[yoyoismee]

Adds a Paxa Labs example to the Speech Settings page (content/docs/configuration/stt_tts.mdx), for both speech.tts.openai and speech.stt.openai. I'm one of the founders of Paxa Labs, so this is our own API.

The existing OpenAI-compatible examples use a local placeholder URL and OpenAI's model and voice names. Our endpoint needs the full url, so a working example with Paxa's own model and voice IDs saves a round of guessing. OpenAI's names (tts-1, alloy, whisper-1) are accepted too.

Checked

• The keys against packages/data-provider/src/config.ts. tts.openai takes url, apiKey, model and voices, and stt.openai takes url, apiKey and model. TTSService and STTService post to url as the full endpoint path.
• The request bodies LibreChat sends, tested against our API. {input, model, voice} with no response_format returns mp3. A multipart model + file + language returns JSON with text. I haven't run a full LibreChat instance with this config.
• The page compiles with @mdx-js/mdx + remark-gfm, and both snippets parse as YAML. They have no version: line, so check:config-version is unaffected. I didn't run the full pnpm build. The change is English only, and I left the translations to the workflow.

If you'd rather keep this page vendor-neutral, feel free to close this.

[vercel]

@yoyoismee is attempting to deploy a commit to the LibreChat's projects Team on Vercel.

A member of the Team first needs to authorize it.