feat: add OpenAI-compatible API channel

pkg/agent/context.go

@@ -497,6 +497,7 @@ func (cb *ContextBuilder) BuildMessages(
 	currentMessage string,
 	media []string,
 	channel, chatID, senderID, senderDisplayName string,
+	extraSystemPrompt string,
 ) []providers.Message {
 	messages := []providers.Message{}
 
@@ -539,6 +540,15 @@ func (cb *ContextBuilder) BuildMessages(
 		contentBlocks = append(contentBlocks, providers.ContentBlock{Type: "text", Text: summaryText})
 	}
 
+	if strings.TrimSpace(extraSystemPrompt) != "" {
+		extraText := fmt.Sprintf(
+			"CLIENT_SYSTEM_INSTRUCTIONS: The following instructions were provided by the external client for this request only.\n\n%s",
+			extraSystemPrompt,
+		)
+		stringParts = append(stringParts, extraText)
+		contentBlocks = append(contentBlocks, providers.ContentBlock{Type: "text", Text: extraText})
+	}
+
 	fullSystemPrompt := strings.Join(stringParts, "\n\n---\n\n")
 
 	// Log system prompt summary for debugging (debug mode only).

pkg/agent/context_cache_test.go

@@ -82,7 +82,7 @@ func TestSingleSystemMessage(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			msgs := cb.BuildMessages(tt.history, tt.summary, tt.message, nil, "test", "chat1", "", "")
+			msgs := cb.BuildMessages(tt.history, tt.summary, tt.message, nil, "test", "chat1", "", "", "")
 
 			systemCount := 0
 			for _, m := range msgs {
@@ -168,7 +168,7 @@ func TestBuildMessages_CurrentSenderDynamicContext(t *testing.T) {
 
 	for _, tt := range tests {
 		t.Run(tt.name, func(t *testing.T) {
-			msgs := cb.BuildMessages(nil, "", "hello", nil, "discord", "chat1", tt.senderID, tt.senderDisplayName)
+			msgs := cb.BuildMessages(nil, "", "hello", nil, "discord", "chat1", tt.senderID, tt.senderDisplayName, "")
 			sys := msgs[0].Content
 
 			if tt.wantSection {
@@ -638,7 +638,7 @@ func TestConcurrentBuildSystemPromptWithCache(t *testing.T) {
 				}
 
 				// Also exercise BuildMessages concurrently
-				msgs := cb.BuildMessages(nil, "", "hello", nil, "test", "chat", "", "")
+				msgs := cb.BuildMessages(nil, "", "hello", nil, "test", "chat", "", "", "")
 				if len(msgs) < 2 {
 					errs <- "BuildMessages returned fewer than 2 messages"
 					return
@@ -726,6 +726,6 @@ func BenchmarkBuildMessagesWithCache(b *testing.B) {
 
 	b.ResetTimer()
 	for i := 0; i < b.N; i++ {
-		_ = cb.BuildMessages(history, "summary", "new message", nil, "cli", "test", "", "")
+		_ = cb.BuildMessages(history, "summary", "new message", nil, "cli", "test", "", "", "")
 	}
 }

pkg/agent/loop.go

@@ -62,20 +62,27 @@ type processOptions struct {
 	SenderDisplayName string   // Current sender display name for dynamic context
 	UserMessage       string   // User message content (may include prefix)
 	Media             []string // media:// refs from inbound message
-	DefaultResponse   string   // Response when LLM returns empty
+	InjectedHistory   []providers.Message
-	EnableSummary     bool     // Whether to trigger summarization
+	ExtraSystemPrompt string
-	SendResponse      bool     // Whether to send response via bus
+	RequestedModel    string
-	NoHistory         bool     // If true, don't load session history (for heartbeat)
+	DefaultResponse   string // Response when LLM returns empty
+	EnableSummary     bool   // Whether to trigger summarization
+	SendResponse      bool   // Whether to send response via bus
+	NoHistory         bool   // If true, don't load or persist session history
 }
 
 const (
-	defaultResponse           = "I've completed processing but have no response to give. Increase `max_tool_iterations` in config.json."
+	defaultResponse              = "I've completed processing but have no response to give. Increase `max_tool_iterations` in config.json."
-	sessionKeyAgentPrefix     = "agent:"
+	sessionKeyAgentPrefix        = "agent:"
-	metadataKeyAccountID      = "account_id"
+	metadataKeyAccountID         = "account_id"
-	metadataKeyGuildID        = "guild_id"
+	metadataKeyGuildID           = "guild_id"
-	metadataKeyTeamID         = "team_id"
+	metadataKeyTeamID            = "team_id"
-	metadataKeyParentPeerKind = "parent_peer_kind"
+	metadataKeyParentPeerKind    = "parent_peer_kind"
-	metadataKeyParentPeerID   = "parent_peer_id"
+	metadataKeyParentPeerID      = "parent_peer_id"
+	metadataKeyNoHistory         = "no_history"
+	metadataKeyRequestedModel    = "requested_model"
+	metadataKeyInjectedHistory   = "injected_history"
+	metadataKeyExtraSystemPrompt = "extra_system_prompt"
 )
 
 func NewAgentLoop(
@@ -758,6 +765,14 @@ func (al *AgentLoop) processMessage(ctx context.Context, msg bus.InboundMessage)
 		SendResponse:      false,
 	}
 
+	if metadataEnabled(msg, metadataKeyNoHistory) {
+		opts.NoHistory = true
+		opts.EnableSummary = false
+	}
+	opts.RequestedModel = inboundMetadata(msg, metadataKeyRequestedModel)
+	opts.ExtraSystemPrompt = inboundMetadata(msg, metadataKeyExtraSystemPrompt)
+	opts.InjectedHistory = decodeInjectedHistory(msg)
+
 	// context-dependent commands check their own Runtime fields and report
 	// "unavailable" when the required capability is nil.
 	if response, handled := al.handleCommand(ctx, msg, agent, &opts); handled {
@@ -884,7 +899,9 @@ func (al *AgentLoop) runAgentLoop(
 	// 1. Build messages (skip history for heartbeat)
 	var history []providers.Message
 	var summary string
-	if !opts.NoHistory {
+	if len(opts.InjectedHistory) > 0 {
+		history = append([]providers.Message(nil), opts.InjectedHistory...)
+	} else if !opts.NoHistory {
 		history = agent.Sessions.GetHistory(opts.SessionKey)
 		summary = agent.Sessions.GetSummary(opts.SessionKey)
 	}
@@ -897,6 +914,7 @@ func (al *AgentLoop) runAgentLoop(
 		opts.ChatID,
 		opts.SenderID,
 		opts.SenderDisplayName,
+		opts.ExtraSystemPrompt,
 	)
 
 	// Resolve media:// refs: images→base64 data URLs, non-images→local paths in content
@@ -905,7 +923,9 @@ func (al *AgentLoop) runAgentLoop(
 	messages = resolveMediaRefs(messages, al.mediaStore, maxMediaSize)
 
 	// 2. Save user message to session
-	agent.Sessions.AddMessage(opts.SessionKey, "user", opts.UserMessage)
+	if !opts.NoHistory {
+		agent.Sessions.AddMessage(opts.SessionKey, "user", opts.UserMessage)
+	}
 
 	// 3. Run LLM iteration loop
 	finalContent, iteration, err := al.runLLMIteration(ctx, agent, messages, opts)
@@ -922,11 +942,13 @@ func (al *AgentLoop) runAgentLoop(
 	}
 
 	// 5. Save final assistant message to session
-	agent.Sessions.AddMessage(opts.SessionKey, "assistant", finalContent)
+	if !opts.NoHistory {
-	agent.Sessions.Save(opts.SessionKey)
+		agent.Sessions.AddMessage(opts.SessionKey, "assistant", finalContent)
+		agent.Sessions.Save(opts.SessionKey)
+	}
 
 	// 6. Optional: summarization
-	if opts.EnableSummary {
+	if opts.EnableSummary && !opts.NoHistory {
 		al.maybeSummarize(agent, opts.SessionKey, opts.Channel, opts.ChatID)
 	}
 
@@ -1022,7 +1044,10 @@ func (al *AgentLoop) runLLMIteration(
 	// selectCandidates evaluates routing once and the decision is sticky for
 	// all tool-follow-up iterations within the same turn so that a multi-step
 	// tool chain doesn't switch models mid-way through.
-	activeCandidates, activeModel := al.selectCandidates(agent, opts.UserMessage, messages)
+	activeCandidates, activeModel, err := al.selectCandidates(agent, opts.RequestedModel, opts.UserMessage, messages)
+	if err != nil {
+		return "", iteration, err
+	}
 
 	for iteration < agent.MaxIterations {
 		iteration++
@@ -1162,7 +1187,7 @@ func (al *AgentLoop) runLLMIteration(
 				continue
 			}
 
-			if isContextError && retry < maxRetries {
+			if isContextError && retry < maxRetries && !opts.NoHistory {
 				logger.WarnCF(
 					"agent",
 					"Context window error detected, attempting compression",
@@ -1185,7 +1210,7 @@ func (al *AgentLoop) runLLMIteration(
 				newSummary := agent.Sessions.GetSummary(opts.SessionKey)
 				messages = agent.ContextBuilder.BuildMessages(
 					newHistory, newSummary, "",
-					nil, opts.Channel, opts.ChatID, opts.SenderID, opts.SenderDisplayName,
+					nil, opts.Channel, opts.ChatID, opts.SenderID, opts.SenderDisplayName, opts.ExtraSystemPrompt,
 				)
 				continue
 			}
@@ -1446,11 +1471,16 @@ func (al *AgentLoop) runLLMIteration(
 // that a multi-step tool chain doesn't switch models mid-way.
 func (al *AgentLoop) selectCandidates(
 	agent *AgentInstance,
+	requestedModel string,
 	userMsg string,
 	history []providers.Message,
-) (candidates []providers.FallbackCandidate, model string) {
+) (candidates []providers.FallbackCandidate, model string, err error) {
+	if strings.TrimSpace(requestedModel) != "" {
+		return al.resolveRequestedModelCandidates(agent, requestedModel)
+	}
+
 	if agent.Router == nil || len(agent.LightCandidates) == 0 {
-		return agent.Candidates, agent.Model
+		return agent.Candidates, agent.Model, nil
 	}
 
 	_, usedLight, score := agent.Router.SelectModel(userMsg, history, agent.Model)
@@ -1461,7 +1491,7 @@ func (al *AgentLoop) selectCandidates(
 				"score":     score,
 				"threshold": agent.Router.Threshold(),
 			})
-		return agent.Candidates, agent.Model
+		return agent.Candidates, agent.Model, nil
 	}
 
 	logger.InfoCF("agent", "Model routing: light model selected",
@@ -1471,7 +1501,66 @@ func (al *AgentLoop) selectCandidates(
 			"score":       score,
 			"threshold":   agent.Router.Threshold(),
 		})
-	return agent.LightCandidates, agent.Router.LightModel()
+	return agent.LightCandidates, agent.Router.LightModel(), nil
+}
+
+func (al *AgentLoop) resolveRequestedModelCandidates(
+	agent *AgentInstance,
+	requestedModel string,
+) ([]providers.FallbackCandidate, string, error) {
+	requestedModel = strings.TrimSpace(requestedModel)
+	if requestedModel == "" {
+		return agent.Candidates, agent.Model, nil
+	}
+
+	cfg := al.GetConfig()
+	resolveFromModelList := func(raw string) (string, bool) {
+		ensureProtocol := func(model string) string {
+			if model == "" {
+				return ""
+			}
+			if strings.Contains(model, "/") {
+				return model
+			}
+			return "openai/" + model
+		}
+
+		raw = strings.TrimSpace(raw)
+		if raw == "" || cfg == nil {
+			return "", false
+		}
+
+		if mc, err := cfg.GetModelConfig(raw); err == nil && mc != nil && strings.TrimSpace(mc.Model) != "" {
+			return ensureProtocol(mc.Model), true
+		}
+
+		for i := range cfg.ModelList {
+			fullModel := strings.TrimSpace(cfg.ModelList[i].Model)
+			if fullModel == "" {
+				continue
+			}
+			if fullModel == raw {
+				return ensureProtocol(fullModel), true
+			}
+			_, modelID := providers.ExtractProtocol(fullModel)
+			if modelID == raw {
+				return ensureProtocol(fullModel), true
+			}
+		}
+
+		return "", false
+	}
+
+	candidates := providers.ResolveCandidatesWithLookup(
+		providers.ModelConfig{Primary: requestedModel},
+		al.cfg.Agents.Defaults.Provider,
+		resolveFromModelList,
+	)
+	if len(candidates) == 0 {
+		return nil, "", fmt.Errorf("requested model %q not found in model_list", requestedModel)
+	}
+
+	return candidates, requestedModel, nil
 }
 
 // maybeSummarize triggers summarization if the session history exceeds thresholds.
@@ -1983,6 +2072,27 @@ func inboundMetadata(msg bus.InboundMessage, key string) string {
 	return msg.Metadata[key]
 }
 
+func metadataEnabled(msg bus.InboundMessage, key string) bool {
+	value := strings.ToLower(strings.TrimSpace(inboundMetadata(msg, key)))
+	return value == "1" || value == "true" || value == "yes" || value == "on"
+}
+
+func decodeInjectedHistory(msg bus.InboundMessage) []providers.Message {
+	raw := inboundMetadata(msg, metadataKeyInjectedHistory)
+	if strings.TrimSpace(raw) == "" {
+		return nil
+	}
+
+	var history []providers.Message
+	if err := json.Unmarshal([]byte(raw), &history); err != nil {
+		logger.WarnCF("agent", "Failed to decode injected history", map[string]any{
+			"error": err.Error(),
+		})
+		return nil
+	}
+	return history
+}
+
 // extractParentPeer extracts the parent peer (reply-to) from inbound message metadata.
 func extractParentPeer(msg bus.InboundMessage) *routing.RoutePeer {
 	parentKind := inboundMetadata(msg, metadataKeyParentPeerKind)

pkg/agent/openai_request_test.go

@@ -0,0 +1,82 @@
+package agent
+
+import (
+	"context"
+	"os"
+	"testing"
+
+	"github.com/sipeed/picoclaw/pkg/bus"
+	"github.com/sipeed/picoclaw/pkg/config"
+	"github.com/sipeed/picoclaw/pkg/providers"
+	"github.com/sipeed/picoclaw/pkg/routing"
+)
+
+type modelRecordingProvider struct {
+	lastModel string
+}
+
+func (p *modelRecordingProvider) Chat(
+	ctx context.Context,
+	messages []providers.Message,
+	tools []providers.ToolDefinition,
+	model string,
+	opts map[string]any,
+) (*providers.LLMResponse, error) {
+	p.lastModel = model
+	return &providers.LLMResponse{
+		Content:   "ok",
+		ToolCalls: nil,
+	}, nil
+}
+
+func (p *modelRecordingProvider) GetDefaultModel() string {
+	return "gpt-5.4"
+}
+
+func TestProcessMessage_OpenAIRequestDoesNotPersistHistoryAndOverridesModel(t *testing.T) {
+	tmpDir, err := os.MkdirTemp("", "agent-openai-*")
+	if err != nil {
+		t.Fatalf("MkdirTemp() error = %v", err)
+	}
+	defer os.RemoveAll(tmpDir)
+
+	cfg := config.DefaultConfig()
+	cfg.Agents.Defaults.Workspace = tmpDir
+	cfg.Agents.Defaults.Model = "gpt-5.4"
+	cfg.Agents.Defaults.Provider = "openai"
+	cfg.Agents.Defaults.MaxTokens = 4096
+	cfg.Agents.Defaults.MaxToolIterations = 4
+
+	messageBus := bus.NewMessageBus()
+	provider := &modelRecordingProvider{}
+	loop := NewAgentLoop(cfg, messageBus, provider)
+
+	_, err = loop.processMessage(context.Background(), bus.InboundMessage{
+		Channel:  "openai_api",
+		SenderID: "client-1",
+		ChatID:   "chat-1",
+		Content:  "hello",
+		Metadata: map[string]string{
+			"no_history":      "true",
+			"requested_model": "deepseek-chat",
+		},
+	})
+	if err != nil {
+		t.Fatalf("processMessage() error = %v", err)
+	}
+
+	if provider.lastModel != "deepseek-chat" {
+		t.Fatalf("provider model = %q, want %q", provider.lastModel, "deepseek-chat")
+	}
+
+	agent := loop.GetRegistry().GetDefaultAgent()
+	if agent == nil {
+		t.Fatal("expected default agent")
+	}
+
+	sessionKey := routing.BuildAgentMainSessionKey(agent.ID)
+	history := agent.Sessions.GetHistory(sessionKey)
+	if len(history) != 0 {
+		t.Fatalf("expected no persisted history, got %d entries", len(history))
+	}
+}

pkg/channels/manager.go

@@ -303,6 +303,10 @@ func (m *Manager) initChannels() error {
 		m.initChannel("pico", "Pico")
 	}
 
+	if m.config.Channels.OpenAIAPI.Enabled && m.config.Channels.OpenAIAPI.APIKey != "" {
+		m.initChannel("openai_api", "OpenAI API")
+	}
+
 	if m.config.Channels.IRC.Enabled && m.config.Channels.IRC.Server != "" {
 		m.initChannel("irc", "IRC")
 	}

pkg/channels/openai_api/init.go

@@ -0,0 +1,13 @@
+package openai_api
+
+import (
+	"github.com/sipeed/picoclaw/pkg/bus"
+	"github.com/sipeed/picoclaw/pkg/channels"
+	"github.com/sipeed/picoclaw/pkg/config"
+)
+
+func init() {
+	channels.RegisterFactory("openai_api", func(cfg *config.Config, b *bus.MessageBus) (channels.Channel, error) {
+		return NewOpenAIAPIChannel(cfg, b)
+	})
+}

pkg/channels/openai_api/openai_api.go

@@ -0,0 +1,750 @@
+package openai_api
+
+import (
+	"context"
+	"crypto/subtle"
+	"encoding/json"
+	"fmt"
+	"net"
+	"net/http"
+	"sort"
+	"strconv"
+	"strings"
+	"sync"
+	"time"
+
+	"github.com/google/uuid"
+
+	"github.com/sipeed/picoclaw/pkg/bus"
+	"github.com/sipeed/picoclaw/pkg/channels"
+	"github.com/sipeed/picoclaw/pkg/config"
+	"github.com/sipeed/picoclaw/pkg/logger"
+	"github.com/sipeed/picoclaw/pkg/providers"
+)
+
+const (
+	maxRequestBodySize  = 4 << 20
+	responseIdleWindow  = 250 * time.Millisecond
+	responseWaitTimeout = 5 * time.Minute
+)
+
+type responseTask struct {
+	ctx     context.Context
+	cancel  context.CancelFunc
+	updates chan string
+}
+
+type chatCompletionRequest struct {
+	Model    string                  `json:"model"`
+	Messages []chatCompletionMessage `json:"messages"`
+	Stream   bool                    `json:"stream,omitempty"`
+	User     string                  `json:"user,omitempty"`
+}
+
+type chatCompletionMessage struct {
+	Role       string         `json:"role"`
+	Content    any            `json:"content"`
+	ToolCallID string         `json:"tool_call_id,omitempty"`
+	ToolCalls  []chatToolCall `json:"tool_calls,omitempty"`
+}
+
+type chatToolCall struct {
+	ID       string            `json:"id,omitempty"`
+	Type     string            `json:"type,omitempty"`
+	Function *chatToolFunction `json:"function,omitempty"`
+}
+
+type chatToolFunction struct {
+	Name      string `json:"name"`
+	Arguments string `json:"arguments"`
+}
+
+type translatedConversation struct {
+	CurrentMessage    string
+	InjectedHistory   []providers.Message
+	ExtraSystemPrompt string
+}
+
+type OpenAIAPIChannel struct {
+	*channels.BaseChannel
+	config     config.OpenAIAPIConfig
+	listenHost string
+	models     []config.ModelConfig
+	messageBus *bus.MessageBus
+	server     *http.Server
+	listener   net.Listener
+	ctx        context.Context
+	cancel     context.CancelFunc
+	taskMu     sync.RWMutex
+	tasks      map[string]*responseTask
+}
+
+func NewOpenAIAPIChannel(cfg *config.Config, messageBus *bus.MessageBus) (*OpenAIAPIChannel, error) {
+	if cfg == nil {
+		return nil, fmt.Errorf("config is required")
+	}
+	if strings.TrimSpace(cfg.Channels.OpenAIAPI.APIKey) == "" {
+		return nil, fmt.Errorf("openai_api api_key is required")
+	}
+
+	listenHost := strings.TrimSpace(cfg.Gateway.Host)
+	if listenHost == "" {
+		listenHost = "127.0.0.1"
+	}
+
+	base := channels.NewBaseChannel("openai_api", cfg.Channels.OpenAIAPI, messageBus, nil)
+
+	return &OpenAIAPIChannel{
+		BaseChannel: base,
+		config:      cfg.Channels.OpenAIAPI,
+		listenHost:  listenHost,
+		models:      append([]config.ModelConfig(nil), cfg.ModelList...),
+		messageBus:  messageBus,
+		tasks:       make(map[string]*responseTask),
+	}, nil
+}
+
+func (c *OpenAIAPIChannel) Start(ctx context.Context) error {
+	addr := net.JoinHostPort(c.listenHost, strconv.Itoa(c.config.Port))
+	listener, err := net.Listen("tcp", addr)
+	if err != nil {
+		return fmt.Errorf("openai_api listen %s: %w", addr, err)
+	}
+
+	c.ctx, c.cancel = context.WithCancel(ctx)
+	c.listener = listener
+
+	mux := http.NewServeMux()
+	mux.HandleFunc("OPTIONS /v1/models", c.handleOptions)
+	mux.HandleFunc("GET /v1/models", c.handleModels)
+	mux.HandleFunc("OPTIONS /v1/chat/completions", c.handleOptions)
+	mux.HandleFunc("POST /v1/chat/completions", c.handleChatCompletions)
+	mux.HandleFunc("GET /health", c.handleHealth)
+
+	c.server = &http.Server{
+		Handler:     mux,
+		ReadTimeout: 30 * time.Second,
+		// Streaming responses stay open until the request finishes.
+		WriteTimeout: 0,
+	}
+
+	c.SetRunning(true)
+	logger.InfoCF("openai_api", "OpenAI API channel listening", map[string]any{
+		"addr": addr,
+	})
+
+	go func() {
+		if err := c.server.Serve(listener); err != nil && err != http.ErrServerClosed {
+			logger.ErrorCF("openai_api", "OpenAI API server stopped unexpectedly", map[string]any{
+				"error": err.Error(),
+			})
+		}
+	}()
+
+	return nil
+}
+
+func (c *OpenAIAPIChannel) Stop(ctx context.Context) error {
+	c.SetRunning(false)
+
+	if c.cancel != nil {
+		c.cancel()
+	}
+
+	c.taskMu.Lock()
+	for chatID, task := range c.tasks {
+		if task.cancel != nil {
+			task.cancel()
+		}
+		delete(c.tasks, chatID)
+	}
+	c.taskMu.Unlock()
+
+	if c.server != nil {
+		if err := c.server.Shutdown(ctx); err != nil {
+			return err
+		}
+	}
+
+	return nil
+}
+
+func (c *OpenAIAPIChannel) Send(ctx context.Context, msg bus.OutboundMessage) error {
+	if !c.IsRunning() {
+		return channels.ErrNotRunning
+	}
+	content := strings.TrimSpace(msg.Content)
+	if content == "" {
+		return nil
+	}
+
+	task := c.getTask(msg.ChatID)
+	if task == nil {
+		logger.DebugCF("openai_api", "Dropping outbound response with no waiting request", map[string]any{
+			"chat_id": msg.ChatID,
+		})
+		return nil
+	}
+
+	select {
+	case task.updates <- content:
+		return nil
+	case <-task.ctx.Done():
+		return nil
+	case <-ctx.Done():
+		return ctx.Err()
+	}
+}
+
+func (c *OpenAIAPIChannel) handleOptions(w http.ResponseWriter, r *http.Request) {
+	setCORSHeaders(w)
+	w.WriteHeader(http.StatusNoContent)
+}
+
+func (c *OpenAIAPIChannel) handleHealth(w http.ResponseWriter, r *http.Request) {
+	w.Header().Set("Content-Type", "application/json")
+	json.NewEncoder(w).Encode(map[string]any{"status": "ok"})
+}
+
+func (c *OpenAIAPIChannel) handleModels(w http.ResponseWriter, r *http.Request) {
+	setCORSHeaders(w)
+	if !c.authenticate(r) {
+		writeOpenAIError(w, http.StatusUnauthorized, "Invalid API key", "invalid_request_error", "invalid_api_key")
+		return
+	}
+
+	type modelObject struct {
+		ID      string `json:"id"`
+		Object  string `json:"object"`
+		Created int64  `json:"created"`
+		OwnedBy string `json:"owned_by"`
+	}
+
+	seen := make(map[string]bool)
+	items := make([]modelObject, 0, len(c.models))
+	for _, model := range c.models {
+		id := strings.TrimSpace(model.ModelName)
+		if id == "" || seen[id] {
+			continue
+		}
+		seen[id] = true
+		items = append(items, modelObject{
+			ID:      id,
+			Object:  "model",
+			Created: 0,
+			OwnedBy: "picoclaw",
+		})
+	}
+
+	sort.Slice(items, func(i, j int) bool {
+		return items[i].ID < items[j].ID
+	})
+
+	w.Header().Set("Content-Type", "application/json")
+	json.NewEncoder(w).Encode(map[string]any{
+		"object": "list",
+		"data":   items,
+	})
+}
+
+func (c *OpenAIAPIChannel) handleChatCompletions(w http.ResponseWriter, r *http.Request) {
+	setCORSHeaders(w)
+	if !c.authenticate(r) {
+		writeOpenAIError(w, http.StatusUnauthorized, "Invalid API key", "invalid_request_error", "invalid_api_key")
+		return
+	}
+
+	var req chatCompletionRequest
+	decoder := json.NewDecoder(http.MaxBytesReader(w, r.Body, maxRequestBodySize))
+	if err := decoder.Decode(&req); err != nil {
+		writeOpenAIError(w, http.StatusBadRequest, "Invalid JSON request body", "invalid_request_error", "invalid_json")
+		return
+	}
+
+	if strings.TrimSpace(req.Model) == "" {
+		writeOpenAIError(w, http.StatusBadRequest, "model is required", "invalid_request_error", "missing_model")
+		return
+	}
+	if !c.supportsModel(req.Model) {
+		writeOpenAIError(w, http.StatusBadRequest, fmt.Sprintf("model %q is not configured", req.Model), "invalid_request_error", "model_not_found")
+		return
+	}
+	if len(req.Messages) == 0 {
+		writeOpenAIError(w, http.StatusBadRequest, "messages must not be empty", "invalid_request_error", "missing_messages")
+		return
+	}
+
+	translated, err := translateConversation(req.Messages)
+	if err != nil {
+		writeOpenAIError(w, http.StatusBadRequest, err.Error(), "invalid_request_error", "invalid_messages")
+		return
+	}
+
+	reqCtx, cancel := context.WithTimeout(r.Context(), responseWaitTimeout)
+	defer cancel()
+
+	chatID := "openai_api:" + uuid.NewString()
+	task := &responseTask{
+		ctx:     reqCtx,
+		cancel:  cancel,
+		updates: make(chan string, 32),
+	}
+	c.setTask(chatID, task)
+	defer c.deleteTask(chatID)
+
+	metadata := map[string]string{
+		"no_history":      "true",
+		"requested_model": req.Model,
+	}
+	if strings.TrimSpace(translated.ExtraSystemPrompt) != "" {
+		metadata["extra_system_prompt"] = translated.ExtraSystemPrompt
+	}
+	if len(translated.InjectedHistory) > 0 {
+		rawHistory, err := json.Marshal(translated.InjectedHistory)
+		if err != nil {
+			writeOpenAIError(w, http.StatusInternalServerError, "Failed to encode conversation history", "server_error", "history_encode_failed")
+			return
+		}
+		metadata["injected_history"] = string(rawHistory)
+	}
+
+	senderID := strings.TrimSpace(req.User)
+	if senderID == "" {
+		senderID = "openai-client"
+	}
+
+	if err := c.messageBus.PublishInbound(reqCtx, bus.InboundMessage{
+		Channel:  c.Name(),
+		SenderID: senderID,
+		Sender: bus.SenderInfo{
+			Platform:    c.Name(),
+			PlatformID:  senderID,
+			CanonicalID: c.Name() + ":" + senderID,
+			DisplayName: senderID,
+		},
+		ChatID:   chatID,
+		Content:  translated.CurrentMessage,
+		Peer:     bus.Peer{Kind: "direct", ID: senderID},
+		Metadata: metadata,
+	}); err != nil {
+		writeOpenAIError(w, http.StatusBadGateway, fmt.Sprintf("Failed to submit request: %v", err), "server_error", "publish_failed")
+		return
+	}
+
+	firstChunk, err := waitForFirstChunk(reqCtx, task)
+	if err != nil {
+		writeOpenAIError(w, http.StatusGatewayTimeout, "Timed out waiting for assistant response", "server_error", "response_timeout")
+		return
+	}
+
+	completionID := "chatcmpl_" + strings.ReplaceAll(uuid.NewString(), "-", "")
+	createdAt := time.Now().Unix()
+
+	if req.Stream {
+		flusher, ok := w.(http.Flusher)
+		if !ok {
+			writeOpenAIError(w, http.StatusInternalServerError, "Streaming is not supported by this server", "server_error", "stream_not_supported")
+			return
+		}
+
+		w.Header().Set("Content-Type", "text/event-stream")
+		w.Header().Set("Cache-Control", "no-cache")
+		w.Header().Set("Connection", "keep-alive")
+
+		if err := writeChatCompletionChunk(w, completionID, createdAt, req.Model, firstChunk, true, false); err != nil {
+			return
+		}
+		flusher.Flush()
+
+		if err := streamRemainingChunks(reqCtx, task, w, flusher, completionID, createdAt, req.Model); err != nil {
+			return
+		}
+
+		_ = writeChatCompletionChunk(w, completionID, createdAt, req.Model, "", false, true)
+		_, _ = fmt.Fprint(w, "data: [DONE]\n\n")
+		flusher.Flush()
+		return
+	}
+
+	chunks, err := collectRemainingChunks(reqCtx, task, []string{firstChunk})
+	if err != nil {
+		writeOpenAIError(w, http.StatusGatewayTimeout, "Timed out waiting for assistant response", "server_error", "response_timeout")
+		return
+	}
+
+	w.Header().Set("Content-Type", "application/json")
+	json.NewEncoder(w).Encode(map[string]any{
+		"id":      completionID,
+		"object":  "chat.completion",
+		"created": createdAt,
+		"model":   req.Model,
+		"choices": []map[string]any{
+			{
+				"index": 0,
+				"message": map[string]any{
+					"role":    "assistant",
+					"content": strings.Join(chunks, "\n\n"),
+				},
+				"finish_reason": "stop",
+			},
+		},
+	})
+}
+
+func translateConversation(messages []chatCompletionMessage) (translatedConversation, error) {
+	var out translatedConversation
+	var nonSystem []providers.Message
+	var systemPrompts []string
+
+	for _, message := range messages {
+		role := strings.ToLower(strings.TrimSpace(message.Role))
+		content := strings.TrimSpace(extractMessageContent(message))
+
+		switch role {
+		case "system", "developer":
+			if content != "" {
+				systemPrompts = append(systemPrompts, content)
+			}
+		case "user", "assistant", "tool":
+			pm := providers.Message{
+				Role:    role,
+				Content: content,
+			}
+			if role == "tool" {
+				pm.ToolCallID = strings.TrimSpace(message.ToolCallID)
+			}
+			nonSystem = append(nonSystem, pm)
+		default:
+			if content == "" {
+				continue
+			}
+			nonSystem = append(nonSystem, providers.Message{
+				Role:    "user",
+				Content: fmt.Sprintf("[%s]\n%s", role, content),
+			})
+		}
+	}
+
+	if len(nonSystem) == 0 {
+		return translatedConversation{}, fmt.Errorf("at least one non-system message is required")
+	}
+
+	out.ExtraSystemPrompt = strings.Join(systemPrompts, "\n\n")
+	last := nonSystem[len(nonSystem)-1]
+
+	if last.Role == "user" && strings.TrimSpace(last.Content) != "" {
+		out.CurrentMessage = last.Content
+		out.InjectedHistory = append([]providers.Message(nil), nonSystem[:len(nonSystem)-1]...)
+		return out, nil
+	}
+
+	out.InjectedHistory = append([]providers.Message(nil), nonSystem...)
+	switch last.Role {
+	case "assistant":
+		out.CurrentMessage = "Continue the conversation with the next assistant response."
+	case "tool":
+		out.CurrentMessage = "Continue the conversation after the tool result above."
+	default:
+		out.CurrentMessage = "Continue the conversation based on the previous messages."
+	}
+	return out, nil
+}
+
+func extractMessageContent(message chatCompletionMessage) string {
+	content := strings.TrimSpace(extractContentText(message.Content))
+	toolCalls := strings.TrimSpace(renderToolCalls(message.ToolCalls))
+
+	switch {
+	case content == "":
+		return toolCalls
+	case toolCalls == "":
+		return content
+	default:
+		return content + "\n\n" + toolCalls
+	}
+}
+
+func renderToolCalls(toolCalls []chatToolCall) string {
+	if len(toolCalls) == 0 {
+		return ""
+	}
+
+	lines := make([]string, 0, len(toolCalls))
+	for _, toolCall := range toolCalls {
+		if toolCall.Function == nil {
+			continue
+		}
+		name := strings.TrimSpace(toolCall.Function.Name)
+		args := strings.TrimSpace(toolCall.Function.Arguments)
+		if name == "" {
+			continue
+		}
+		if args == "" {
+			lines = append(lines, fmt.Sprintf("[assistant_tool_call] %s()", name))
+			continue
+		}
+		lines = append(lines, fmt.Sprintf("[assistant_tool_call] %s(%s)", name, args))
+	}
+	return strings.Join(lines, "\n")
+}
+
+func extractContentText(content any) string {
+	switch value := content.(type) {
+	case nil:
+		return ""
+	case string:
+		return value
+	case []any:
+		parts := make([]string, 0, len(value))
+		for _, rawPart := range value {
+			part, ok := rawPart.(map[string]any)
+			if !ok {
+				continue
+			}
+			partType, _ := part["type"].(string)
+			switch partType {
+			case "text", "input_text", "output_text":
+				if text, ok := stringValue(part["text"]); ok && text != "" {
+					parts = append(parts, text)
+					continue
+				}
+				if text, ok := stringValue(part["input_text"]); ok && text != "" {
+					parts = append(parts, text)
+					continue
+				}
+				if text, ok := stringValue(part["output_text"]); ok && text != "" {
+					parts = append(parts, text)
+				}
+			case "image_url", "input_image":
+				if url := extractImageURL(part["image_url"]); url != "" {
+					parts = append(parts, "[image] "+url)
+				}
+			default:
+				if text, ok := stringValue(part["text"]); ok && text != "" {
+					parts = append(parts, text)
+				}
+			}
+		}
+		return strings.Join(parts, "\n")
+	default:
+		return ""
+	}
+}
+
+func extractImageURL(value any) string {
+	switch v := value.(type) {
+	case string:
+		return strings.TrimSpace(v)
+	case map[string]any:
+		if url, ok := stringValue(v["url"]); ok {
+			return url
+		}
+	}
+	return ""
+}
+
+func stringValue(value any) (string, bool) {
+	text, ok := value.(string)
+	if !ok {
+		return "", false
+	}
+	text = strings.TrimSpace(text)
+	if text == "" {
+		return "", false
+	}
+	return text, true
+}
+
+func waitForFirstChunk(ctx context.Context, task *responseTask) (string, error) {
+	for {
+		select {
+		case chunk := <-task.updates:
+			if strings.TrimSpace(chunk) != "" {
+				return chunk, nil
+			}
+		case <-ctx.Done():
+			return "", ctx.Err()
+		}
+	}
+}
+
+func collectRemainingChunks(ctx context.Context, task *responseTask, initial []string) ([]string, error) {
+	chunks := append([]string(nil), initial...)
+	timer := time.NewTimer(responseIdleWindow)
+	defer timer.Stop()
+
+	for {
+		select {
+		case chunk := <-task.updates:
+			if strings.TrimSpace(chunk) == "" {
+				continue
+			}
+			chunks = append(chunks, chunk)
+			if !timer.Stop() {
+				select {
+				case <-timer.C:
+				default:
+				}
+			}
+			timer.Reset(responseIdleWindow)
+		case <-timer.C:
+			return chunks, nil
+		case <-ctx.Done():
+			return nil, ctx.Err()
+		}
+	}
+}
+
+func streamRemainingChunks(
+	ctx context.Context,
+	task *responseTask,
+	w http.ResponseWriter,
+	flusher http.Flusher,
+	completionID string,
+	createdAt int64,
+	model string,
+) error {
+	timer := time.NewTimer(responseIdleWindow)
+	defer timer.Stop()
+
+	for {
+		select {
+		case chunk := <-task.updates:
+			if strings.TrimSpace(chunk) == "" {
+				continue
+			}
+			if err := writeChatCompletionChunk(w, completionID, createdAt, model, chunk, false, false); err != nil {
+				return err
+			}
+			flusher.Flush()
+			if !timer.Stop() {
+				select {
+				case <-timer.C:
+				default:
+				}
+			}
+			timer.Reset(responseIdleWindow)
+		case <-timer.C:
+			return nil
+		case <-ctx.Done():
+			return ctx.Err()
+		}
+	}
+}
+
+func writeChatCompletionChunk(
+	w http.ResponseWriter,
+	completionID string,
+	createdAt int64,
+	model string,
+	content string,
+	includeRole bool,
+	finished bool,
+) error {
+	delta := map[string]any{}
+	if includeRole {
+		delta["role"] = "assistant"
+	}
+	if content != "" {
+		delta["content"] = content
+	}
+
+	finishReason := any(nil)
+	if finished {
+		finishReason = "stop"
+	}
+
+	payload, err := json.Marshal(map[string]any{
+		"id":      completionID,
+		"object":  "chat.completion.chunk",
+		"created": createdAt,
+		"model":   model,
+		"choices": []map[string]any{
+			{
+				"index":         0,
+				"delta":         delta,
+				"finish_reason": finishReason,
+			},
+		},
+	})
+	if err != nil {
+		return err
+	}
+
+	_, err = fmt.Fprintf(w, "data: %s\n\n", payload)
+	return err
+}
+
+func writeOpenAIError(w http.ResponseWriter, status int, message, errorType, code string) {
+	setCORSHeaders(w)
+	w.Header().Set("Content-Type", "application/json")
+	w.WriteHeader(status)
+	_ = json.NewEncoder(w).Encode(map[string]any{
+		"error": map[string]any{
+			"message": message,
+			"type":    errorType,
+			"code":    code,
+		},
+	})
+}
+
+func setCORSHeaders(w http.ResponseWriter) {
+	w.Header().Set("Access-Control-Allow-Origin", "*")
+	w.Header().Set("Access-Control-Allow-Headers", "Authorization, Content-Type")
+	w.Header().Set("Access-Control-Allow-Methods", "GET, POST, OPTIONS")
+}
+
+func (c *OpenAIAPIChannel) authenticate(r *http.Request) bool {
+	token := strings.TrimSpace(strings.TrimPrefix(r.Header.Get("Authorization"), "Bearer "))
+	if token == "" {
+		return false
+	}
+	expected := strings.TrimSpace(c.config.APIKey)
+	if expected == "" {
+		return false
+	}
+	return subtle.ConstantTimeCompare([]byte(token), []byte(expected)) == 1
+}
+
+func (c *OpenAIAPIChannel) supportsModel(requested string) bool {
+	requested = strings.TrimSpace(requested)
+	if requested == "" {
+		return false
+	}
+
+	for _, model := range c.models {
+		if strings.TrimSpace(model.ModelName) == requested {
+			return true
+		}
+		if strings.TrimSpace(model.Model) == requested {
+			return true
+		}
+		_, modelID := providers.ExtractProtocol(model.Model)
+		if modelID == requested {
+			return true
+		}
+	}
+
+	return false
+}
+
+func (c *OpenAIAPIChannel) setTask(chatID string, task *responseTask) {
+	c.taskMu.Lock()
+	defer c.taskMu.Unlock()
+	c.tasks[chatID] = task
+}
+
+func (c *OpenAIAPIChannel) getTask(chatID string) *responseTask {
+	c.taskMu.RLock()
+	defer c.taskMu.RUnlock()
+	return c.tasks[chatID]
+}
+
+func (c *OpenAIAPIChannel) deleteTask(chatID string) {
+	c.taskMu.Lock()
+	defer c.taskMu.Unlock()
+	if task, ok := c.tasks[chatID]; ok && task.cancel != nil {
+		task.cancel()
+	}
+	delete(c.tasks, chatID)
+}

pkg/channels/openai_api/openai_api_test.go

@@ -0,0 +1,166 @@
+package openai_api
+
+import (
+	"context"
+	"encoding/json"
+	"net/http"
+	"net/http/httptest"
+	"strings"
+	"testing"
+	"time"
+
+	"github.com/sipeed/picoclaw/pkg/bus"
+	"github.com/sipeed/picoclaw/pkg/config"
+)
+
+func newTestChannel(t *testing.T) (*OpenAIAPIChannel, *bus.MessageBus) {
+	t.Helper()
+
+	cfg := config.DefaultConfig()
+	cfg.Channels.OpenAIAPI.APIKey = "test-key"
+	cfg.Channels.OpenAIAPI.Port = 0
+
+	messageBus := bus.NewMessageBus()
+	channel, err := NewOpenAIAPIChannel(cfg, messageBus)
+	if err != nil {
+		t.Fatalf("NewOpenAIAPIChannel() error = %v", err)
+	}
+	channel.SetRunning(true)
+
+	return channel, messageBus
+}
+
+func TestHandleChatCompletions_NonStreaming(t *testing.T) {
+	channel, messageBus := newTestChannel(t)
+
+	go func() {
+		msg := <-messageBus.InboundChan()
+		if msg.Content != "Final user question" {
+			t.Errorf("inbound content = %q, want %q", msg.Content, "Final user question")
+		}
+		if got := msg.Metadata["requested_model"]; got != "gpt-5.4" {
+			t.Errorf("requested_model = %q, want %q", got, "gpt-5.4")
+		}
+		if got := msg.Metadata["no_history"]; got != "true" {
+			t.Errorf("no_history = %q, want %q", got, "true")
+		}
+		if msg.Metadata["extra_system_prompt"] == "" {
+			t.Error("expected extra_system_prompt metadata to be populated")
+		}
+		if msg.Metadata["injected_history"] == "" {
+			t.Error("expected injected_history metadata to be populated")
+		}
+
+		_ = channel.Send(context.Background(), bus.OutboundMessage{
+			Channel: channel.Name(),
+			ChatID:  msg.ChatID,
+			Content: "Assistant response",
+		})
+	}()
+
+	body := `{
+		"model": "gpt-5.4",
+		"messages": [
+			{"role": "system", "content": "Be concise."},
+			{"role": "user", "content": "Earlier question"},
+			{"role": "assistant", "content": "Earlier answer"},
+			{"role": "user", "content": "Final user question"}
+		]
+	}`
+
+	req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", strings.NewReader(body))
+	req.Header.Set("Authorization", "Bearer test-key")
+	rec := httptest.NewRecorder()
+
+	channel.handleChatCompletions(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusOK, rec.Body.String())
+	}
+
+	var response struct {
+		Model   string `json:"model"`
+		Choices []struct {
+			Message struct {
+				Role    string `json:"role"`
+				Content string `json:"content"`
+			} `json:"message"`
+		} `json:"choices"`
+	}
+	if err := json.Unmarshal(rec.Body.Bytes(), &response); err != nil {
+		t.Fatalf("Unmarshal() error = %v", err)
+	}
+	if response.Model != "gpt-5.4" {
+		t.Fatalf("model = %q, want %q", response.Model, "gpt-5.4")
+	}
+	if len(response.Choices) != 1 {
+		t.Fatalf("len(choices) = %d, want 1", len(response.Choices))
+	}
+	if response.Choices[0].Message.Role != "assistant" {
+		t.Fatalf("role = %q, want %q", response.Choices[0].Message.Role, "assistant")
+	}
+	if response.Choices[0].Message.Content != "Assistant response" {
+		t.Fatalf("content = %q, want %q", response.Choices[0].Message.Content, "Assistant response")
+	}
+}
+
+func TestHandleChatCompletions_Streaming(t *testing.T) {
+	channel, messageBus := newTestChannel(t)
+
+	go func() {
+		msg := <-messageBus.InboundChan()
+		_ = channel.Send(context.Background(), bus.OutboundMessage{
+			Channel: channel.Name(),
+			ChatID:  msg.ChatID,
+			Content: "part one",
+		})
+		time.Sleep(50 * time.Millisecond)
+		_ = channel.Send(context.Background(), bus.OutboundMessage{
+			Channel: channel.Name(),
+			ChatID:  msg.ChatID,
+			Content: "part two",
+		})
+	}()
+
+	body := `{
+		"model": "gpt-5.4",
+		"stream": true,
+		"messages": [
+			{"role": "user", "content": "Stream this please"}
+		]
+	}`
+
+	req := httptest.NewRequest(http.MethodPost, "/v1/chat/completions", strings.NewReader(body))
+	req.Header.Set("Authorization", "Bearer test-key")
+	rec := httptest.NewRecorder()
+
+	channel.handleChatCompletions(rec, req)
+
+	if rec.Code != http.StatusOK {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusOK, rec.Body.String())
+	}
+
+	bodyText := rec.Body.String()
+	if !strings.Contains(bodyText, `"content":"part one"`) {
+		t.Fatalf("stream body missing first chunk: %s", bodyText)
+	}
+	if !strings.Contains(bodyText, `"content":"part two"`) {
+		t.Fatalf("stream body missing second chunk: %s", bodyText)
+	}
+	if !strings.Contains(bodyText, "data: [DONE]") {
+		t.Fatalf("stream body missing [DONE]: %s", bodyText)
+	}
+}
+
+func TestHandleModels_RequiresAuth(t *testing.T) {
+	channel, _ := newTestChannel(t)
+
+	req := httptest.NewRequest(http.MethodGet, "/v1/models", nil)
+	rec := httptest.NewRecorder()
+
+	channel.handleModels(rec, req)
+
+	if rec.Code != http.StatusUnauthorized {
+		t.Fatalf("status = %d, want %d", rec.Code, http.StatusUnauthorized)
+	}
+}

pkg/config/config.go

@@ -272,6 +272,7 @@ type ChannelsConfig struct {
 	WeComApp   WeComAppConfig   `json:"wecom_app"`
 	WeComAIBot WeComAIBotConfig `json:"wecom_aibot"`
 	Pico       PicoConfig       `json:"pico"`
+	OpenAIAPI  OpenAIAPIConfig  `json:"openai_api"`
 	IRC        IRCConfig        `json:"irc"`
 }
 
@@ -473,6 +474,12 @@ type PicoConfig struct {
 	Placeholder     PlaceholderConfig   `json:"placeholder,omitempty"`
 }
 
+type OpenAIAPIConfig struct {
+	Enabled bool   `json:"enabled" env:"PICOCLAW_CHANNELS_OPENAI_API_ENABLED"`
+	Port    int    `json:"port"    env:"PICOCLAW_CHANNELS_OPENAI_API_PORT"`
+	APIKey  string `json:"api_key" env:"PICOCLAW_CHANNELS_OPENAI_API_API_KEY"`
+}
+
 type IRCConfig struct {
 	Enabled            bool                `json:"enabled"                 env:"PICOCLAW_CHANNELS_IRC_ENABLED"`
 	Server             string              `json:"server"                  env:"PICOCLAW_CHANNELS_IRC_SERVER"`

pkg/config/defaults.go

@@ -175,6 +175,11 @@ func DefaultConfig() *Config {
 				MaxConnections: 100,
 				AllowFrom:      FlexibleStringSlice{},
 			},
+			OpenAIAPI: OpenAIAPIConfig{
+				Enabled: false,
+				Port:    18794,
+				APIKey:  "",
+			},
 		},
 		Providers: ProvidersConfig{
 			OpenAI: OpenAIProviderConfig{WebSearch: true},

pkg/gateway/gateway.go

@@ -21,6 +21,7 @@ import (
 	_ "github.com/sipeed/picoclaw/pkg/channels/maixcam"
 	_ "github.com/sipeed/picoclaw/pkg/channels/matrix"
 	_ "github.com/sipeed/picoclaw/pkg/channels/onebot"
+	_ "github.com/sipeed/picoclaw/pkg/channels/openai_api"
 	_ "github.com/sipeed/picoclaw/pkg/channels/pico"
 	_ "github.com/sipeed/picoclaw/pkg/channels/qq"
 	_ "github.com/sipeed/picoclaw/pkg/channels/slack"

web/backend/api/channels.go

@@ -26,6 +26,7 @@ var channelCatalog = []channelCatalogItem{
 	{Name: "whatsapp", ConfigKey: "whatsapp", Variant: "bridge"},
 	{Name: "whatsapp_native", ConfigKey: "whatsapp", Variant: "native"},
 	{Name: "pico", ConfigKey: "pico"},
+	{Name: "openai_api", ConfigKey: "openai_api"},
 	{Name: "maixcam", ConfigKey: "maixcam"},
 	{Name: "matrix", ConfigKey: "matrix"},
 	{Name: "irc", ConfigKey: "irc"},

web/backend/api/config.go

@@ -179,6 +179,20 @@ func validateConfig(cfg *config.Config) []string {
 		errs = append(errs, "channels.pico.token is required when pico channel is enabled")
 	}
 
+	if cfg.Channels.OpenAIAPI.Enabled && (cfg.Channels.OpenAIAPI.Port < 1 || cfg.Channels.OpenAIAPI.Port > 65535) {
+		errs = append(errs, fmt.Sprintf("channels.openai_api.port %d is out of valid range (1-65535)", cfg.Channels.OpenAIAPI.Port))
+	} else if cfg.Channels.OpenAIAPI.Port != 0 && (cfg.Channels.OpenAIAPI.Port < 1 || cfg.Channels.OpenAIAPI.Port > 65535) {
+		errs = append(errs, fmt.Sprintf("channels.openai_api.port %d is out of valid range (1-65535)", cfg.Channels.OpenAIAPI.Port))
+	}
+
+	if cfg.Channels.OpenAIAPI.Enabled && cfg.Channels.OpenAIAPI.APIKey == "" {
+		errs = append(errs, "channels.openai_api.api_key is required when openai_api channel is enabled")
+	}
+
+	if cfg.Channels.OpenAIAPI.Enabled && cfg.Channels.OpenAIAPI.Port == cfg.Gateway.Port {
+		errs = append(errs, "channels.openai_api.port must differ from gateway.port")
+	}
+
 	// Telegram: token required when enabled
 	if cfg.Channels.Telegram.Enabled && cfg.Channels.Telegram.Token == "" {
 		errs = append(errs, "channels.telegram.token is required when telegram channel is enabled")

web/backend/api/config_test.go

@@ -165,3 +165,64 @@ func TestHandlePatchConfig_AllowsInvalidDenyRegexPatternsWhenDenyPatternsDisable
 		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusOK, rec.Body.String())
 	}
 }
+
+func TestHandlePatchConfig_RejectsOpenAIAPIWithoutKeyWhenEnabled(t *testing.T) {
+	configPath, cleanup := setupOAuthTestEnv(t)
+	defer cleanup()
+
+	h := NewHandler(configPath)
+	mux := http.NewServeMux()
+	h.RegisterRoutes(mux)
+
+	req := httptest.NewRequest(http.MethodPatch, "/api/config", bytes.NewBufferString(`{
+		"channels": {
+			"openai_api": {
+				"enabled": true,
+				"port": 18794,
+				"api_key": ""
+			}
+		}
+	}`))
+	req.Header.Set("Content-Type", "application/json")
+
+	rec := httptest.NewRecorder()
+	mux.ServeHTTP(rec, req)
+	if rec.Code != http.StatusBadRequest {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusBadRequest, rec.Body.String())
+	}
+	if !bytes.Contains(rec.Body.Bytes(), []byte("channels.openai_api.api_key")) {
+		t.Fatalf("expected openai_api api_key validation error, body=%s", rec.Body.String())
+	}
+}
+
+func TestHandlePatchConfig_RejectsOpenAIAPIPortConflictWithGateway(t *testing.T) {
+	configPath, cleanup := setupOAuthTestEnv(t)
+	defer cleanup()
+
+	h := NewHandler(configPath)
+	mux := http.NewServeMux()
+	h.RegisterRoutes(mux)
+
+	req := httptest.NewRequest(http.MethodPatch, "/api/config", bytes.NewBufferString(`{
+		"gateway": {
+			"port": 18790
+		},
+		"channels": {
+			"openai_api": {
+				"enabled": true,
+				"port": 18790,
+				"api_key": "test-key"
+			}
+		}
+	}`))
+	req.Header.Set("Content-Type", "application/json")
+
+	rec := httptest.NewRecorder()
+	mux.ServeHTTP(rec, req)
+	if rec.Code != http.StatusBadRequest {
+		t.Fatalf("status = %d, want %d, body=%s", rec.Code, http.StatusBadRequest, rec.Body.String())
+	}
+	if !bytes.Contains(rec.Body.Bytes(), []byte("channels.openai_api.port must differ from gateway.port")) {
+		t.Fatalf("expected openai_api port conflict validation error, body=%s", rec.Body.String())
+	}
+}

web/frontend/src/components/channels/channel-config-page.tsx

@@ -36,6 +36,7 @@ const SECRET_FIELD_MAP: Record<string, string> = {
   access_token: "_access_token",
   bot_token: "_bot_token",
   app_token: "_app_token",
+  api_key: "_api_key",
   encoding_aes_key: "_encoding_aes_key",
   encrypt_key: "_encrypt_key",
   verification_token: "_verification_token",
@@ -156,6 +157,8 @@ function isConfigured(
       return asBool(config.use_native)
     case "pico":
       return asString(config.token) !== ""
+    case "openai_api":
+      return asString(config.api_key) !== ""
     case "maixcam":
       return asString(config.host) !== ""
     case "matrix":
@@ -199,6 +202,8 @@ function getRequiredFieldKeys(channelName: string): string[] {
       return ["bridge_url"]
     case "pico":
       return ["token"]
+    case "openai_api":
+      return ["api_key", "port"]
     case "maixcam":
       return ["host"]
     case "matrix":
@@ -229,6 +234,7 @@ function getChannelDocSlug(channelName: string): string {
 
 const CHANNELS_WITHOUT_DOCS = new Set([
   "pico",
+  "openai_api",
   "wecom",
   "matrix",
   "irc",

web/frontend/src/components/channels/channel-forms/generic-form.tsx

@@ -17,6 +17,7 @@ interface GenericFormProps {
 // Secret field names that should use masked input.
 const SECRET_FIELDS = new Set([
   "token",
+  "api_key",
   "app_secret",
   "client_secret",
   "corp_secret",
@@ -47,6 +48,23 @@ const OBJECT_FIELDS = new Set([
 ])
 
 function formatLabel(key: string): string {
+  const overrides: Record<string, string> = {
+    api_key: "API Key",
+    ws_url: "WS URL",
+    app_id: "App ID",
+    user_id: "User ID",
+    device_id: "Device ID",
+    client_id: "Client ID",
+    corp_id: "Corp ID",
+    agent_id: "Agent ID",
+    webhook_url: "Webhook URL",
+    webhook_host: "Webhook Host",
+    webhook_port: "Webhook Port",
+    webhook_path: "Webhook Path",
+  }
+  if (overrides[key]) {
+    return overrides[key]
+  }
   return key
     .split("_")
     .map((w) => w.charAt(0).toUpperCase() + w.slice(1))
@@ -104,6 +122,7 @@ export function GenericForm({
 
   const buildHint = (key: string): string => {
     const descriptions: Record<string, string> = {
+      api_key: t("channels.form.desc.apiKey"),
       ws_url: t("channels.form.desc.wsUrl"),
       reconnect_interval: t("channels.form.desc.reconnectInterval"),
       bridge_url: t("channels.form.desc.bridgeUrl"),

web/frontend/src/hooks/use-sidebar-channels.ts

@@ -42,6 +42,7 @@ const CHANNEL_IMPORTANCE_ORDER = [
   "onebot",
   "matrix",
   "pico",
+  "openai_api",
   "maixcam",
   "irc",
   "whatsapp",
@@ -84,6 +85,7 @@ const CHANNEL_ICON_MAP: Record<
   maixcam: IconCamera,
   onebot: IconRobot,
   pico: IconBrandChrome,
+  openai_api: IconPlug,
   irc: IconMessages,
 }
 

web/frontend/src/i18n/locales/en.json

@@ -238,6 +238,7 @@
       "whatsapp": "WhatsApp",
       "whatsapp_native": "WhatsApp Native",
       "pico": "Web",
+      "openai_api": "OpenAI API",
       "maixcam": "MaixCam",
       "matrix": "Matrix",
       "irc": "IRC"
@@ -277,6 +278,7 @@
     "form": {
       "desc": {
         "token": "Bot access token used to connect to the platform API.",
+        "apiKey": "Authentication key used by external clients to access the OpenAI-compatible endpoint.",
         "botToken": "Bot token used to send and receive messages.",
         "appToken": "App token used for Socket Mode connections.",
         "appId": "Unique application ID used for authentication.",

web/frontend/src/i18n/locales/zh.json

@@ -238,6 +238,7 @@
       "whatsapp": "WhatsApp",
       "whatsapp_native": "WhatsApp Native",
       "pico": "Web",
+      "openai_api": "OpenAI API",
       "maixcam": "MaixCam",
       "matrix": "Matrix",
       "irc": "IRC"
@@ -277,6 +278,7 @@
     "form": {
       "desc": {
         "token": "机器人访问令牌，用于连接平台 API。",
+        "apiKey": "用于访问 OpenAI 兼容接口的鉴权密钥。",
         "botToken": "Bot Token，用于发送与接收消息。",
         "appToken": "App Token，用于 Socket 模式连接。",
         "appId": "应用唯一标识，用于平台鉴权。",