forked from sashabaranov/go-openai
-
Notifications
You must be signed in to change notification settings - Fork 0
/
audio.go
191 lines (163 loc) · 5.43 KB
/
audio.go
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
package openai
import (
"bytes"
"context"
"fmt"
"io"
"net/http"
"os"
utils "github.com/sashabaranov/go-openai/internal"
)
// Whisper Defines the models provided by OpenAI to use when processing audio with OpenAI.
const (
Whisper1 = "whisper-1"
)
// Response formats; Whisper uses AudioResponseFormatJSON by default.
type AudioResponseFormat string
const (
AudioResponseFormatJSON AudioResponseFormat = "json"
AudioResponseFormatText AudioResponseFormat = "text"
AudioResponseFormatSRT AudioResponseFormat = "srt"
AudioResponseFormatVerboseJSON AudioResponseFormat = "verbose_json"
AudioResponseFormatVTT AudioResponseFormat = "vtt"
)
// AudioRequest represents a request structure for audio API.
// ResponseFormat is not supported for now. We only return JSON text, which may be sufficient.
type AudioRequest struct {
Model string
// FilePath is either an existing file in your filesystem or a filename representing the contents of Reader.
FilePath string
// Reader is an optional io.Reader when you do not want to use an existing file.
Reader io.Reader
Prompt string // For translation, it should be in English
Temperature float32
Language string // For translation, just do not use it. It seems "en" works, not confirmed...
Format AudioResponseFormat
}
// AudioResponse represents a response structure for audio API.
type AudioResponse struct {
Task string `json:"task"`
Language string `json:"language"`
Duration float64 `json:"duration"`
Segments []struct {
ID int `json:"id"`
Seek int `json:"seek"`
Start float64 `json:"start"`
End float64 `json:"end"`
Text string `json:"text"`
Tokens []int `json:"tokens"`
Temperature float64 `json:"temperature"`
AvgLogprob float64 `json:"avg_logprob"`
CompressionRatio float64 `json:"compression_ratio"`
NoSpeechProb float64 `json:"no_speech_prob"`
Transient bool `json:"transient"`
} `json:"segments"`
Text string `json:"text"`
}
// CreateTranscription — API call to create a transcription. Returns transcribed text.
func (c *Client) CreateTranscription(
ctx context.Context,
request AudioRequest,
) (response AudioResponse, err error) {
return c.callAudioAPI(ctx, request, "transcriptions")
}
// CreateTranslation — API call to translate audio into English.
func (c *Client) CreateTranslation(
ctx context.Context,
request AudioRequest,
) (response AudioResponse, err error) {
return c.callAudioAPI(ctx, request, "translations")
}
// callAudioAPI — API call to an audio endpoint.
func (c *Client) callAudioAPI(
ctx context.Context,
request AudioRequest,
endpointSuffix string,
) (response AudioResponse, err error) {
var formBody bytes.Buffer
builder := c.createFormBuilder(&formBody)
if err = audioMultipartForm(request, builder); err != nil {
return AudioResponse{}, err
}
urlSuffix := fmt.Sprintf("/audio/%s", endpointSuffix)
req, err := c.newRequest(ctx, http.MethodPost, c.fullURL(urlSuffix, request.Model),
withBody(&formBody), withContentType(builder.FormDataContentType()))
if err != nil {
return AudioResponse{}, err
}
if request.HasJSONResponse() {
err = c.sendRequest(req, &response)
} else {
err = c.sendRequest(req, &response.Text)
}
if err != nil {
return AudioResponse{}, err
}
return
}
// HasJSONResponse returns true if the response format is JSON.
func (r AudioRequest) HasJSONResponse() bool {
return r.Format == "" || r.Format == AudioResponseFormatJSON || r.Format == AudioResponseFormatVerboseJSON
}
// audioMultipartForm creates a form with audio file contents and the name of the model to use for
// audio processing.
func audioMultipartForm(request AudioRequest, b utils.FormBuilder) error {
err := createFileField(request, b)
if err != nil {
return err
}
err = b.WriteField("model", request.Model)
if err != nil {
return fmt.Errorf("writing model name: %w", err)
}
// Create a form field for the prompt (if provided)
if request.Prompt != "" {
err = b.WriteField("prompt", request.Prompt)
if err != nil {
return fmt.Errorf("writing prompt: %w", err)
}
}
// Create a form field for the format (if provided)
if request.Format != "" {
err = b.WriteField("response_format", string(request.Format))
if err != nil {
return fmt.Errorf("writing format: %w", err)
}
}
// Create a form field for the temperature (if provided)
if request.Temperature != 0 {
err = b.WriteField("temperature", fmt.Sprintf("%.2f", request.Temperature))
if err != nil {
return fmt.Errorf("writing temperature: %w", err)
}
}
// Create a form field for the language (if provided)
if request.Language != "" {
err = b.WriteField("language", request.Language)
if err != nil {
return fmt.Errorf("writing language: %w", err)
}
}
// Close the multipart writer
return b.Close()
}
// createFileField creates the "file" form field from either an existing file or by using the reader.
func createFileField(request AudioRequest, b utils.FormBuilder) error {
if request.Reader != nil {
err := b.CreateFormFileReader("file", request.Reader, request.FilePath)
if err != nil {
return fmt.Errorf("creating form using reader: %w", err)
}
return nil
}
f, err := os.Open(request.FilePath)
if err != nil {
return fmt.Errorf("opening audio file: %w", err)
}
defer f.Close()
err = b.CreateFormFile("file", f)
if err != nil {
return fmt.Errorf("creating form file: %w", err)
}
return nil
}