1 | type Openai = { |
2 | api_key: string; |
3 | organization_id: string; |
4 | }; |
5 | type Base64 = string; |
6 | |
7 | * Create transcription |
8 | * Transcribes audio into the input language. |
9 | */ |
10 | export async function main( |
11 | auth: Openai, |
12 | body: { |
13 | file: { |
14 | base64: Base64; |
15 | type: |
16 | | "image/png" |
17 | | "image/jpeg" |
18 | | "image/gif" |
19 | | "application/pdf" |
20 | | "appication/json" |
21 | | "text/csv" |
22 | | "text/plain" |
23 | | "audio/mpeg" |
24 | | "audio/wav" |
25 | | "video/mp4"; |
26 | name: string; |
27 | }; |
28 | model: string | "whisper-1"; |
29 | language?: string; |
30 | prompt?: string; |
31 | response_format?: "json" | "text" | "srt" | "verbose_json" | "vtt"; |
32 | temperature?: number; |
33 | "timestamp_granularities[]"?: ("word" | "segment")[]; |
34 | } |
35 | ) { |
36 | const url = new URL(`https://api.openai.com/v1/audio/transcriptions`); |
37 |
|
38 | const formData = new FormData(); |
39 | for (const [k, v] of Object.entries(body)) { |
40 | if (v !== undefined && v !== "") { |
41 | if (["file"].includes(k)) { |
42 | const { base64, type, name } = v as { |
43 | base64: Base64; |
44 | type: string; |
45 | name: string; |
46 | }; |
47 | formData.append( |
48 | k, |
49 | new Blob([Uint8Array.from(atob(base64), (m) => m.codePointAt(0)!)], { |
50 | type, |
51 | }), |
52 | name |
53 | ); |
54 | } else { |
55 | formData.append(k, String(v)); |
56 | } |
57 | } |
58 | } |
59 | const response = await fetch(url, { |
60 | method: "POST", |
61 | headers: { |
62 | "OpenAI-Organization": auth.organization_id, |
63 | Authorization: "Bearer " + auth.api_key, |
64 | }, |
65 | body: formData, |
66 | }); |
67 | if (!response.ok) { |
68 | const text = await response.text(); |
69 | throw new Error(`${response.status} ${text}`); |
70 | } |
71 | return await response.json(); |
72 | } |
73 |
|