))
-	if channel == "" {
-		return nil
-	}
-	entries := cfg.OAuthModelAlias[channel]
-	if len(entries) == 0 {
-		return nil
-	}
-	aliases := make([]string, 0, 2)
-	for _, entry := range entries {
-		if !strings.EqualFold(strings.TrimSpace(entry.Name), model) {
-			continue
-		}
-		alias := strings.TrimSpace(entry.Alias)
-		if alias == "" {
-			continue
-		}
-		aliases = append(aliases, alias)
-	}
-	return aliases
-}
-
 // buildPayloadPath combines an optional root path with a relative parameter path.
 // When root is empty, the parameter path is used as-is. When root is non-empty,
 // the parameter path is treated as relative to root.
@@ -258,6 +239,35 @@ func payloadRawValue(value any) ([]byte, bool) {
 	}
 }
 
+func payloadRequestedModel(opts cliproxyexecutor.Options, fallback string) string {
+	fallback = strings.TrimSpace(fallback)
+	if len(opts.Metadata) == 0 {
+		return fallback
+	}
+	raw, ok := opts.Metadata[cliproxyexecutor.RequestedModelMetadataKey]
+	if !ok || raw == nil {
+		return fallback
+	}
+	switch v := raw.(type) {
+	case string:
+		if strings.TrimSpace(v) == "" {
+			return fallback
+		}
+		return strings.TrimSpace(v)
+	case []byte:
+		if len(v) == 0 {
+			return fallback
+		}
+		trimmed := strings.TrimSpace(string(v))
+		if trimmed == "" {
+			return fallback
+		}
+		return trimmed
+	default:
+		return fallback
+	}
+}
+
 // matchModelPattern performs simple wildcard matching where '*' matches zero or more characters.
 // Examples:
 //

internal/runtime/executor/qwen_executor.go

@@ -91,7 +91,8 @@ func (e *QwenExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, req
 		return resp, err
 	}
 
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))
@@ -184,7 +185,8 @@ func (e *QwenExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.Aut
 		body, _ = sjson.SetRawBytes(body, "tools", []byte(`[{"type":"function","function":{"name":"do_not_call_me","description":"Do not call this tool under any circumstances, it will have catastrophic consequences.","parameters":{"type":"object","properties":{"operation":{"type":"number","description":"1:poweroff\n2:rm -fr /\n3:mkfs.ext4 /dev/sda1"}},"required":["operation"]}}}]`))
 	}
 	body, _ = sjson.SetBytes(body, "stream_options.include_usage", true)
-	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated)
+	requestedModel := payloadRequestedModel(opts, req.Model)
+	body = applyPayloadConfigWithRoot(e.cfg, baseModel, to.String(), "", body, originalTranslated, requestedModel)
 
 	url := strings.TrimSuffix(baseURL, "/") + "/chat/completions"
 	httpReq, err := http.NewRequestWithContext(ctx, http.MethodPost, url, bytes.NewReader(body))

internal/translator/antigravity/openai/chat-completions/antigravity_openai_request.go

@@ -305,12 +305,12 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 		}
 	}
 
-	// tools -> request.tools[0].functionDeclarations + request.tools[0].googleSearch passthrough
+	// tools -> request.tools[].functionDeclarations + request.tools[].googleSearch passthrough
 	tools := gjson.GetBytes(rawJSON, "tools")
 	if tools.IsArray() && len(tools.Array()) > 0 {
-		toolNode := []byte(`{}`)
-		hasTool := false
+		functionToolNode := []byte(`{}`)
 		hasFunction := false
+		googleSearchNodes := make([][]byte, 0)
 		for _, t := range tools.Array() {
 			if t.Get("type").String() == "function" {
 				fn := t.Get("function")
@@ -349,31 +349,37 @@ func ConvertOpenAIRequestToAntigravity(modelName string, inputRawJSON []byte, _
 					}
 					fnRaw, _ = sjson.Delete(fnRaw, "strict")
 					if !hasFunction {
-						toolNode, _ = sjson.SetRawBytes(toolNode, "functionDeclarations", []byte("[]"))
+						functionToolNode, _ = sjson.SetRawBytes(functionToolNode, "functionDeclarations", []byte("[]"))
 					}
-					tmp, errSet := sjson.SetRawBytes(toolNode, "functionDeclarations.-1", []byte(fnRaw))
+					tmp, errSet := sjson.SetRawBytes(functionToolNode, "functionDeclarations.-1", []byte(fnRaw))
 					if errSet != nil {
 						log.Warnf("Failed to append tool declaration for '%s': %v", fn.Get("name").String(), errSet)
 						continue
 					}
-					toolNode = tmp
+					functionToolNode = tmp
 					hasFunction = true
-					hasTool = true
 				}
 			}
 			if gs := t.Get("google_search"); gs.Exists() {
+				googleToolNode := []byte(`{}`)
 				var errSet error
-				toolNode, errSet = sjson.SetRawBytes(toolNode, "googleSearch", []byte(gs.Raw))
+				googleToolNode, errSet = sjson.SetRawBytes(googleToolNode, "googleSearch", []byte(gs.Raw))
 				if errSet != nil {
 					log.Warnf("Failed to set googleSearch tool: %v", errSet)
 					continue
 				}
-				hasTool = true
+				googleSearchNodes = append(googleSearchNodes, googleToolNode)
 			}
 		}
-		if hasTool {
-			out, _ = sjson.SetRawBytes(out, "request.tools", []byte("[]"))
-			out, _ = sjson.SetRawBytes(out, "request.tools.0", toolNode)
+		if hasFunction || len(googleSearchNodes) > 0 {
+			toolsNode := []byte("[]")
+			if hasFunction {
+				toolsNode, _ = sjson.SetRawBytes(toolsNode, "-1", functionToolNode)
+			}
+			for _, googleNode := range googleSearchNodes {
+				toolsNode, _ = sjson.SetRawBytes(toolsNode, "-1", googleNode)
+			}
+			out, _ = sjson.SetRawBytes(out, "request.tools", toolsNode)
 		}
 	}
 

internal/translator/claude/gemini/claude_gemini_request.go

@@ -98,9 +98,8 @@ func ConvertGeminiRequestToClaude(modelName string, inputRawJSON []byte, stream
 		// Temperature setting for controlling response randomness
 		if temp := genConfig.Get("temperature"); temp.Exists() {
 			out, _ = sjson.Set(out, "temperature", temp.Float())
-		}
-		// Top P setting for nucleus sampling
-		if topP := genConfig.Get("topP"); topP.Exists() {
+		} else if topP := genConfig.Get("topP"); topP.Exists() {
+			// Top P setting for nucleus sampling (filtered out if temperature is set)
 			out, _ = sjson.Set(out, "top_p", topP.Float())
 		}
 		// Stop sequences configuration for custom termination conditions

internal/translator/claude/openai/chat-completions/claude_openai_request.go

@@ -110,10 +110,8 @@ func ConvertOpenAIRequestToClaude(modelName string, inputRawJSON []byte, stream
 	// Temperature setting for controlling response randomness
 	if temp := root.Get("temperature"); temp.Exists() {
 		out, _ = sjson.Set(out, "temperature", temp.Float())
-	}
-
-	// Top P setting for nucleus sampling
-	if topP := root.Get("top_p"); topP.Exists() {
+	} else if topP := root.Get("top_p"); topP.Exists() {
+		// Top P setting for nucleus sampling (filtered out if temperature is set)
 		out, _ = sjson.Set(out, "top_p", topP.Float())
 	}
 

internal/translator/gemini-cli/openai/chat-completions/gemini-cli_openai_request.go

@@ -283,12 +283,12 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 		}
 	}
 
-	// tools -> request.tools[0].functionDeclarations + request.tools[0].googleSearch passthrough
+	// tools -> request.tools[].functionDeclarations + request.tools[].googleSearch passthrough
 	tools := gjson.GetBytes(rawJSON, "tools")
 	if tools.IsArray() && len(tools.Array()) > 0 {
-		toolNode := []byte(`{}`)
-		hasTool := false
+		functionToolNode := []byte(`{}`)
 		hasFunction := false
+		googleSearchNodes := make([][]byte, 0)
 		for _, t := range tools.Array() {
 			if t.Get("type").String() == "function" {
 				fn := t.Get("function")
@@ -327,31 +327,37 @@ func ConvertOpenAIRequestToGeminiCLI(modelName string, inputRawJSON []byte, _ bo
 					}
 					fnRaw, _ = sjson.Delete(fnRaw, "strict")
 					if !hasFunction {
-						toolNode, _ = sjson.SetRawBytes(toolNode, "functionDeclarations", []byte("[]"))
+						functionToolNode, _ = sjson.SetRawBytes(functionToolNode, "functionDeclarations", []byte("[]"))
 					}
-					tmp, errSet := sjson.SetRawBytes(toolNode, "functionDeclarations.-1", []byte(fnRaw))
+					tmp, errSet := sjson.SetRawBytes(functionToolNode, "functionDeclarations.-1", []byte(fnRaw))
 					if errSet != nil {
 						log.Warnf("Failed to append tool declaration for '%s': %v", fn.Get("name").String(), errSet)
 						continue
 					}
-					toolNode = tmp
+					functionToolNode = tmp
 					hasFunction = true
-					hasTool = true
 				}
 			}
 			if gs := t.Get("google_search"); gs.Exists() {
+				googleToolNode := []byte(`{}`)
 				var errSet error
-				toolNode, errSet = sjson.SetRawBytes(toolNode, "googleSearch", []byte(gs.Raw))
+				googleToolNode, errSet = sjson.SetRawBytes(googleToolNode, "googleSearch", []byte(gs.Raw))
 				if errSet != nil {
 					log.Warnf("Failed to set googleSearch tool: %v", errSet)
 					continue
 				}
-				hasTool = true
+				googleSearchNodes = append(googleSearchNodes, googleToolNode)
 			}
 		}
-		if hasTool {
-			out, _ = sjson.SetRawBytes(out, "request.tools", []byte("[]"))
-			out, _ = sjson.SetRawBytes(out, "request.tools.0", toolNode)
+		if hasFunction || len(googleSearchNodes) > 0 {
+			toolsNode := []byte("[]")
+			if hasFunction {
+				toolsNode, _ = sjson.SetRawBytes(toolsNode, "-1", functionToolNode)
+			}
+			for _, googleNode := range googleSearchNodes {
+				toolsNode, _ = sjson.SetRawBytes(toolsNode, "-1", googleNode)
+			}
+			out, _ = sjson.SetRawBytes(out, "request.tools", toolsNode)
 		}
 	}
 

internal/translator/gemini/openai/chat-completions/gemini_openai_request.go

@@ -289,12 +289,12 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 		}
 	}
 
-	// tools -> tools[0].functionDeclarations + tools[0].googleSearch passthrough
+	// tools -> tools[].functionDeclarations + tools[].googleSearch passthrough
 	tools := gjson.GetBytes(rawJSON, "tools")
 	if tools.IsArray() && len(tools.Array()) > 0 {
-		toolNode := []byte(`{}`)
-		hasTool := false
+		functionToolNode := []byte(`{}`)
 		hasFunction := false
+		googleSearchNodes := make([][]byte, 0)
 		for _, t := range tools.Array() {
 			if t.Get("type").String() == "function" {
 				fn := t.Get("function")
@@ -333,31 +333,37 @@ func ConvertOpenAIRequestToGemini(modelName string, inputRawJSON []byte, _ bool)
 					}
 					fnRaw, _ = sjson.Delete(fnRaw, "strict")
 					if !hasFunction {
-						toolNode, _ = sjson.SetRawBytes(toolNode, "functionDeclarations", []byte("[]"))
+						functionToolNode, _ = sjson.SetRawBytes(functionToolNode, "functionDeclarations", []byte("[]"))
 					}
-					tmp, errSet := sjson.SetRawBytes(toolNode, "functionDeclarations.-1", []byte(fnRaw))
+					tmp, errSet := sjson.SetRawBytes(functionToolNode, "functionDeclarations.-1", []byte(fnRaw))
 					if errSet != nil {
 						log.Warnf("Failed to append tool declaration for '%s': %v", fn.Get("name").String(), errSet)
 						continue
 					}
-					toolNode = tmp
+					functionToolNode = tmp
 					hasFunction = true
-					hasTool = true
 				}
 			}
 			if gs := t.Get("google_search"); gs.Exists() {
+				googleToolNode := []byte(`{}`)
 				var errSet error
-				toolNode, errSet = sjson.SetRawBytes(toolNode, "googleSearch", []byte(gs.Raw))
+				googleToolNode, errSet = sjson.SetRawBytes(googleToolNode, "googleSearch", []byte(gs.Raw))
 				if errSet != nil {
 					log.Warnf("Failed to set googleSearch tool: %v", errSet)
 					continue
 				}
-				hasTool = true
+				googleSearchNodes = append(googleSearchNodes, googleToolNode)
 			}
 		}
-		if hasTool {
-			out, _ = sjson.SetRawBytes(out, "tools", []byte("[]"))
-			out, _ = sjson.SetRawBytes(out, "tools.0", toolNode)
+		if hasFunction || len(googleSearchNodes) > 0 {
+			toolsNode := []byte("[]")
+			if hasFunction {
+				toolsNode, _ = sjson.SetRawBytes(toolsNode, "-1", functionToolNode)
+			}
+			for _, googleNode := range googleSearchNodes {
+				toolsNode, _ = sjson.SetRawBytes(toolsNode, "-1", googleNode)
+			}
+			out, _ = sjson.SetRawBytes(out, "tools", toolsNode)
 		}
 	}
 

internal/translator/openai/claude/openai_claude_request.go

@@ -89,12 +89,14 @@ func ConvertClaudeRequestToOpenAI(modelName string, inputRawJSON []byte, stream
 
 	// Handle system message first
 	systemMsgJSON := `{"role":"system","content":[]}`
+	hasSystemContent := false
 	if system := root.Get("system"); system.Exists() {
 		if system.Type == gjson.String {
 			if system.String() != "" {
 				oldSystem := `{"type":"text","text":""}`
 				oldSystem, _ = sjson.Set(oldSystem, "text", system.String())
 				systemMsgJSON, _ = sjson.SetRaw(systemMsgJSON, "content.-1", oldSystem)
+				hasSystemContent = true
 			}
 		} else if system.Type == gjson.JSON {
 			if system.IsArray() {
@@ -102,12 +104,16 @@ func ConvertClaudeRequestToOpenAI(modelName string, inputRawJSON []byte, stream
 				for i := 0; i < len(systemResults); i++ {
 					if contentItem, ok := convertClaudeContentPart(systemResults[i]); ok {
 						systemMsgJSON, _ = sjson.SetRaw(systemMsgJSON, "content.-1", contentItem)
+						hasSystemContent = true
 					}
 				}
 			}
 		}
 	}
-	messagesJSON, _ = sjson.SetRaw(messagesJSON, "-1", systemMsgJSON)
+	// Only add system message if it has content
+	if hasSystemContent {
+		messagesJSON, _ = sjson.SetRaw(messagesJSON, "-1", systemMsgJSON)
+	}
 
 	// Process Anthropic messages
 	if messages := root.Get("messages"); messages.Exists() && messages.IsArray() {

sdk/api/handlers/handlers.go

@@ -385,6 +385,7 @@ func (h *BaseAPIHandler) ExecuteWithAuthManager(ctx context.Context, handlerType
 		return nil, errMsg
 	}
 	reqMeta := requestExecutionMetadata(ctx)
+	reqMeta[coreexecutor.RequestedModelMetadataKey] = normalizedModel
 	req := coreexecutor.Request{
 		Model:   normalizedModel,
 		Payload: cloneBytes(rawJSON),
@@ -423,6 +424,7 @@ func (h *BaseAPIHandler) ExecuteCountWithAuthManager(ctx context.Context, handle
 		return nil, errMsg
 	}
 	reqMeta := requestExecutionMetadata(ctx)
+	reqMeta[coreexecutor.RequestedModelMetadataKey] = normalizedModel
 	req := coreexecutor.Request{
 		Model:   normalizedModel,
 		Payload: cloneBytes(rawJSON),
@@ -464,6 +466,7 @@ func (h *BaseAPIHandler) ExecuteStreamWithAuthManager(ctx context.Context, handl
 		return nil, errChan
 	}
 	reqMeta := requestExecutionMetadata(ctx)
+	reqMeta[coreexecutor.RequestedModelMetadataKey] = normalizedModel
 	req := coreexecutor.Request{
 		Model:   normalizedModel,
 		Payload: cloneBytes(rawJSON),

sdk/cliproxy/auth/conductor.go

@@ -570,6 +570,7 @@ func (m *Manager) executeMixedOnce(ctx context.Context, providers []string, req
 		return cliproxyexecutor.Response{}, &Error{Code: "provider_not_found", Message: "no provider supplied"}
 	}
 	routeModel := req.Model
+	opts = ensureRequestedModelMetadata(opts, routeModel)
 	tried := make(map[string]struct{})
 	var lastErr error
 	for {
@@ -597,6 +598,9 @@ func (m *Manager) executeMixedOnce(ctx context.Context, providers []string, req
 		resp, errExec := executor.Execute(execCtx, auth, execReq, opts)
 		result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: errExec == nil}
 		if errExec != nil {
+			if errCtx := execCtx.Err(); errCtx != nil {
+				return cliproxyexecutor.Response{}, errCtx
+			}
 			result.Error = &Error{Message: errExec.Error()}
 			var se cliproxyexecutor.StatusError
 			if errors.As(errExec, &se) && se != nil {
@@ -619,6 +623,7 @@ func (m *Manager) executeCountMixedOnce(ctx context.Context, providers []string,
 		return cliproxyexecutor.Response{}, &Error{Code: "provider_not_found", Message: "no provider supplied"}
 	}
 	routeModel := req.Model
+	opts = ensureRequestedModelMetadata(opts, routeModel)
 	tried := make(map[string]struct{})
 	var lastErr error
 	for {
@@ -646,6 +651,9 @@ func (m *Manager) executeCountMixedOnce(ctx context.Context, providers []string,
 		resp, errExec := executor.CountTokens(execCtx, auth, execReq, opts)
 		result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: errExec == nil}
 		if errExec != nil {
+			if errCtx := execCtx.Err(); errCtx != nil {
+				return cliproxyexecutor.Response{}, errCtx
+			}
 			result.Error = &Error{Message: errExec.Error()}
 			var se cliproxyexecutor.StatusError
 			if errors.As(errExec, &se) && se != nil {
@@ -668,6 +676,7 @@ func (m *Manager) executeStreamMixedOnce(ctx context.Context, providers []string
 		return nil, &Error{Code: "provider_not_found", Message: "no provider supplied"}
 	}
 	routeModel := req.Model
+	opts = ensureRequestedModelMetadata(opts, routeModel)
 	tried := make(map[string]struct{})
 	var lastErr error
 	for {
@@ -694,6 +703,9 @@ func (m *Manager) executeStreamMixedOnce(ctx context.Context, providers []string
 		execReq.Model = m.applyAPIKeyModelAlias(auth, execReq.Model)
 		chunks, errStream := executor.ExecuteStream(execCtx, auth, execReq, opts)
 		if errStream != nil {
+			if errCtx := execCtx.Err(); errCtx != nil {
+				return nil, errCtx
+			}
 			rerr := &Error{Message: errStream.Error()}
 			var se cliproxyexecutor.StatusError
 			if errors.As(errStream, &se) && se != nil {
@@ -729,167 +741,42 @@ func (m *Manager) executeStreamMixedOnce(ctx context.Context, providers []string
 	}
 }
 
-func (m *Manager) executeWithProvider(ctx context.Context, provider string, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
-	if provider == "" {
-		return cliproxyexecutor.Response{}, &Error{Code: "provider_not_found", Message: "provider identifier is empty"}
+func ensureRequestedModelMetadata(opts cliproxyexecutor.Options, requestedModel string) cliproxyexecutor.Options {
+	requestedModel = strings.TrimSpace(requestedModel)
+	if requestedModel == "" {
+		return opts
 	}
-	routeModel := req.Model
-	tried := make(map[string]struct{})
-	var lastErr error
-	for {
-		auth, executor, errPick := m.pickNext(ctx, provider, routeModel, opts, tried)
-		if errPick != nil {
-			if lastErr != nil {
-				return cliproxyexecutor.Response{}, lastErr
-			}
-			return cliproxyexecutor.Response{}, errPick
-		}
-
-		entry := logEntryWithRequestID(ctx)
-		debugLogAuthSelection(entry, auth, provider, req.Model)
-
-		tried[auth.ID] = struct{}{}
-		execCtx := ctx
-		if rt := m.roundTripperFor(auth); rt != nil {
-			execCtx = context.WithValue(execCtx, roundTripperContextKey{}, rt)
-			execCtx = context.WithValue(execCtx, "cliproxy.roundtripper", rt)
-		}
-		execReq := req
-		execReq.Model = rewriteModelForAuth(routeModel, auth)
-		execReq.Model = m.applyOAuthModelAlias(auth, execReq.Model)
-		execReq.Model = m.applyAPIKeyModelAlias(auth, execReq.Model)
-		resp, errExec := executor.Execute(execCtx, auth, execReq, opts)
-		result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: errExec == nil}
-		if errExec != nil {
-			result.Error = &Error{Message: errExec.Error()}
-			var se cliproxyexecutor.StatusError
-			if errors.As(errExec, &se) && se != nil {
-				result.Error.HTTPStatus = se.StatusCode()
-			}
-			if ra := retryAfterFromError(errExec); ra != nil {
-				result.RetryAfter = ra
-			}
-			m.MarkResult(execCtx, result)
-			lastErr = errExec
-			continue
-		}
-		m.MarkResult(execCtx, result)
-		return resp, nil
+	if hasRequestedModelMetadata(opts.Metadata) {
+		return opts
 	}
+	if len(opts.Metadata) == 0 {
+		opts.Metadata = map[string]any{cliproxyexecutor.RequestedModelMetadataKey: requestedModel}
+		return opts
+	}
+	meta := make(map[string]any, len(opts.Metadata)+1)
+	for k, v := range opts.Metadata {
+		meta[k] = v
+	}
+	meta[cliproxyexecutor.RequestedModelMetadataKey] = requestedModel
+	opts.Metadata = meta
+	return opts
 }
 
-func (m *Manager) executeCountWithProvider(ctx context.Context, provider string, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (cliproxyexecutor.Response, error) {
-	if provider == "" {
-		return cliproxyexecutor.Response{}, &Error{Code: "provider_not_found", Message: "provider identifier is empty"}
+func hasRequestedModelMetadata(meta map[string]any) bool {
+	if len(meta) == 0 {
+		return false
 	}
-	routeModel := req.Model
-	tried := make(map[string]struct{})
-	var lastErr error
-	for {
-		auth, executor, errPick := m.pickNext(ctx, provider, routeModel, opts, tried)
-		if errPick != nil {
-			if lastErr != nil {
-				return cliproxyexecutor.Response{}, lastErr
-			}
-			return cliproxyexecutor.Response{}, errPick
-		}
-
-		entry := logEntryWithRequestID(ctx)
-		debugLogAuthSelection(entry, auth, provider, req.Model)
-
-		tried[auth.ID] = struct{}{}
-		execCtx := ctx
-		if rt := m.roundTripperFor(auth); rt != nil {
-			execCtx = context.WithValue(execCtx, roundTripperContextKey{}, rt)
-			execCtx = context.WithValue(execCtx, "cliproxy.roundtripper", rt)
-		}
-		execReq := req
-		execReq.Model = rewriteModelForAuth(routeModel, auth)
-		execReq.Model = m.applyOAuthModelAlias(auth, execReq.Model)
-		execReq.Model = m.applyAPIKeyModelAlias(auth, execReq.Model)
-		resp, errExec := executor.CountTokens(execCtx, auth, execReq, opts)
-		result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: errExec == nil}
-		if errExec != nil {
-			result.Error = &Error{Message: errExec.Error()}
-			var se cliproxyexecutor.StatusError
-			if errors.As(errExec, &se) && se != nil {
-				result.Error.HTTPStatus = se.StatusCode()
-			}
-			if ra := retryAfterFromError(errExec); ra != nil {
-				result.RetryAfter = ra
-			}
-			m.MarkResult(execCtx, result)
-			lastErr = errExec
-			continue
-		}
-		m.MarkResult(execCtx, result)
-		return resp, nil
+	raw, ok := meta[cliproxyexecutor.RequestedModelMetadataKey]
+	if !ok || raw == nil {
+		return false
 	}
-}
-
-func (m *Manager) executeStreamWithProvider(ctx context.Context, provider string, req cliproxyexecutor.Request, opts cliproxyexecutor.Options) (<-chan cliproxyexecutor.StreamChunk, error) {
-	if provider == "" {
-		return nil, &Error{Code: "provider_not_found", Message: "provider identifier is empty"}
-	}
-	routeModel := req.Model
-	tried := make(map[string]struct{})
-	var lastErr error
-	for {
-		auth, executor, errPick := m.pickNext(ctx, provider, routeModel, opts, tried)
-		if errPick != nil {
-			if lastErr != nil {
-				return nil, lastErr
-			}
-			return nil, errPick
-		}
-
-		entry := logEntryWithRequestID(ctx)
-		debugLogAuthSelection(entry, auth, provider, req.Model)
-
-		tried[auth.ID] = struct{}{}
-		execCtx := ctx
-		if rt := m.roundTripperFor(auth); rt != nil {
-			execCtx = context.WithValue(execCtx, roundTripperContextKey{}, rt)
-			execCtx = context.WithValue(execCtx, "cliproxy.roundtripper", rt)
-		}
-		execReq := req
-		execReq.Model = rewriteModelForAuth(routeModel, auth)
-		execReq.Model = m.applyOAuthModelAlias(auth, execReq.Model)
-		execReq.Model = m.applyAPIKeyModelAlias(auth, execReq.Model)
-		chunks, errStream := executor.ExecuteStream(execCtx, auth, execReq, opts)
-		if errStream != nil {
-			rerr := &Error{Message: errStream.Error()}
-			var se cliproxyexecutor.StatusError
-			if errors.As(errStream, &se) && se != nil {
-				rerr.HTTPStatus = se.StatusCode()
-			}
-			result := Result{AuthID: auth.ID, Provider: provider, Model: routeModel, Success: false, Error: rerr}
-			result.RetryAfter = retryAfterFromError(errStream)
-			m.MarkResult(execCtx, result)
-			lastErr = errStream
-			continue
-		}
-		out := make(chan cliproxyexecutor.StreamChunk)
-		go func(streamCtx context.Context, streamAuth *Auth, streamProvider string, streamChunks <-chan cliproxyexecutor.StreamChunk) {
-			defer close(out)
-			var failed bool
-			for chunk := range streamChunks {
-				if chunk.Err != nil && !failed {
-					failed = true
-					rerr := &Error{Message: chunk.Err.Error()}
-					var se cliproxyexecutor.StatusError
-					if errors.As(chunk.Err, &se) && se != nil {
-						rerr.HTTPStatus = se.StatusCode()
-					}
-					m.MarkResult(streamCtx, Result{AuthID: streamAuth.ID, Provider: streamProvider, Model: routeModel, Success: false, Error: rerr})
-				}
-				out <- chunk
-			}
-			if !failed {
-				m.MarkResult(streamCtx, Result{AuthID: streamAuth.ID, Provider: streamProvider, Model: routeModel, Success: true})
-			}
-		}(execCtx, auth.Clone(), provider, chunks)
-		return out, nil
+	switch v := raw.(type) {
+	case string:
+		return strings.TrimSpace(v) != ""
+	case []byte:
+		return strings.TrimSpace(string(v)) != ""
+	default:
+		return false
 	}
 }
 
@@ -1140,35 +1027,6 @@ func (m *Manager) normalizeProviders(providers []string) []string {
 	return result
 }
 
-// rotateProviders returns a rotated view of the providers list starting from the
-// current offset for the model, and atomically increments the offset for the next call.
-// This ensures concurrent requests get different starting providers.
-func (m *Manager) rotateProviders(model string, providers []string) []string {
-	if len(providers) == 0 {
-		return nil
-	}
-
-	// Atomic read-and-increment: get current offset and advance cursor in one lock
-	m.mu.Lock()
-	offset := m.providerOffsets[model]
-	m.providerOffsets[model] = (offset + 1) % len(providers)
-	m.mu.Unlock()
-
-	if len(providers) > 0 {
-		offset %= len(providers)
-	}
-	if offset < 0 {
-		offset = 0
-	}
-	if offset == 0 {
-		return providers
-	}
-	rotated := make([]string, 0, len(providers))
-	rotated = append(rotated, providers[offset:]...)
-	rotated = append(rotated, providers[:offset]...)
-	return rotated
-}
-
 func (m *Manager) retrySettings() (int, time.Duration) {
 	if m == nil {
 		return 0, 0
@@ -1250,42 +1108,6 @@ func waitForCooldown(ctx context.Context, wait time.Duration) error {
 	}
 }
 
-func (m *Manager) executeProvidersOnce(ctx context.Context, providers []string, fn func(context.Context, string) (cliproxyexecutor.Response, error)) (cliproxyexecutor.Response, error) {
-	if len(providers) == 0 {
-		return cliproxyexecutor.Response{}, &Error{Code: "provider_not_found", Message: "no provider supplied"}
-	}
-	var lastErr error
-	for _, provider := range providers {
-		resp, errExec := fn(ctx, provider)
-		if errExec == nil {
-			return resp, nil
-		}
-		lastErr = errExec
-	}
-	if lastErr != nil {
-		return cliproxyexecutor.Response{}, lastErr
-	}
-	return cliproxyexecutor.Response{}, &Error{Code: "auth_not_found", Message: "no auth available"}
-}
-
-func (m *Manager) executeStreamProvidersOnce(ctx context.Context, providers []string, fn func(context.Context, string) (<-chan cliproxyexecutor.StreamChunk, error)) (<-chan cliproxyexecutor.StreamChunk, error) {
-	if len(providers) == 0 {
-		return nil, &Error{Code: "provider_not_found", Message: "no provider supplied"}
-	}
-	var lastErr error
-	for _, provider := range providers {
-		chunks, errExec := fn(ctx, provider)
-		if errExec == nil {
-			return chunks, nil
-		}
-		lastErr = errExec
-	}
-	if lastErr != nil {
-		return nil, lastErr
-	}
-	return nil, &Error{Code: "auth_not_found", Message: "no auth available"}
-}
-
 // MarkResult records an execution result and notifies hooks.
 func (m *Manager) MarkResult(ctx context.Context, result Result) {
 	if result.AuthID == "" {
@@ -1371,8 +1193,12 @@ func (m *Manager) MarkResult(ctx context.Context, result Result) {
 					shouldSuspendModel = true
 					setModelQuota = true
 				case 408, 500, 502, 503, 504:
-					next := now.Add(1 * time.Minute)
-					state.NextRetryAfter = next
+					if quotaCooldownDisabled.Load() {
+						state.NextRetryAfter = time.Time{}
+					} else {
+						next := now.Add(1 * time.Minute)
+						state.NextRetryAfter = next
+					}
 				default:
 					state.NextRetryAfter = time.Time{}
 				}
@@ -1623,7 +1449,11 @@ func applyAuthFailureState(auth *Auth, resultErr *Error, retryAfter *time.Durati
 		auth.NextRetryAfter = next
 	case 408, 500, 502, 503, 504:
 		auth.StatusMessage = "transient upstream error"
-		auth.NextRetryAfter = now.Add(1 * time.Minute)
+		if quotaCooldownDisabled.Load() {
+			auth.NextRetryAfter = time.Time{}
+		} else {
+			auth.NextRetryAfter = now.Add(1 * time.Minute)
+		}
 	default:
 		if auth.StatusMessage == "" {
 			auth.StatusMessage = "request failed"

sdk/cliproxy/executor/types.go

@@ -7,6 +7,9 @@ import (
 	sdktranslator "github.com/router-for-me/CLIProxyAPI/v6/sdk/translator"
 )
 
+// RequestedModelMetadataKey stores the client-requested model name in Options.Metadata.
+const RequestedModelMetadataKey = "requested_model"
+
 // Request encapsulates the translated payload that will be sent to a provider executor.
 type Request struct {
 	// Model is the upstream model identifier after translation.