admin/ai-service
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Switch transcription to Whisper large v3
Some checks failed
CI / test (push) Failing after 10s
Details
Build and Deploy / build-and-deploy (push) Successful in 24s
Details

Browse Source
This commit is contained in:
Grendgi
2026-06-10 10:10:13 +03:00
parent 1b63dcdbf5
commit 8d6cd84403
12 changed files with 85 additions and 93 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
internal/config/config.go
View File

@@ -18,11 +18,11 @@ type Config struct {
LLMAPIKey string
LLMModel string
LLMTimeout time.Duration
VoxtralBaseURL string
VoxtralAPIKey string
VoxtralModel string
VoxtralTimeout time.Duration
AudioLLMPrompt string
AudioBaseURL string
AudioAPIKey string
AudioModel string
AudioTimeout time.Duration
AudioPrompt string
AIStatsSidecarURL string
AIStatsTimeout time.Duration
@@ -48,11 +48,11 @@ func Load() Config {
LLMAPIKey: envString("LLM_API_KEY", ""),
LLMModel: envString("LLM_MODEL", "qwen2.5-14b"),
LLMTimeout: envDuration("LLM_TIMEOUT", 5*time.Minute),
VoxtralBaseURL: envString("VOXTRAL_BASE_URL", envString("AUDIO_LLM_BASE_URL", "")),
VoxtralAPIKey: envString("VOXTRAL_API_KEY", envString("AUDIO_LLM_API_KEY", envString("LLM_API_KEY", ""))),
VoxtralModel: envString("VOXTRAL_MODEL", "mistralai/Voxtral-Small-24B-2507"),
VoxtralTimeout: envDuration("VOXTRAL_TIMEOUT", envDuration("AUDIO_LLM_TIMEOUT", 10*time.Minute)),
AudioLLMPrompt: envString("AUDIO_LLM_PROMPT", defaultAudioLLMPrompt()),
AudioBaseURL: envString("AUDIO_TRANSCRIPTION_BASE_URL", envString("AUDIO_LLM_BASE_URL", "")),
AudioAPIKey: envString("AUDIO_TRANSCRIPTION_API_KEY", envString("AUDIO_LLM_API_KEY", envString("LLM_API_KEY", ""))),
AudioModel: envString("AUDIO_TRANSCRIPTION_MODEL", "openai/whisper-large-v3"),
AudioTimeout: envDuration("AUDIO_TRANSCRIPTION_TIMEOUT", envDuration("AUDIO_LLM_TIMEOUT", 10*time.Minute)),
AudioPrompt: envString("AUDIO_TRANSCRIPTION_PROMPT", envString("AUDIO_LLM_PROMPT", defaultAudioPrompt())),
AIStatsSidecarURL: envString("AI_STATS_SIDECAR_URL", ""),
AIStatsTimeout: envDuration("AI_STATS_TIMEOUT", 8*time.Second),
@@ -132,7 +132,7 @@ func envCSVDefault(key string, fallback []string) []string {
return fallback
}
func defaultAudioLLMPrompt() string {
func defaultAudioPrompt() string {
return "Расшифруй речь из аудио максимально точно. Сохрани русский язык, имена, телефоны, суммы и смысловые паузы. Не добавляй комментарии, анализ, Markdown или JSON. Верни только чистый текст расшифровки."
}

3
internal/httpapi/dashboard.go
View File

@@ -5,6 +5,7 @@ import (
"time"
"ai-service/internal/model"
"ai-service/internal/transcription"
)
type dashboardResponse struct {
@@ -51,7 +52,7 @@ func (s *Server) handleDashboard(w http.ResponseWriter, r *http.Request) {
At: now,
Providers: []providerStatus{
s.checkLLM(ctx),
s.checkAudioLLM(ctx, "voxtral-small", s.cfg.VoxtralBaseURL, s.cfg.VoxtralAPIKey, s.cfg.VoxtralModel, s.cfg.VoxtralTimeout),
s.checkAudioLLM(ctx, transcription.ProviderWhisperLargeV3, s.cfg.AudioBaseURL, s.cfg.AudioAPIKey, s.cfg.AudioModel, s.cfg.AudioTimeout),
},
},
Infra: loadInfraSnapshot(r, s.cfg),

4
internal/httpapi/providers.go
View File

@@ -8,6 +8,8 @@ import (
"net/http"
"strings"
"time"
"ai-service/internal/transcription"
)
type providerStatus struct {
@@ -42,7 +44,7 @@ func (s *Server) handleProviderStatus(w http.ResponseWriter, r *http.Request) {
At: time.Now().UTC(),
Providers: []providerStatus{
s.checkLLM(ctx),
s.checkAudioLLM(ctx, "voxtral-small", s.cfg.VoxtralBaseURL, s.cfg.VoxtralAPIKey, s.cfg.VoxtralModel, s.cfg.VoxtralTimeout),
s.checkAudioLLM(ctx, transcription.ProviderWhisperLargeV3, s.cfg.AudioBaseURL, s.cfg.AudioAPIKey, s.cfg.AudioModel, s.cfg.AudioTimeout),
},
}
writeJSON(w, http.StatusOK, resp)

41
internal/transcription/client.go
View File

@@ -19,16 +19,19 @@ type Client struct {
http *http.Client
}
const ProviderVoxtral = "voxtral-small"
const (
ProviderWhisperLargeV3 = "whisper-large-v3"
defaultWhisperModel = "openai/whisper-large-v3"
)
var speakerLabelPattern = regexp.MustCompile(`(?i)(?:^|[\n\r ]+)((?:speaker|спикер|говорящий)\s*\d+)\s*[:：-]`)
type Options struct {
VoxtralBaseURL string
VoxtralAPIKey string
VoxtralModel string
VoxtralTimeout time.Duration
AudioLLMPrompt string
AudioBaseURL string
AudioAPIKey string
AudioModel string
AudioTimeout time.Duration
AudioPrompt string
}
type ProviderConfig struct {
@@ -102,17 +105,17 @@ type audioTranscriptionSegment struct {
func New(baseURL string, timeout time.Duration, ffmpegPath string, leadSilence time.Duration) *Client {
return NewWithOptions(Options{
VoxtralBaseURL: baseURL,
VoxtralTimeout: timeout,
AudioBaseURL: baseURL,
AudioTimeout: timeout,
})
}
func NewWithOptions(opts Options) *Client {
audioLLMPrompt := strings.TrimSpace(opts.AudioLLMPrompt)
if audioLLMPrompt == "" {
audioLLMPrompt = "Transcribe the audio exactly. Return only the transcript text."
audioPrompt := strings.TrimSpace(opts.AudioPrompt)
if audioPrompt == "" {
audioPrompt = "Transcribe the audio exactly. Return only the transcript text."
}
provider := buildVoxtralProvider(opts, audioLLMPrompt)
provider := buildAudioProvider(opts, audioPrompt)
if provider.BaseURL == "" {
return nil
}
@@ -122,18 +125,18 @@ func NewWithOptions(opts Options) *Client {
}
}
func buildVoxtralProvider(opts Options, prompt string) ProviderConfig {
baseURL := strings.TrimRight(strings.TrimSpace(opts.VoxtralBaseURL), "/")
func buildAudioProvider(opts Options, prompt string) ProviderConfig {
baseURL := strings.TrimRight(strings.TrimSpace(opts.AudioBaseURL), "/")
if baseURL == "" {
return ProviderConfig{}
}
model := firstNonEmpty(opts.VoxtralModel, "mistralai/Voxtral-Small-24B-2507")
model := firstNonEmpty(opts.AudioModel, defaultWhisperModel)
return ProviderConfig{
Name: ProviderVoxtral,
Name: ProviderWhisperLargeV3,
BaseURL: baseURL,
APIKey: strings.TrimSpace(opts.VoxtralAPIKey),
APIKey: strings.TrimSpace(opts.AudioAPIKey),
Model: model,
Timeout: defaultDuration(opts.VoxtralTimeout, 10*time.Minute),
Timeout: defaultDuration(opts.AudioTimeout, 10*time.Minute),
Prompt: prompt,
}
}
@@ -147,7 +150,7 @@ func defaultDuration(v, fallback time.Duration) time.Duration {
func (c *Client) Transcribe(ctx context.Context, in Input) (*Result, error) {
if c == nil || c.provider.BaseURL == "" {
return nil, fmt.Errorf("voxtral transcription provider not configured")
return nil, fmt.Errorf("audio transcription provider not configured")
}
if strings.TrimSpace(in.AudioURL) == "" {
return nil, fmt.Errorf("audio_url is required")

18
internal/transcription/client_test.go
View File

@@ -7,22 +7,22 @@ import (
"testing"
)
func TestNewWithOptionsBuildsVoxtralProvider(t *testing.T) {
func TestNewWithOptionsBuildsWhisperProvider(t *testing.T) {
client := NewWithOptions(Options{
VoxtralBaseURL: "http://voxtral",
AudioBaseURL: "http://whisper",
})
if client == nil {
t.Fatal("client is nil")
}
if client.provider.Name != ProviderVoxtral {
t.Fatalf("provider = %q, want %q", client.provider.Name, ProviderVoxtral)
if client.provider.Name != ProviderWhisperLargeV3 {
t.Fatalf("provider = %q, want %q", client.provider.Name, ProviderWhisperLargeV3)
}
if client.provider.Model != "mistralai/Voxtral-Small-24B-2507" {
if client.provider.Model != "openai/whisper-large-v3" {
t.Fatalf("model = %q", client.provider.Model)
}
}
func TestVoxtralUsesAudioTranscriptionsEndpoint(t *testing.T) {
func TestWhisperUsesAudioTranscriptionsEndpoint(t *testing.T) {
audioSrv := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
_, _ = w.Write([]byte("fake audio"))
}))
@@ -50,8 +50,8 @@ func TestVoxtralUsesAudioTranscriptionsEndpoint(t *testing.T) {
defer providerSrv.Close()
client := NewWithOptions(Options{
VoxtralBaseURL: providerSrv.URL,
VoxtralModel: "mistralai/Voxtral-Small-24B-2507",
AudioBaseURL: providerSrv.URL,
AudioModel: "openai/whisper-large-v3",
})
if client == nil {
t.Fatal("client is nil")
@@ -63,7 +63,7 @@ func TestVoxtralUsesAudioTranscriptionsEndpoint(t *testing.T) {
if gotPath != "/v1/audio/transcriptions" {
t.Fatalf("path = %q, want /v1/audio/transcriptions", gotPath)
}
if gotModel != "mistralai/Voxtral-Small-24B-2507" {
if gotModel != "openai/whisper-large-v3" {
t.Fatalf("model = %q", gotModel)
}
if gotResponseFormat != "json" {

2
internal/worker/worker.go
View File

@@ -20,7 +20,7 @@ const (
TaskCallAnalysis = "call_analysis"
TaskTranscription = "transcription"
TranscriptionProfile = "voxtral-small"
TranscriptionProfile = "whisper-large-v3"
)
type Worker struct {