carlosduplar
Replace tags with official hackathon format
01601b9
|
Raw History Blame Contribute Delete
2.48 kB

A newer version of the Gradio SDK is available: 6.29.0

Upgrade
metadata
title: Patient Virtuel · Français en médecine dentaire
emoji: 🦷
colorFrom: red
colorTo: gray
sdk: gradio
sdk_version: 6.16.0
app_file: src/server_app.py
pinned: false
license: apache-2.0
tags:
  - hackathon
  - build-small-hackathon
  - backyard-ai
  - voice-ai
  - french-learning
  - dental
  - gradio
  - modal
  - llama-cpp
  - gemma
  - track:backyard
  - sponsor:modal
  - achievement:offgrid
  - achievement:offbrand
  - achievement:llama
  - achievement:sharing
  - achievement:fieldnotes

Patient Virtuel · Français en médecine dentaire

Hackathon submission for build-small-hackathon.

Track: Backyard AI — a real tool for a real learner: a dental hygienist training professional French.

Uses gradio.Server with a custom HTML/CSS/JS frontend (Off-Brand Award entry).

Llama Champion badge: Model runs through the llama.cpp runtime via llama-cpp-python with flash attention and L4 GPU optimizations.

Off the Grid badge: Zero cloud APIs. STT → faster-whisper (local), LLM → Gemma 4 (local), TTS → piper-tts (local neural, fr_FR-siwis-medium voice).

See space_README.md for the full description.

📖 Field Notes — architecture decisions, lessons learned, and the story behind the app. 📡 Agent Trace — curated build log on HF Hub (Sharing is Caring badge).

Demo video · LinkedIn post

Architecture

src/
  server_app.py       → gr.Server backend (API endpoints)
  core.py             → Shared session logic
  llm_engine.py       → Gemma 4 26B-A4B via Modal (llama.cpp runtime)
  stt_engine.py       → faster-whisper via Modal
  tts_engine.py       → piper-tts (local neural, fr_FR-siwis-medium)
  prompts.py          → System prompts
  parse_feedback.py   → Feedback parsing
frontend/
  custom_index.html   → Custom vanilla HTML/CSS/JS frontend
  style.css           → Styles
modal/
  modal_app.py        → Modal deployments (llama.cpp + CUDA)

Running locally

pip install -r requirements.txt
python src/server_app.py

Deploying to HF Spaces

The Space uses space_README.md as its README and src/server_app.py as the entry point.