story/internal/mockprovider/server.go

package mockprovider

import (
	"bytes"
	"context"
	"crypto/hmac"
	"crypto/sha256"
	"encoding/hex"
	"encoding/json"
	"errors"
	"io"
	"log/slog"
	"net/http"
	"sync"
	"time"

	"data-terminal-cloud/internal/voiceprovider"
)

type Options struct {
	APIKey             string
	CallbackURL        string
	CallbackSecret     string
	StartDelay         time.Duration
	DuplicateCallbacks int
}

type task struct {
	Request voiceprovider.StartRequest
	Status  string
}

type Server struct {
	options    Options
	logger     *slog.Logger
	client     *http.Client
	mu         sync.Mutex
	tasks      map[string]*task
	available  bool
	startCalls int
	stopCalls  int
	now        func() time.Time
}

func New(options Options, logger *slog.Logger) (*Server, error) {
	if len(options.APIKey) < 16 || options.CallbackURL == "" || len(options.CallbackSecret) < 32 || options.StartDelay < 0 || options.DuplicateCallbacks < 1 {
		return nil, errors.New("valid mock provider options are required")
	}
	if logger == nil {
		logger = slog.Default()
	}
	return &Server{options: options, logger: logger, client: &http.Client{Timeout: 5 * time.Second}, tasks: map[string]*task{}, available: true, now: time.Now}, nil
}

func (s *Server) Handler() http.Handler {
	mux := http.NewServeMux()
	mux.HandleFunc("GET /health/live", func(writer http.ResponseWriter, _ *http.Request) {
		writer.Header().Set("Content-Type", "application/json")
		_, _ = writer.Write([]byte(`{"status":"alive","service":"mock-provider"}`))
	})
	mux.HandleFunc("POST /mock/v1/voice-chats:start", s.start)
	mux.HandleFunc("POST /mock/v1/voice-chats:query", s.query)
	mux.HandleFunc("POST /mock/v1/voice-chats:stop", s.stop)
	mux.HandleFunc("GET /mock/v1/admin/state", s.adminState)
	mux.HandleFunc("POST /mock/v1/admin/availability", s.adminAvailability)
	mux.HandleFunc("POST /mock/v1/admin/reset", s.adminReset)
	mux.HandleFunc("POST /mock/v1/admin/replay", s.adminReplay)
	return mux
}

func (s *Server) query(writer http.ResponseWriter, request *http.Request) {
	if !s.authorized(request) {
		http.Error(writer, "unauthorized", http.StatusUnauthorized)
		return
	}
	var input voiceprovider.QueryRequest
	decoder := json.NewDecoder(io.LimitReader(request.Body, 1<<20))
	decoder.DisallowUnknownFields()
	if err := decoder.Decode(&input); err != nil || input.TaskRef == "" || input.ConversationID == "" {
		http.Error(writer, "invalid request", http.StatusBadRequest)
		return
	}
	s.mu.Lock()
	available := s.available
	existing := s.tasks[input.TaskRef]
	s.mu.Unlock()
	if !available {
		http.Error(writer, "provider temporarily unavailable", http.StatusServiceUnavailable)
		return
	}
	if existing == nil {
		http.Error(writer, "task not found", http.StatusNotFound)
		return
	}
	if existing.Request.ConversationID != input.ConversationID {
		http.Error(writer, "task reference conflict", http.StatusConflict)
		return
	}
	writer.Header().Set("Content-Type", "application/json")
	_ = json.NewEncoder(writer).Encode(voiceprovider.QueryResponse{TaskRef: input.TaskRef, Status: existing.Status})
}

func (s *Server) start(writer http.ResponseWriter, request *http.Request) {
	if !hmac.Equal([]byte(request.Header.Get("X-Mock-Provider-Key")), []byte(s.options.APIKey)) {
		http.Error(writer, "unauthorized", http.StatusUnauthorized)
		return
	}
	s.mu.Lock()
	s.startCalls++
	available := s.available
	s.mu.Unlock()
	if !available {
		http.Error(writer, "provider temporarily unavailable", http.StatusServiceUnavailable)
		return
	}
	decoder := json.NewDecoder(io.LimitReader(request.Body, 1<<20))
	decoder.DisallowUnknownFields()
	var input voiceprovider.StartRequest
	if err := decoder.Decode(&input); err != nil || input.TaskRef == "" || input.RoomID == "" || input.DeviceUserID == "" || input.AgentUserID == "" || input.ConversationID == "" || !supportedEngineMode(input.EngineMode) {
		http.Error(writer, "invalid request", http.StatusBadRequest)
		return
	}
	s.mu.Lock()
	existing, found := s.tasks[input.TaskRef]
	if found && existing.Request != input {
		s.mu.Unlock()
		http.Error(writer, "task reference conflict", http.StatusConflict)
		return
	}
	if !found {
		existing = &task{Request: input, Status: "STARTING"}
		s.tasks[input.TaskRef] = existing
	}
	status := existing.Status
	s.mu.Unlock()
	if !found {
		go s.completeStart(input)
	}
	writer.Header().Set("Content-Type", "application/json")
	writer.WriteHeader(http.StatusAccepted)
	_ = json.NewEncoder(writer).Encode(voiceprovider.StartResponse{TaskRef: input.TaskRef, Status: status})
}

func supportedEngineMode(mode string) bool {
	return mode == "s2s_pure" || mode == "s2s_hybrid" || mode == "cascade"
}

func (s *Server) completeStart(input voiceprovider.StartRequest) {
	timer := time.NewTimer(s.options.StartDelay)
	defer timer.Stop()
	<-timer.C
	s.mu.Lock()
	if existing := s.tasks[input.TaskRef]; existing != nil {
		if existing.Status != "STARTING" {
			s.mu.Unlock()
			return
		}
		existing.Status = "ACTIVE"
	}
	s.mu.Unlock()
	digest := sha256.Sum256([]byte(input.TaskRef))
	event := voiceprovider.CallbackEvent{
		Provider: "mock", EventID: "mock-started-" + hex.EncodeToString(digest[:12]),
		TaskRef: input.TaskRef, ConversationID: input.ConversationID,
		EventType: "voice_chat.started", Status: "ACTIVE", OccurredAt: s.now().UTC(),
	}
	body, err := json.Marshal(event)
	if err != nil {
		return
	}
	for index := 0; index < s.options.DuplicateCallbacks; index++ {
		if err := s.sendCallback(body); err != nil {
			s.logger.Warn("mock provider callback failed", "task_ref", input.TaskRef, "attempt", index+1, "error", err)
		}
	}
}

func (s *Server) stop(writer http.ResponseWriter, request *http.Request) {
	if !hmac.Equal([]byte(request.Header.Get("X-Mock-Provider-Key")), []byte(s.options.APIKey)) {
		http.Error(writer, "unauthorized", http.StatusUnauthorized)
		return
	}
	s.mu.Lock()
	s.stopCalls++
	available := s.available
	s.mu.Unlock()
	if !available {
		http.Error(writer, "provider temporarily unavailable", http.StatusServiceUnavailable)
		return
	}
	decoder := json.NewDecoder(io.LimitReader(request.Body, 1<<20))
	decoder.DisallowUnknownFields()
	var input voiceprovider.StopRequest
	if err := decoder.Decode(&input); err != nil || input.TaskRef == "" || input.ConversationID == "" || input.Reason == "" {
		http.Error(writer, "invalid request", http.StatusBadRequest)
		return
	}
	s.mu.Lock()
	existing, found := s.tasks[input.TaskRef]
	if !found {
		s.mu.Unlock()
		http.Error(writer, "task not found", http.StatusNotFound)
		return
	}
	if existing.Request.ConversationID != input.ConversationID {
		s.mu.Unlock()
		http.Error(writer, "task reference conflict", http.StatusConflict)
		return
	}
	shouldComplete := existing.Status != "STOPPING" && existing.Status != "COMPLETED"
	if shouldComplete {
		existing.Status = "STOPPING"
	}
	status := existing.Status
	s.mu.Unlock()
	if shouldComplete {
		go s.completeStop(existing.Request)
	}
	writer.Header().Set("Content-Type", "application/json")
	writer.WriteHeader(http.StatusAccepted)
	_ = json.NewEncoder(writer).Encode(voiceprovider.StopResponse{TaskRef: input.TaskRef, Status: status})
}

func (s *Server) adminState(writer http.ResponseWriter, request *http.Request) {
	if !s.authorized(request) {
		http.Error(writer, "unauthorized", http.StatusUnauthorized)
		return
	}
	s.mu.Lock()
	state := struct {
		Available  bool `json:"available"`
		StartCalls int  `json:"start_calls"`
		StopCalls  int  `json:"stop_calls"`
		Tasks      int  `json:"tasks"`
	}{Available: s.available, StartCalls: s.startCalls, StopCalls: s.stopCalls, Tasks: len(s.tasks)}
	s.mu.Unlock()
	writer.Header().Set("Content-Type", "application/json")
	_ = json.NewEncoder(writer).Encode(state)
}

func (s *Server) adminAvailability(writer http.ResponseWriter, request *http.Request) {
	if !s.authorized(request) {
		http.Error(writer, "unauthorized", http.StatusUnauthorized)
		return
	}
	var input struct {
		Available *bool `json:"available"`
	}
	decoder := json.NewDecoder(io.LimitReader(request.Body, 1024))
	decoder.DisallowUnknownFields()
	if err := decoder.Decode(&input); err != nil || input.Available == nil {
		http.Error(writer, "invalid request", http.StatusBadRequest)
		return
	}
	s.mu.Lock()
	s.available = *input.Available
	s.mu.Unlock()
	writer.Header().Set("Content-Type", "application/json")
	_ = json.NewEncoder(writer).Encode(map[string]bool{"available": *input.Available})
}

func (s *Server) adminReset(writer http.ResponseWriter, request *http.Request) {
	if !s.authorized(request) {
		http.Error(writer, "unauthorized", http.StatusUnauthorized)
		return
	}
	s.mu.Lock()
	s.available = true
	s.startCalls = 0
	s.stopCalls = 0
	s.tasks = map[string]*task{}
	s.mu.Unlock()
	writer.WriteHeader(http.StatusNoContent)
}

func (s *Server) adminReplay(writer http.ResponseWriter, request *http.Request) {
	if !s.authorized(request) {
		http.Error(writer, "unauthorized", http.StatusUnauthorized)
		return
	}
	var input struct {
		TaskRef        string `json:"task_ref"`
		Sample         string `json:"sample"`
		DuplicateCount int    `json:"duplicate_count"`
	}
	decoder := json.NewDecoder(io.LimitReader(request.Body, 1<<20))
	decoder.DisallowUnknownFields()
	if err := decoder.Decode(&input); err != nil || input.TaskRef == "" || input.Sample == "" || input.DuplicateCount < 0 || input.DuplicateCount > 10 {
		http.Error(writer, "invalid request", http.StatusBadRequest)
		return
	}
	if input.DuplicateCount == 0 {
		input.DuplicateCount = 1
	}
	s.mu.Lock()
	existing := s.tasks[input.TaskRef]
	s.mu.Unlock()
	if existing == nil {
		http.Error(writer, "task not found", http.StatusNotFound)
		return
	}
	event, err := s.sampleEvent(existing.Request, input.Sample)
	if err != nil {
		http.Error(writer, err.Error(), http.StatusBadRequest)
		return
	}
	body, err := json.Marshal(event)
	if err != nil {
		http.Error(writer, "encode sample", http.StatusInternalServerError)
		return
	}
	for attempt := 0; attempt < input.DuplicateCount; attempt++ {
		if err := s.sendCallback(body); err != nil {
			http.Error(writer, "callback delivery failed", http.StatusBadGateway)
			return
		}
	}
	writer.Header().Set("Content-Type", "application/json")
	writer.WriteHeader(http.StatusAccepted)
	_ = json.NewEncoder(writer).Encode(map[string]any{"sample": input.Sample, "deliveries": input.DuplicateCount, "event_id": event.EventID})
}

func (s *Server) sampleEvent(input voiceprovider.StartRequest, sample string) (voiceprovider.CallbackEvent, error) {
	now := s.now().UTC()
	digest := sha256.Sum256([]byte("sample:" + sample + ":" + input.TaskRef))
	event := voiceprovider.CallbackEvent{
		Provider: "mock", EventID: "mock-sample-" + sample + "-" + hex.EncodeToString(digest[:8]),
		TaskRef: input.TaskRef, ConversationID: input.ConversationID, OccurredAt: now,
	}
	switch sample {
	case "started":
		event.EventType, event.Status = "voice_chat.started", "ACTIVE"
	case "final_transcript":
		event.EventType, event.Status = "voice_chat.transcript.final", "FINAL"
		event.Turns = []voiceprovider.FinalTurn{{TurnKey: "sample-subject-101", SequenceNo: 101, Speaker: "subject", StartAt: now.Add(-3 * time.Second), EndAt: now.Add(-2 * time.Second), FinalText: "这是脱敏的长者最终字幕样例。", ProviderRef: "sample-subject-101"}}
	case "agent_answering":
		event.EventType, event.Status = "voice_chat.transcript.final", "FINAL"
		event.Turns = []voiceprovider.FinalTurn{{TurnKey: "sample-agent-102", SequenceNo: 102, Speaker: "agent", StartAt: now.Add(-2 * time.Second), EndAt: now.Add(-time.Second), FinalText: "这是智能体开始回答的脱敏样例。", ProviderRef: "sample-agent-102"}}
	case "interrupted":
		event.EventType, event.Status = "voice_chat.transcript.final", "FINAL"
		event.Turns = []voiceprovider.FinalTurn{{TurnKey: "sample-agent-103", SequenceNo: 103, Speaker: "agent", StartAt: now.Add(-time.Second), EndAt: now, FinalText: "这是被打断的智能体回答样例。", Interrupted: true, ProviderRef: "sample-agent-103"}}
	case "usage":
		event.EventType, event.Status = "voice_chat.usage", "FINAL"
		event.Usage = []voiceprovider.UsageMetric{{UsageCategory: "RTC", MetricCode: "rtc.minutes", Quantity: "1", Unit: "minutes"}}
	case "error":
		event.EventType, event.Status = "voice_chat.failed", "FAILED"
	case "stopped":
		event.EventType, event.Status = "voice_chat.stopped", "COMPLETED"
	default:
		return voiceprovider.CallbackEvent{}, errors.New("unknown sample")
	}
	return event, nil
}

func (s *Server) authorized(request *http.Request) bool {
	return hmac.Equal([]byte(request.Header.Get("X-Mock-Provider-Key")), []byte(s.options.APIKey))
}

func (s *Server) completeStop(input voiceprovider.StartRequest) {
	timer := time.NewTimer(s.options.StartDelay)
	defer timer.Stop()
	<-timer.C
	s.mu.Lock()
	if existing := s.tasks[input.TaskRef]; existing != nil {
		existing.Status = "COMPLETED"
	}
	s.mu.Unlock()
	s.sendFinalTranscript(input)
	digest := sha256.Sum256([]byte("stop:" + input.TaskRef))
	event := voiceprovider.CallbackEvent{
		Provider: "mock", EventID: "mock-stopped-" + hex.EncodeToString(digest[:12]),
		TaskRef: input.TaskRef, ConversationID: input.ConversationID,
		EventType: "voice_chat.stopped", Status: "COMPLETED", OccurredAt: s.now().UTC(),
	}
	body, err := json.Marshal(event)
	if err != nil {
		return
	}
	for index := 0; index < s.options.DuplicateCallbacks; index++ {
		if err := s.sendCallback(body); err != nil {
			s.logger.Warn("mock provider stop callback failed", "task_ref", input.TaskRef, "attempt", index+1, "error", err)
		}
	}
}

func (s *Server) sendFinalTranscript(input voiceprovider.StartRequest) {
	base := s.now().UTC().Add(-20 * time.Second)
	digest := sha256.Sum256([]byte("transcript:" + input.TaskRef))
	event := voiceprovider.CallbackEvent{
		Provider: "mock", EventID: "mock-transcript-" + hex.EncodeToString(digest[:12]),
		TaskRef: input.TaskRef, ConversationID: input.ConversationID,
		EventType: "voice_chat.transcript.final", Status: "FINAL", OccurredAt: s.now().UTC(),
		Turns: []voiceprovider.FinalTurn{
			{TurnKey: "turn-1", SequenceNo: 1, Speaker: "subject", StartAt: base, EndAt: base.Add(6 * time.Second), FinalText: "我年轻时常和姐姐在家乡河边散步。有一次春天突然下雨，我们躲在老柳树下，后来一路笑着跑回家，现在想起来还是很高兴。", ProviderRef: "mock-turn-1"},
			{TurnKey: "turn-2", SequenceNo: 2, Speaker: "agent", StartAt: base.Add(7 * time.Second), EndAt: base.Add(13 * time.Second), FinalText: "听起来那是一段很温暖的回忆，下次我们可以继续聊聊家乡的春天。", ProviderRef: "mock-turn-2"},
		},
	}
	body, err := json.Marshal(event)
	if err != nil {
		return
	}
	for index := 0; index < s.options.DuplicateCallbacks; index++ {
		if err := s.sendCallback(body); err != nil {
			s.logger.Warn("mock provider transcript callback failed", "task_ref", input.TaskRef, "attempt", index+1, "error", err)
		}
	}
}

func (s *Server) sendCallback(body []byte) error {
	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
	defer cancel()
	request, err := http.NewRequestWithContext(ctx, http.MethodPost, s.options.CallbackURL, bytes.NewReader(body))
	if err != nil {
		return err
	}
	timestamp := s.now().UTC().Format(time.RFC3339)
	request.Header.Set("Content-Type", "application/json")
	request.Header.Set(voiceprovider.HeaderCallbackTimestamp, timestamp)
	request.Header.Set(voiceprovider.HeaderCallbackSignature,
		voiceprovider.SignCallback([]byte(s.options.CallbackSecret), timestamp, body))
	response, err := s.client.Do(request)
	if err != nil {
		return err
	}
	defer response.Body.Close()
	_, _ = io.Copy(io.Discard, io.LimitReader(response.Body, 1<<20))
	if response.StatusCode < 200 || response.StatusCode >= 300 {
		return errors.New("callback endpoint rejected event")
	}
	return nil
}