ci(homebrew): add workflow to auto-bump Homebrew formula on release

feat(cliproxy): Rebind auth executors on config change

.github/workflows/homebrew.yml

@@ -0,0 +1,16 @@
+name: Bump Homebrew formula
+
+on:
+  release:
+    types: [ released ]
+
+jobs:
+  homebrew:
+    name: Bump Homebrew formula
+    runs-on: ubuntu-latest
+    steps:
+      - name: Update Homebrew Formula
+        uses: dawidd6/action-homebrew-bump-formula@v5
+        with:
+          token: ${{ secrets.HOMEBREW_TOKEN }}
+          formula: cliproxyapi

README.md

@@ -328,7 +328,7 @@ The server uses a YAML configuration file (`config.yaml`) located in the project
 | `openai-compatibility.*.models.*.alias`            | string   | ""                 | The alias used in the API.                                                                                                                                                                |
 | `gemini-web`                            | object   | {}                 | Configuration specific to the Gemini Web client.                                                                                                                                          |
 | `gemini-web.context`                    | boolean  | true               | Enables conversation context reuse for continuous dialogue.                                                                                                                               |
-| `gemini-web.code-mode`                  | boolean  | false              | Enables code mode for optimized responses in coding-related tasks.                                                                                                                        |
+| `gemini-web.gem-mode`                   | string   | ""                | Selects a predefined Gem to attach for Gemini Web requests; allowed values: `coding-partner`, `writing-editor`. When empty, no Gem is attached.                                           |
 | `gemini-web.max-chars-per-request`      | integer  | 1,000,000          | The maximum number of characters to send to Gemini Web in a single request.                                                                                                               |
 | `gemini-web.disable-continuation-hint`  | boolean  | false              | Disables the continuation hint for split prompts.                                                                                                                                         |
 
@@ -378,7 +378,7 @@ quota-exceeded:
 # Gemini Web client configuration
 gemini-web:
   context: true # Enable conversation context reuse
-  code-mode: false # Enable code mode
+  gem-mode: "" # Select Gem: "coding-partner" or "writing-editor"; empty means no Gem
   max-chars-per-request: 1000000 # Max characters per request
 
 # API keys for official Generative Language API
@@ -720,6 +720,17 @@ Contributions are welcome! Please feel free to submit a Pull Request.
 4. Push to the branch (`git push origin feature/amazing-feature`)
 5. Open a Pull Request
 
+## Who is with us?
+
+Those projects are based on CLIProxyAPI:
+
+### [vibeproxy](https://github.com/automazeio/vibeproxy)
+
+Native macOS menu bar app to use your Claude Code & ChatGPT subscriptions with AI coding tools - no API keys needed
+
+> [!NOTE]  
+> If you developed a project based on CLIProxyAPI, please open a PR to add it to this list.
+
 ## License
 
 This project is licensed under the MIT License - see the [LICENSE](LICENSE) file for details.

README_CN.md

@@ -340,7 +340,7 @@ console.log(await claudeResponse.json());
 | `openai-compatibility.*.models.*.alias`               | string   | ""                 | 在API中使用的别名。                                                         |
 | `gemini-web`                            | object   | {}                 | Gemini Web 客户端的特定配置。                                                 |
 | `gemini-web.context`                    | boolean  | true               | 是否启用会话上下文重用，以实现连续对话。                                        |
-| `gemini-web.code-mode`                  | boolean  | false              | 是否启用代码模式，优化代码相关任务的响应。                                      |
+| `gemini-web.gem-mode`                   | string   | ""                | 选择要附加的预设 Gem（`coding-partner` 或 `writing-editor`）；为空表示不附加。 |
 | `gemini-web.max-chars-per-request`      | integer  | 1,000,000          | 单次请求发送给 Gemini Web 的最大字符数。                                        |
 | `gemini-web.disable-continuation-hint`  | boolean  | false              | 当提示被拆分时，是否禁用连续提示的暗示。                                        |
 
@@ -390,7 +390,7 @@ quota-exceeded:
 # Gemini Web 客户端配置
 gemini-web:
   context: true # 启用会话上下文重用
-  code-mode: false # 启用代码模式
+  gem-mode: "" # 选择 Gem："coding-partner" 或 "writing-editor"；为空表示不附加
   max-chars-per-request: 1000000 # 单次请求最大字符数
 
 # AIStduio Gemini API 的 API 密钥
@@ -729,6 +729,18 @@ docker run --rm -p 8317:8317 -v /path/to/your/config.yaml:/CLIProxyAPI/config.ya
 4. 推送到分支（`git push origin feature/amazing-feature`）
 5. 打开 Pull Request
 
+## 谁与我们在一起？
+
+这些项目基于 CLIProxyAPI:
+
+### [vibeproxy](https://github.com/automazeio/vibeproxy)
+
+一个原生 macOS 菜单栏应用，让您可以使用 Claude Code & ChatGPT 订阅服务和 AI 编程工具，无需 API 密钥。
+
+> [!NOTE]  
+> 如果你开发了基于 CLIProxyAPI 的项目，请提交一个 PR（拉取请求）将其添加到此列表中。
+
+
 ## 许可证
 
 此项目根据 MIT 许可证授权 - 有关详细信息，请参阅 [LICENSE](LICENSE) 文件。

config.example.yaml

@@ -90,10 +90,8 @@ quota-exceeded:
 #    # Disable the short continuation hint appended to intermediate chunks
 #    # when splitting long prompts. Default is false (hint enabled by default).
 #    disable-continuation-hint: false
-#    # Code mode:
-#    #   - true: enable XML wrapping hint and attach the coding-partner Gem.
-#    #           Thought merging (<think> into visible content) applies to STREAMING only;
-#    #           non-stream responses keep reasoning/thought parts separate for clients
-#    #           that expect explicit reasoning fields.
-#    #   - false: disable XML hint and keep <think> separate
-#    code-mode: false
+#    # Gem selection (Gem Mode):
+#    #   - "coding-partner": attach the predefined Coding partner Gem
+#    #   - "writing-editor": attach the predefined Writing editor Gem
+#    #   - empty: do not attach any Gem
+#    gem-mode: ""

internal/api/handlers/management/auth_files.go

@@ -255,7 +255,9 @@ func (h *Handler) ListAuthFiles(c *gin.Context) {
 			full := filepath.Join(h.cfg.AuthDir, name)
 			if data, errRead := os.ReadFile(full); errRead == nil {
 				typeValue := gjson.GetBytes(data, "type").String()
+				emailValue := gjson.GetBytes(data, "email").String()
 				fileData["type"] = typeValue
+				fileData["email"] = emailValue
 			}
 
 			files = append(files, fileData)

internal/config/config.go

@@ -62,10 +62,19 @@ type GeminiWebConfig struct {
 	// Defaults to true if not set in YAML (see LoadConfig).
 	Context bool `yaml:"context" json:"context"`
 
-	// CodeMode, when true, enables coding mode behaviors for Gemini Web:
-	// - Attach the predefined "Coding partner" Gem
-	// - Enable XML wrapping hint for tool markup
-	// - Merge <think> content into visible content for tool-friendly output
+	// GemMode selects a predefined Gem to attach for Gemini Web requests.
+	// Allowed values:
+	// - "coding-partner"
+	// - "writing-editor"
+	// When empty, no Gem is attached by configuration.
+	// This is independent from CodeMode below, which is kept for backwards compatibility.
+	GemMode string `yaml:"gem-mode" json:"gem-mode"`
+
+	// CodeMode enables legacy coding-mode behaviors for Gemini Web.
+	// Backwards compatibility: when true, the service behaves as before by
+	// attaching the predefined "Coding partner" Gem and enabling extra
+	// conveniences (e.g., XML wrapping hints). Prefer GemMode for selecting
+	// a Gem going forward.
 	CodeMode bool `yaml:"code-mode" json:"code-mode"`
 
 	// MaxCharsPerRequest caps the number of characters (runes) sent to

internal/provider/gemini-web/client.go

@@ -225,7 +225,7 @@ func MaskToken28(s string) string {
 }
 
 var NanoBananaModel = map[string]struct{}{
-	"gemini-2.5-flash-image-preview": {},
+	"gemini-2.5-flash-image-web": {},
 }
 
 // NewGeminiClient creates a client. Pass empty strings to auto-detect via browser cookies (not implemented in Go port).
@@ -380,6 +380,15 @@ func (c *GeminiClient) generateOnce(prompt string, files []string, model Model,
 	}
 
 	inner := []any{item0, nil, item2}
+	// Attach Gem first to keep index alignment with reference implementation
+	// so the Gemini Web UI can recognize the selected Gem.
+	if gem != nil {
+		// pad with 16 nils then gem ID
+		for i := 0; i < 16; i++ {
+			inner = append(inner, nil)
+		}
+		inner = append(inner, gem.ID)
+	}
 	requestedModel := strings.ToLower(model.Name)
 	if chat != nil && chat.RequestedModel() != "" {
 		requestedModel = chat.RequestedModel()
@@ -388,13 +397,6 @@ func (c *GeminiClient) generateOnce(prompt string, files []string, model Model,
 		inner = ensureAnyLen(inner, 49)
 		inner[49] = 14
 	}
-	if gem != nil {
-		// pad with 16 nils then gem ID
-		for i := 0; i < 16; i++ {
-			inner = append(inner, nil)
-		}
-		inner = append(inner, gem.ID)
-	}
 	innerJSON, _ := json.Marshal(inner)
 	outer := []any{nil, string(innerJSON)}
 	outerJSON, _ := json.Marshal(outer)

internal/provider/gemini-web/conversation/alias.go

@@ -21,7 +21,7 @@ func EnsureGeminiWebAliasMap() {
 				continue
 			}
 			if m.ID == "gemini-2.5-flash" {
-				aliasMap["gemini-2.5-flash-image-preview"] = "gemini-2.5-flash"
+				aliasMap["gemini-2.5-flash-image-web"] = "gemini-2.5-flash"
 			}
 			alias := AliasFromModelID(m.ID)
 			aliasMap[strings.ToLower(alias)] = strings.ToLower(m.ID)
@@ -65,8 +65,8 @@ func GetGeminiWebAliasedModels() []*registry.ModelInfo {
 			continue
 		} else if m.ID == "gemini-2.5-flash" {
 			cpy := *m
-			cpy.ID = "gemini-2.5-flash-image-preview"
-			cpy.Name = "gemini-2.5-flash-image-preview"
+			cpy.ID = "gemini-2.5-flash-image-web"
+			cpy.Name = "gemini-2.5-flash-image-web"
 			cpy.DisplayName = "Nano Banana"
 			cpy.Description = "Gemini 2.5 Flash Preview Image"
 			aliased = append(aliased, &cpy)

internal/provider/gemini-web/state.go

@@ -460,10 +460,10 @@ func (s *GeminiWebState) Send(ctx context.Context, modelName string, reqPayload
 		return nil, s.wrapSendError(err), nil
 	}
 
-	// Hook: For gemini-2.5-flash-image-preview, if the API returns only images without any text,
+	// Hook: For gemini-2.5-flash-image-web, if the API returns only images without any text,
 	// inject a small textual summary so that conversation persistence has non-empty assistant text.
 	// This helps conversation recovery (conv store) to match sessions reliably.
-	if strings.EqualFold(modelName, "gemini-2.5-flash-image-preview") {
+	if strings.EqualFold(modelName, "gemini-2.5-flash-image-web") {
 		if len(output.Candidates) > 0 {
 			c := output.Candidates[output.Chosen]
 			hasNoText := strings.TrimSpace(c.Text) == ""
@@ -696,7 +696,22 @@ func (s *GeminiWebState) findReusableSession(modelName string, msgs []RoleText)
 }
 
 func (s *GeminiWebState) getConfiguredGem() *Gem {
-	if s.cfg != nil && s.cfg.GeminiWeb.CodeMode {
+	if s.cfg == nil {
+		return nil
+	}
+	// New behavior: attach Gem based on explicit GemMode selection.
+	// Only attaches the Gem; does not toggle any other behavior.
+	if gm := strings.ToLower(strings.TrimSpace(s.cfg.GeminiWeb.GemMode)); gm != "" {
+		switch gm {
+		case "coding-partner":
+			return &Gem{ID: "coding-partner", Name: "Coding partner", Predefined: true}
+		case "writing-editor":
+			return &Gem{ID: "writing-editor", Name: "Writing editor", Predefined: true}
+		}
+	}
+	// Backwards compatibility: legacy CodeMode still attaches Coding partner
+	// and may enable extra behaviors elsewhere.
+	if s.cfg.GeminiWeb.CodeMode {
 		return &Gem{ID: "coding-partner", Name: "Coding partner", Predefined: true}
 	}
 	return nil
@@ -1015,3 +1030,10 @@ func FindReusableSessionIn(items map[string]ConversationRecord, index map[string
 	}
 	return ConversationRecord{}, nil, 0, false
 }
+
+// SetConfig updates the configuration reference used by the state.
+// This allows hot-reload of configuration to take effect for existing
+// runtime states that were cached on auth during previous requests.
+func (s *GeminiWebState) SetConfig(cfg *config.Config) {
+	s.cfg = cfg
+}

internal/runtime/executor/gemini_cli_executor.go

@@ -89,6 +89,7 @@ func (e *GeminiCLIExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth
 			payload = setJSONField(payload, "project", projectID)
 			payload = setJSONField(payload, "model", attemptModel)
 		}
+		payload = disableGeminiThinkingConfig(payload, attemptModel)
 
 		tok, errTok := tokenSource.Token()
 		if errTok != nil {
@@ -165,6 +166,7 @@ func (e *GeminiCLIExecutor) ExecuteStream(ctx context.Context, auth *cliproxyaut
 		payload := append([]byte(nil), basePayload...)
 		payload = setJSONField(payload, "project", projectID)
 		payload = setJSONField(payload, "model", attemptModel)
+		payload = disableGeminiThinkingConfig(payload, attemptModel)
 
 		tok, errTok := tokenSource.Token()
 		if errTok != nil {
@@ -291,6 +293,7 @@ func (e *GeminiCLIExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.
 		payload := sdktranslator.TranslateRequest(from, to, attemptModel, bytes.Clone(req.Payload), false)
 		payload = deleteJSONField(payload, "project")
 		payload = deleteJSONField(payload, "model")
+		payload = disableGeminiThinkingConfig(payload, attemptModel)
 
 		tok, errTok := tokenSource.Token()
 		if errTok != nil {
@@ -500,6 +503,29 @@ func cliPreviewFallbackOrder(model string) []string {
 	}
 }
 
+func disableGeminiThinkingConfig(body []byte, model string) []byte {
+	if !geminiModelDisallowsThinking(model) {
+		return body
+	}
+
+	updated := deleteJSONField(body, "request.generationConfig.thinkingConfig")
+	updated = deleteJSONField(updated, "generationConfig.thinkingConfig")
+	return updated
+}
+
+func geminiModelDisallowsThinking(model string) bool {
+	if model == "" {
+		return false
+	}
+	lower := strings.ToLower(model)
+	for _, marker := range []string{"gemini-2.5-flash-image-preview"} {
+		if strings.Contains(lower, marker) {
+			return true
+		}
+	}
+	return false
+}
+
 // setJSONField sets a top-level JSON field on a byte slice payload via sjson.
 func setJSONField(body []byte, key, value string) []byte {
 	if key == "" {

internal/runtime/executor/gemini_executor.go

@@ -77,6 +77,7 @@ func (e *GeminiExecutor) Execute(ctx context.Context, auth *cliproxyauth.Auth, r
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	body = disableGeminiThinkingConfig(body, req.Model)
 
 	action := "generateContent"
 	if req.Metadata != nil {
@@ -134,6 +135,7 @@ func (e *GeminiExecutor) ExecuteStream(ctx context.Context, auth *cliproxyauth.A
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	body := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), true)
+	body = disableGeminiThinkingConfig(body, req.Model)
 
 	url := fmt.Sprintf("%s/%s/models/%s:%s", glEndpoint, glAPIVersion, req.Model, "streamGenerateContent")
 	if opts.Alt == "" {
@@ -204,6 +206,7 @@ func (e *GeminiExecutor) CountTokens(ctx context.Context, auth *cliproxyauth.Aut
 	from := opts.SourceFormat
 	to := sdktranslator.FromString("gemini")
 	translatedReq := sdktranslator.TranslateRequest(from, to, req.Model, bytes.Clone(req.Payload), false)
+	translatedReq = disableGeminiThinkingConfig(translatedReq, req.Model)
 	respCtx := context.WithValue(ctx, "alt", opts.Alt)
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "tools")
 	translatedReq, _ = sjson.DeleteBytes(translatedReq, "generationConfig")

internal/runtime/executor/gemini_web_executor.go

@@ -168,6 +168,8 @@ func (e *GeminiWebExecutor) stateFor(auth *cliproxyauth.Auth) (*geminiwebapi.Gem
 		return nil, fmt.Errorf("gemini-web executor: auth is nil")
 	}
 	if runtime, ok := auth.Runtime.(*geminiWebRuntime); ok && runtime != nil && runtime.state != nil {
+		// Hot-reload: ensure cached state sees the latest config
+		runtime.state.SetConfig(e.cfg)
 		return runtime.state, nil
 	}
 
@@ -175,6 +177,8 @@ func (e *GeminiWebExecutor) stateFor(auth *cliproxyauth.Auth) (*geminiwebapi.Gem
 	defer e.mu.Unlock()
 
 	if runtime, ok := auth.Runtime.(*geminiWebRuntime); ok && runtime != nil && runtime.state != nil {
+		// Hot-reload: ensure cached state sees the latest config
+		runtime.state.SetConfig(e.cfg)
 		return runtime.state, nil
 	}
 

internal/translator/codex/openai/responses/codex_openai-responses_request.go

@@ -34,9 +34,17 @@ func ConvertOpenAIResponsesRequestToCodex(modelName string, inputRawJSON []byte,
 	}
 
 	inputResult := gjson.GetBytes(rawJSON, "input")
-	inputResults := []gjson.Result{}
-	if inputResult.Exists() && inputResult.IsArray() {
-		inputResults = inputResult.Array()
+	var inputResults []gjson.Result
+	if inputResult.Exists() {
+		if inputResult.IsArray() {
+			inputResults = inputResult.Array()
+		} else if inputResult.Type == gjson.String {
+			newInput := `[{"type":"message","role":"user","content":[{"type":"input_text","text":""}]}]`
+			newInput, _ = sjson.Set(newInput, "0.content.0.text", inputResult.String())
+			inputResults = gjson.Parse(newInput).Array()
+		}
+	} else {
+		inputResults = []gjson.Result{}
 	}
 
 	extractedSystemInstructions := false

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

@@ -8,6 +8,7 @@ package chat_completions
 import (
 	"bytes"
 	"context"
+	"encoding/json"
 	"fmt"
 	"time"
 
@@ -100,6 +101,10 @@ func ConvertCliResponseToOpenAI(_ context.Context, _ string, originalRequestRawJ
 			partResult := partResults[i]
 			partTextResult := partResult.Get("text")
 			functionCallResult := partResult.Get("functionCall")
+			inlineDataResult := partResult.Get("inlineData")
+			if !inlineDataResult.Exists() {
+				inlineDataResult = partResult.Get("inline_data")
+			}
 
 			if partTextResult.Exists() {
 				// Handle text content, distinguishing between regular content and reasoning/thoughts.
@@ -125,6 +130,34 @@ func ConvertCliResponseToOpenAI(_ context.Context, _ string, originalRequestRawJ
 				}
 				template, _ = sjson.Set(template, "choices.0.delta.role", "assistant")
 				template, _ = sjson.SetRaw(template, "choices.0.delta.tool_calls.-1", functionCallTemplate)
+			} else if inlineDataResult.Exists() {
+				data := inlineDataResult.Get("data").String()
+				if data == "" {
+					continue
+				}
+				mimeType := inlineDataResult.Get("mimeType").String()
+				if mimeType == "" {
+					mimeType = inlineDataResult.Get("mime_type").String()
+				}
+				if mimeType == "" {
+					mimeType = "image/png"
+				}
+				imageURL := fmt.Sprintf("data:%s;base64,%s", mimeType, data)
+				imagePayload, err := json.Marshal(map[string]any{
+					"type": "image_url",
+					"image_url": map[string]string{
+						"url": imageURL,
+					},
+				})
+				if err != nil {
+					continue
+				}
+				imagesResult := gjson.Get(template, "choices.0.delta.images")
+				if !imagesResult.Exists() || !imagesResult.IsArray() {
+					template, _ = sjson.SetRaw(template, "choices.0.delta.images", `[]`)
+				}
+				template, _ = sjson.Set(template, "choices.0.delta.role", "assistant")
+				template, _ = sjson.SetRaw(template, "choices.0.delta.images.-1", string(imagePayload))
 			}
 		}
 	}

internal/watcher/watcher.go

@@ -521,6 +521,9 @@ func (w *Watcher) reloadConfig() bool {
 		if oldConfig.GeminiWeb.DisableContinuationHint != newConfig.GeminiWeb.DisableContinuationHint {
 			log.Debugf("  gemini-web.disable-continuation-hint: %t -> %t", oldConfig.GeminiWeb.DisableContinuationHint, newConfig.GeminiWeb.DisableContinuationHint)
 		}
+		if oldConfig.GeminiWeb.GemMode != newConfig.GeminiWeb.GemMode {
+			log.Debugf("  gemini-web.gem-mode: %s -> %s", oldConfig.GeminiWeb.GemMode, newConfig.GeminiWeb.GemMode)
+		}
 		if oldConfig.GeminiWeb.CodeMode != newConfig.GeminiWeb.CodeMode {
 			log.Debugf("  gemini-web.code-mode: %t -> %t", oldConfig.GeminiWeb.CodeMode, newConfig.GeminiWeb.CodeMode)
 		}

sdk/cliproxy/service.go

@@ -291,6 +291,17 @@ func (s *Service) ensureExecutorsForAuth(a *coreauth.Auth) {
 	}
 }
 
+// rebindExecutors refreshes provider executors so they observe the latest configuration.
+func (s *Service) rebindExecutors() {
+	if s == nil || s.coreManager == nil {
+		return
+	}
+	auths := s.coreManager.List()
+	for _, auth := range auths {
+		s.ensureExecutorsForAuth(auth)
+	}
+}
+
 // Run starts the service and blocks until the context is cancelled or the server stops.
 // It initializes all components including authentication, file watching, HTTP server,
 // and starts processing requests. The method blocks until the context is cancelled.
@@ -389,6 +400,7 @@ func (s *Service) Run(ctx context.Context) error {
 		s.cfgMu.Lock()
 		s.cfg = newCfg
 		s.cfgMu.Unlock()
+		s.rebindExecutors()
 
 	}