RainbowBird
8e572ad93c
Signed-off-by: RainbowBird <git@luoling.moe> Commit-Message-Assisted-by: Claude (via Claude Code)
115 lines
4.3 KiB
TypeScript
115 lines
4.3 KiB
TypeScript
import type { Voice } from 'unspeech'
|
|
|
|
import type { TtsAdapter, TtsAdapterContext, TtsInput, TtsResult, TtsVoiceCatalogContext } from './types'
|
|
|
|
import { createBadRequestError, createInternalError } from '../../../utils/error'
|
|
import { nanoid } from '../../../utils/id'
|
|
import { audioMimeFromFormat } from './audio-format'
|
|
import { listVoicesViaUnSpeech, sendSpeechViaUnSpeech } from './unspeech'
|
|
|
|
/**
|
|
* Default Volcengine TTS voice id. `BV001_streaming` is Volcengine's standard
|
|
* Chinese general-purpose streaming voice referenced in their docs.
|
|
*/
|
|
const DEFAULT_VOLCENGINE_VOICE = 'BV001_streaming'
|
|
|
|
/**
|
|
* Default Volcengine audio encoding. Matches our OpenAI-shape `mp3` default.
|
|
*/
|
|
const DEFAULT_VOLCENGINE_FORMAT = 'mp3'
|
|
|
|
/**
|
|
* Default Volcengine cluster. Documented as `volcano_tts` for the generic
|
|
* hosted TTS endpoint; ops can override via `adapterParams.cluster`.
|
|
*/
|
|
const DEFAULT_VOLCENGINE_CLUSTER = 'volcano_tts'
|
|
|
|
/**
|
|
* Volcengine non-streaming REST adapter.
|
|
*
|
|
* Use when:
|
|
* - Routing a hosted TTS request to Volcengine OpenSpeech.
|
|
*
|
|
* Expects:
|
|
* - `ctx.baseURL` is the Volcengine TTS endpoint, e.g.
|
|
* `https://openspeech.bytedance.com/api/v1/tts`.
|
|
* - `ctx.keyPlaintext` is the access token. The auth header uses Volcengine's
|
|
* non-standard `Bearer; <token>` format (semicolon after `Bearer`).
|
|
* - `ctx.adapterParams.appid` is the Volcengine application id (required).
|
|
* - `ctx.adapterParams.cluster` overrides the default cluster id when set.
|
|
*
|
|
* Returns:
|
|
* - {@link TtsResult} with the audio bytes as an `ArrayBuffer`. Body is
|
|
* decoded from the upstream JSON `data` base64 field.
|
|
*/
|
|
export const volcengineAdapter: TtsAdapter = {
|
|
id: 'volcengine',
|
|
|
|
async send(input: TtsInput, ctx: TtsAdapterContext): Promise<TtsResult> {
|
|
const appid = ctx.adapterParams.appid
|
|
if (typeof appid !== 'string' || !appid)
|
|
throw createInternalError('volcengine tts: adapterParams.appid is required')
|
|
|
|
const cluster = typeof ctx.adapterParams.cluster === 'string'
|
|
? ctx.adapterParams.cluster
|
|
: DEFAULT_VOLCENGINE_CLUSTER
|
|
|
|
const apiResourceId = typeof ctx.adapterParams.model === 'string'
|
|
? ctx.adapterParams.model
|
|
: undefined
|
|
|
|
const voice = input.voice ?? DEFAULT_VOLCENGINE_VOICE
|
|
if (typeof input.extraOptions?.pitch === 'number' || typeof input.extraOptions?.volume === 'number') {
|
|
throw createBadRequestError(
|
|
'volcengine does not support Voice Pack pitch or volume parameters',
|
|
'BAD_REQUEST',
|
|
)
|
|
}
|
|
const encoding = input.responseFormat ?? DEFAULT_VOLCENGINE_FORMAT
|
|
const speed = input.speed ?? 1
|
|
|
|
// unspeech volcengine backend (unspeech/pkg/backend/volcengine/speech.go):
|
|
// - reads token from `Authorization: Bearer <token>` (strips "Bearer "
|
|
// prefix), then re-attaches as `Bearer; <token>` to the upstream — so
|
|
// we send a normal Bearer here, NOT the `Bearer; ` form.
|
|
// - takes `app.appid`, `app.cluster`, `user.uid`, `request.reqid`,
|
|
// `audio.encoding`, `audio.speed_ratio` from `extra_body` jsonpath.
|
|
// - decodes the upstream base64 audio frame itself and returns binary.
|
|
return sendSpeechViaUnSpeech({
|
|
ctx,
|
|
model: apiResourceId ? `volcengine/${apiResourceId}` : 'volcengine',
|
|
input: input.text,
|
|
voice,
|
|
responseFormat: encoding,
|
|
extraBody: {
|
|
app: { appid, cluster },
|
|
user: { uid: 'airi-server' },
|
|
audio: { speed_ratio: speed },
|
|
request: { reqid: nanoid(), operation: 'query' },
|
|
},
|
|
fallbackContentType: audioMimeFromFormat(encoding),
|
|
providerLabel: 'volcengine',
|
|
})
|
|
},
|
|
|
|
async getVoiceCatalog(ctx: TtsVoiceCatalogContext): Promise<Voice[]> {
|
|
// unspeech embeds the Volcengine catalog at build time
|
|
// (unspeech/pkg/backend/volcengine/voices.go), filtered server-side to
|
|
// streaming-compatible voices. Passing `model=<api_resource_id>` narrows
|
|
// further by `compatible_models` — adapterParams.model is the operator-
|
|
// configured resource id (e.g. `seed-tts-2.0`).
|
|
const params = new URLSearchParams({ provider: 'volcengine' })
|
|
const apiResourceId = typeof ctx.adapterParams?.model === 'string'
|
|
? ctx.adapterParams.model
|
|
: undefined
|
|
if (apiResourceId)
|
|
params.set('model', apiResourceId)
|
|
|
|
return listVoicesViaUnSpeech({
|
|
ctx,
|
|
query: params.toString(),
|
|
providerLabel: 'volcengine',
|
|
})
|
|
},
|
|
}
|