Repository navigation
feat(voice): send transcription, speakers and speech to cloud providers - #116
Merged
Merged
Conversation
Local SenseVoice/Whisper stay the default. Each voice task can now go to an OpenAI-compatible server, ElevenLabs, AssemblyAI or Deepgram, each with its own base URL (proxies, regional or compatible servers), model and sealed key. - transcribe.engine: openai, elevenlabs, assemblyai, deepgram; transcripts carry provider speakers, language, audio events and sentiment as emotion - diarize.engine provider (and auto) use the provider's speaker labels - voice.stt picks the voice chat engine; voice.tts_provider adds ElevenLabs and Deepgram Aura next to the OpenAI-compatible speech server - Settings > Speech providers with a Test button per provider - POST /settings/speech/test; docs/speech-providers.md; DEVLOG plan Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_014Fki157VgxFETzSSpwFwYe
5 of 8 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Requested by Adeel · project thread
Description
Before: transcription, speaker separation and emotion only ran locally (SenseVoice, Whisper, SpeechBrain), and spoken answers only came from an OpenAI-compatible speech server.
After: local engines stay the default and behave exactly as before. An admin can send each voice task to a provider instead: an OpenAI-compatible Whisper endpoint, ElevenLabs (speech to text and text to speech), AssemblyAI or Deepgram (speech to text and Aura text to speech). Every provider has its own base URL, so a proxy, a regional endpoint or a compatible server works too.
How:
speechsettings section (base URL, model, sealed API key per provider; sentiment switch; timeout) and Settings → Speech providers with a Test button per provider (POST /settings/speech/test, lists models, nothing billed).transcribe.enginegainsopenai,elevenlabs,assemblyai,deepgram(app/domain/speech.py). Provider transcripts carry speaker labels, language, audio events (ElevenLabs) and sentiment as emotion (AssemblyAI, Deepgram). Audio goes out as Ogg Opus 32k (FLAC fallback); long audio is chunked for OpenAI-compatible servers (25 MB limit).diarize.enginegainsprovider;autouses provider labels when a transcript has them, otherwise unchanged. Voice IDs still come from local voiceprints.voice.stt(defaultsame= transcription engine) andvoice.tts_provider(openaidefault,elevenlabs,deepgram).docs/speech-providers.md. DEVLOG plan added; the local GGUF/llama.cpp part comes as a separate PR.No migration needed: new settings keys have defaults, and the new
provider_speakersegment field lives on a schemaless table.Motivation and Context
Adeel asked for voice tasks to be outsourceable to other platforms, with custom endpoints for any service, while keeping local processing as is.
Types of changes
!in the title)Checklist:
tests/domain/test_speech.pywith a fake server for all four providers at custom addresses;__tests__/settings-speech.test.tsx).docs/accordingly.## UnreleasedinCHANGELOG.md.Local runs: backend 545 passed, 1 failed (the known environmental LibreOffice conversion test); frontend 79 suites, 656 tests passed.
🤖 Generated with Claude Code
https://claude.ai/code/session_014Fki157VgxFETzSSpwFwYe
Generated by Claude Code