webvoice: browser voice client spike (phone-direct Live) - #10
Merged
Conversation
…-brain) Validated the phone-client architecture end to end: the browser holds the Gemini Live session (OS does AEC + barge-in — confirmed clean on iOS WebKit, where native desktop Python couldn't), and the Aria brain is reached via one escalate_to_aria tool -> POST /agent. Ephemeral token keeps the API key server-side. - server.py: /live-token (ephemeral, tool baked in), /agent (run the full brain), serves the client. No new deps (fastapi/uvicorn/google-genai already present). - index.html + capture-worklet.js: getUserMedia(echoCancellation) -> 16k PCM -> Live; 24k playback; escalate handoff; on-page log. - README: what it proved, how to run (uvicorn + ngrok), and that it's a no-auth prototype. Spike, not production: single thread, no auth — don't expose publicly. Assisted-by: Claude Code (Opus 4.8)
- PWA manifest + apple-touch-icon + meta tags so iOS "Add to Home Screen" gives a fullscreen, tap-to-launch Aria icon (no App Store, no native code). - Generated app icon (stdlib PNG, no deps); server serves /manifest.json and /icon.png. - run.sh: starts uvicorn + ngrok in one command; ARIA_NGROK_DOMAIN gives a STABLE url so the home-screen icon keeps working. - README: the phone flow (run.sh → open → Add to Home Screen) + the stable-domain setup. Assisted-by: Claude Code (Opus 4.8)
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
A working prototype of the phone-direct voice architecture and the artifact from spike #2.
What it proved
/live-token).escalate_to_aria→POST /agent→ the full Claude brain (memory/email/calendar/tools) → spoken back, snappy.Contents
server.py—/live-token(ephemeral, escalate tool baked into the token),/agent(runs the brain), serves the client. No new deps.index.html+capture-worklet.js—getUserMedia({echoCancellation})→ 16 kHz PCM → Live; 24 kHz playback; the escalate handoff; on-page log.README.md— what it proved + how to run (uvicorn+ngrok).Spike, not production: single shared thread, no auth — don't expose publicly. Inert to the main app (separate entry point, no new deps). Next steps toward a real client: auth, per-user threads, push, native shell.
🤖 Generated with Claude Code