|
| 1 | +package ai |
| 2 | + |
| 3 | +import ( |
| 4 | + "context" |
| 5 | + "io" |
| 6 | + "mime" |
| 7 | + "net/url" |
| 8 | + "strings" |
| 9 | + |
| 10 | + "github.com/pkg/errors" |
| 11 | + "google.golang.org/genai" |
| 12 | +) |
| 13 | + |
| 14 | +const ( |
| 15 | + defaultGeminiEndpoint = "https://generativelanguage.googleapis.com/v1beta" |
| 16 | + geminiTranscriptionPrompt = `Transcribe the audio accurately. Return only the transcript text. Do not summarize, explain, or add content that is not spoken.` |
| 17 | + maxGeminiInlineAudioSize = 14 * 1024 * 1024 |
| 18 | + defaultGeminiAPIVersion = "v1beta" |
| 19 | + geminiProviderDisplayName = "Gemini" |
| 20 | + geminiDefaultTemperature = float32(0) |
| 21 | +) |
| 22 | + |
| 23 | +var geminiSupportedContentTypes = map[string]string{ |
| 24 | + "audio/wav": "audio/wav", |
| 25 | + "audio/x-wav": "audio/wav", |
| 26 | + "audio/mp3": "audio/mp3", |
| 27 | + "audio/mpeg": "audio/mp3", |
| 28 | + "audio/aiff": "audio/aiff", |
| 29 | + "audio/aac": "audio/aac", |
| 30 | + "audio/ogg": "audio/ogg", |
| 31 | + "audio/flac": "audio/flac", |
| 32 | + "audio/x-flac": "audio/flac", |
| 33 | +} |
| 34 | + |
| 35 | +type geminiTranscriber struct { |
| 36 | + client *genai.Client |
| 37 | +} |
| 38 | + |
| 39 | +func newGeminiTranscriber(config ProviderConfig, options transcriberOptions) (*geminiTranscriber, error) { |
| 40 | + endpoint, err := normalizeEndpoint(config.Endpoint, defaultGeminiEndpoint, geminiProviderDisplayName) |
| 41 | + if err != nil { |
| 42 | + return nil, err |
| 43 | + } |
| 44 | + if err := requireAPIKey(config.APIKey, geminiProviderDisplayName); err != nil { |
| 45 | + return nil, err |
| 46 | + } |
| 47 | + baseURL, apiVersion, err := normalizeGeminiEndpoint(endpoint) |
| 48 | + if err != nil { |
| 49 | + return nil, err |
| 50 | + } |
| 51 | + httpOptions := genai.HTTPOptions{ |
| 52 | + BaseURL: baseURL, |
| 53 | + APIVersion: apiVersion, |
| 54 | + } |
| 55 | + if options.httpClient.Timeout > 0 { |
| 56 | + timeout := options.httpClient.Timeout |
| 57 | + httpOptions.Timeout = &timeout |
| 58 | + } |
| 59 | + |
| 60 | + client, err := genai.NewClient(context.Background(), &genai.ClientConfig{ |
| 61 | + APIKey: config.APIKey, |
| 62 | + Backend: genai.BackendGeminiAPI, |
| 63 | + HTTPClient: options.httpClient, |
| 64 | + HTTPOptions: httpOptions, |
| 65 | + }) |
| 66 | + if err != nil { |
| 67 | + return nil, errors.Wrap(err, "failed to create Gemini client") |
| 68 | + } |
| 69 | + return &geminiTranscriber{client: client}, nil |
| 70 | +} |
| 71 | + |
| 72 | +// Transcribe transcribes audio with Gemini generateContent. |
| 73 | +func (t *geminiTranscriber) Transcribe(ctx context.Context, request TranscribeRequest) (*TranscribeResponse, error) { |
| 74 | + if strings.TrimSpace(request.Model) == "" { |
| 75 | + return nil, errors.New("model is required") |
| 76 | + } |
| 77 | + if request.Audio == nil { |
| 78 | + return nil, errors.New("audio is required") |
| 79 | + } |
| 80 | + audio, err := io.ReadAll(request.Audio) |
| 81 | + if err != nil { |
| 82 | + return nil, errors.Wrap(err, "failed to read audio") |
| 83 | + } |
| 84 | + if len(audio) == 0 { |
| 85 | + return nil, errors.New("audio is required") |
| 86 | + } |
| 87 | + if len(audio) > maxGeminiInlineAudioSize { |
| 88 | + return nil, errors.Errorf("audio is too large for Gemini inline transcription; maximum size is %d bytes", maxGeminiInlineAudioSize) |
| 89 | + } |
| 90 | + |
| 91 | + contentType, err := normalizeGeminiContentType(request.ContentType) |
| 92 | + if err != nil { |
| 93 | + return nil, err |
| 94 | + } |
| 95 | + prompt := buildGeminiTranscriptionPrompt(request.Prompt, request.Language) |
| 96 | + temperature := geminiDefaultTemperature |
| 97 | + response, err := t.client.Models.GenerateContent(ctx, normalizeGeminiModelName(request.Model), []*genai.Content{ |
| 98 | + genai.NewContentFromParts([]*genai.Part{ |
| 99 | + genai.NewPartFromBytes(audio, contentType), |
| 100 | + genai.NewPartFromText(prompt), |
| 101 | + }, genai.RoleUser), |
| 102 | + }, &genai.GenerateContentConfig{ |
| 103 | + Temperature: &temperature, |
| 104 | + }) |
| 105 | + if err != nil { |
| 106 | + return nil, errors.Wrap(err, "failed to send Gemini transcription request") |
| 107 | + } |
| 108 | + text := strings.TrimSpace(response.Text()) |
| 109 | + if text == "" { |
| 110 | + return nil, errors.New("Gemini transcription response did not include text") |
| 111 | + } |
| 112 | + return &TranscribeResponse{ |
| 113 | + Text: text, |
| 114 | + }, nil |
| 115 | +} |
| 116 | + |
| 117 | +func normalizeGeminiEndpoint(endpoint string) (string, string, error) { |
| 118 | + parsed, err := url.Parse(endpoint) |
| 119 | + if err != nil { |
| 120 | + return "", "", errors.Wrap(err, "invalid Gemini endpoint") |
| 121 | + } |
| 122 | + path := strings.TrimRight(parsed.Path, "/") |
| 123 | + apiVersion := defaultGeminiAPIVersion |
| 124 | + for _, supportedVersion := range []string{"v1alpha", "v1beta", "v1"} { |
| 125 | + if path == "/"+supportedVersion || strings.HasSuffix(path, "/"+supportedVersion) { |
| 126 | + apiVersion = supportedVersion |
| 127 | + parsed.Path = strings.TrimSuffix(path, "/"+supportedVersion) |
| 128 | + break |
| 129 | + } |
| 130 | + } |
| 131 | + return strings.TrimRight(parsed.String(), "/"), apiVersion, nil |
| 132 | +} |
| 133 | + |
| 134 | +func normalizeGeminiContentType(contentType string) (string, error) { |
| 135 | + mediaType, _, err := mime.ParseMediaType(strings.TrimSpace(contentType)) |
| 136 | + if err != nil { |
| 137 | + return "", errors.Wrap(err, "invalid audio content type") |
| 138 | + } |
| 139 | + mediaType = strings.ToLower(mediaType) |
| 140 | + normalized, ok := geminiSupportedContentTypes[mediaType] |
| 141 | + if !ok { |
| 142 | + return "", errors.Errorf("audio content type %q is not supported by Gemini", mediaType) |
| 143 | + } |
| 144 | + return normalized, nil |
| 145 | +} |
| 146 | + |
| 147 | +func buildGeminiTranscriptionPrompt(prompt string, language string) string { |
| 148 | + parts := []string{geminiTranscriptionPrompt} |
| 149 | + language = strings.TrimSpace(language) |
| 150 | + if language != "" { |
| 151 | + parts = append(parts, "The input language is "+language+".") |
| 152 | + } |
| 153 | + prompt = strings.TrimSpace(prompt) |
| 154 | + if prompt != "" { |
| 155 | + parts = append(parts, "Context and spelling hints:\n"+prompt) |
| 156 | + } |
| 157 | + return strings.Join(parts, "\n\n") |
| 158 | +} |
| 159 | + |
| 160 | +func normalizeGeminiModelName(model string) string { |
| 161 | + return strings.TrimPrefix(strings.TrimSpace(model), "models/") |
| 162 | +} |
0 commit comments