Browser-local practice backend

Aperta
#621 2 commenti 0 reazioni 1 assegnatario Vedi su GitHub

I maintainer di solito rispondono entro 1 giorno

@tony2037 ci sta già lavorando.

Dal 8/11/2026.

Valutazione

Questa issue non è ancora stata valutata.

Descrizione

I have a working Windows/Edge prototype that approaches #330 from the browser side. Would this be welcome as an optional Practice + Editor backend?

Design

  • Run interviewer turns through the browser's LanguageModel Prompt API, using the current problem, code, recent conversation and actual test results as context.
  • Recognize English speech on device, using native local recognition or a checksum-pinned Whisper WASM worker. Use installed local voices for replies.
  • Reuse the existing transcript, test runner and report UI. Store clearly labelled, unverified practice feedback only in device history.

Benefits

  • No Gemini API key or additional inference server to configure.
  • Model inference and speech recognition stay on the user's device.
  • Model, speech, report validation and UI integration are separate ES modules, with no frontend build step.

Scope and evidence
This prototype supports Practice with Editor. LiveKit still transports media, and C/C++/Java execution still uses Compiler Explorer. Strict interviews, whiteboard assessment and server recording remain on the cloud path. Its simplified practice feedback does not implement the cloud assessment contract.

Windows/Edge has been exercised end to end. 327 targeted tests pass, including actual Whisper recognition with networking disabled. Chrome/Linux end-to-end validation and the complete gate remain pending; no quality or latency parity with Gemini Live is claimed.

Would you prefer this as a browser-local backend in CodeTrial, or an out-of-tree experiment alongside the server-side pipeline?

Also see:

Lingua principale
Rust
Stelle
223
Fork
148
Merge medio
1g 4h
PR unite (30g)
223

Preparare l'ambiente

Come iniziare

  1. Leggi tutta la issue e poi la guida ai contributi del progetto.
  2. Commenta sulla issue per dire che te ne occupi tu — evita che due persone facciano lo stesso lavoro.
  3. Fai un fork del repository e lavora su un branch.
  4. Apri una pull request che faccia riferimento al numero della issue.

Altre issue di sysprog21/codetrial

Tutte le issue di sysprog21/codetrial

Issue simili

Altre issue su Rust

Ricevi le nuove issue nella tua casella

Un breve riepilogo di issue GitHub adatte ai principianti.