package callflow import ( "context" "testing" "time" "git.ipao.vip/rogee/go-sip/contracts" "git.ipao.vip/rogee/go-sip/internal/ai" "git.ipao.vip/rogee/go-sip/internal/media" ) func TestInitialMediaWaitIsBounded(t *testing.T) { raw, err := contracts.Read("examples/agent-version-full-explicit.json") if err != nil { t.Fatal(err) } snapshot, err := ai.ValidateForMode(raw, ai.ModeFullAI) if err != nil { t.Fatal(err) } _, err = Execute(context.Background(), NewMemorySession(nil), ai.MockPipeline{}, snapshot, "测试开场", time.Millisecond) if err == nil { t.Fatal("expected bounded initial media wait to fail") } } func TestSharedFlowRunsThreeConversationTurns(t *testing.T) { raw, err := contracts.Read("examples/agent-version-full-explicit.json") if err != nil { t.Fatal(err) } snapshot, err := ai.ValidateForMode(raw, ai.ModeFullAI) if err != nil { t.Fatal(err) } session := &scriptedTurnSession{turns: [][]byte{make([]byte, 6400), make([]byte, 6400), make([]byte, 6400)}} result, err := ExecuteWithCapture(context.Background(), session, ai.MockPipeline{MaxAudioBytes: 16 << 20}, snapshot, "测试开场", CaptureConfig{ FirstSpeechTimeout: time.Second, MaxDuration: 2 * time.Millisecond, MaxTurns: 3, }) if err != nil { t.Fatal(err) } if len(result.Turns) != 3 || len(result.InboundTurns) != 3 || len(result.OutboundTurns) != 4 { t.Fatalf("expected opening plus three shared turns, got turns=%d inbound=%d outbound=%d", len(result.Turns), len(result.InboundTurns), len(result.OutboundTurns)) } if result.Turn.Transcript == "" || result.Turn.Reply == "" || result.RTP.ReceivedBytes != 19200 { t.Fatalf("three-turn flow facts are incomplete: %+v", result) } } type scriptedTurnSession struct { turns [][]byte index int served bool stats media.RTPStats } func (s *scriptedTurnSession) ReadPayload(ctx context.Context) ([]byte, error) { if !s.served && s.index < len(s.turns) { s.served = true payload := append([]byte(nil), s.turns[s.index]...) s.index++ s.stats.ReceivedPackets++ s.stats.ReceivedBytes += uint64(len(payload)) return payload, nil } <-ctx.Done() s.served = false return nil, ctx.Err() } func (s *scriptedTurnSession) SendPCM16(ctx context.Context, pcm []byte, _ int) error { if err := ctx.Err(); err != nil { return err } s.stats.SentPackets++ s.stats.SentBytes += uint64(len(pcm)) return nil } func (s *scriptedTurnSession) Stats() media.RTPStats { return s.stats } type invalidCallPipeline struct{} func (invalidCallPipeline) Synthesize(context.Context, ai.Snapshot, string) ([]byte, error) { return make([]byte, 6400), nil } func (invalidCallPipeline) RunTurn(context.Context, ai.Snapshot, []byte) (ai.TurnResult, error) { return ai.TurnResult{Transcript: "打错了", Reply: ai.InvalidCallMarker, AudioPCM16: make([]byte, 6400)}, nil } func TestInvalidCallStopsBeforeReply(t *testing.T) { raw, err := contracts.Read("examples/agent-version-full-explicit.json") if err != nil { t.Fatal(err) } snapshot, err := ai.ValidateForMode(raw, ai.ModeFullAI) if err != nil { t.Fatal(err) } session := &scriptedTurnSession{turns: [][]byte{make([]byte, 6400)}} result, err := ExecuteWithCapture(context.Background(), session, invalidCallPipeline{}, snapshot, "开场", CaptureConfig{ FirstSpeechTimeout: time.Second, MaxDuration: 2 * time.Millisecond, MaxTurns: 3, }) if err != nil { t.Fatal(err) } if !result.Turn.InvalidCall || result.Turn.InvalidReason != "llm_invalid_call_marker" || len(result.Turns) != 1 { t.Fatalf("invalid call was not classified: %+v", result) } if session.stats.SentPackets != 1 { t.Fatalf("invalid call must not send a reply, sent packets=%d", session.stats.SentPackets) } } func TestASROnlySkipsOpeningLLMAndTTS(t *testing.T) { raw, err := contracts.Read("examples/agent-version-asr-only.json") if err != nil { t.Fatal(err) } snapshot, err := ai.ValidateForMode(raw, ai.ModeASROnly) if err != nil { t.Fatal(err) } pipeline := &asrOnlyPipeline{} session := &scriptedTurnSession{turns: [][]byte{make([]byte, 6400)}} result, err := ExecuteWithCapture(context.Background(), session, pipeline, snapshot, "不得播放", CaptureConfig{ FirstSpeechTimeout: time.Second, MaxDuration: time.Millisecond, MaxTurns: 1, }) if err != nil { t.Fatal(err) } if pipeline.synthesizeCalls != 0 || pipeline.turnCalls != 1 { t.Fatalf("ASR-only provider calls = synthesize:%d turn:%d", pipeline.synthesizeCalls, pipeline.turnCalls) } if result.Turn.Transcript != "asr-only transcript" || result.Turn.Reply != "" || len(result.OutboundTurns) != 0 { t.Fatalf("unexpected ASR-only result: %+v", result) } if session.stats.SentPackets != 0 { t.Fatalf("ASR-only flow must not send opening/reply audio, sent packets=%d", session.stats.SentPackets) } } type asrOnlyPipeline struct { synthesizeCalls int turnCalls int } func (p *asrOnlyPipeline) Synthesize(context.Context, ai.Snapshot, string) ([]byte, error) { p.synthesizeCalls++ return nil, nil } func (p *asrOnlyPipeline) RunTurn(context.Context, ai.Snapshot, []byte) (ai.TurnResult, error) { p.turnCalls++ return ai.TurnResult{Transcript: "asr-only transcript"}, nil } func TestMockModeUsesTheSharedConversationFlow(t *testing.T) { raw, err := contracts.Read("examples/agent-version-full-explicit.json") if err != nil { t.Fatal(err) } snapshot, err := ai.ValidateForMode(raw, ai.ModeFullAI) if err != nil { t.Fatal(err) } session := NewMemorySession(make([]byte, 6400)) result, err := Execute(context.Background(), session, ai.MockPipeline{MaxAudioBytes: 16 << 20}, snapshot, "测试开场", 10*time.Millisecond) if err != nil { t.Fatal(err) } if result.Turn.Transcript == "" || result.Turn.Reply == "" || len(result.Turn.AudioPCM16) == 0 { t.Fatalf("shared flow did not complete mock ASR/LLM/TTS: %+v", result.Turn) } if result.RTP.ReceivedBytes != 6400 || result.RTP.SentPackets == 0 || len(session.OutboundPCM()) == 0 { t.Fatalf("shared media boundary was not exercised: %+v", result.RTP) } }