Persona AI

Psychologist avatar — voice-first chat, OpenAI-compatible TTS/STT, Rhubarb lip-sync, and FAQ-guided replies

Installation · Development · Build · Verify releases · Changelog

Persona AI is an open-source research demo: a supportive psychologist-style assistant with a 3D VRM/GLB avatar, Rhubarb lip-sync, and OpenAI-compatible chat / TTS / STT. A FastAPI backend loads a small FAQ corpus as a high-priority conversational roadmap (short clarifying questions), synthesizes speech, and derives mouth cues from the audio. The Tauri desktop app bundles the same stack and starts the Python sidecar automatically on launch. The UI opens in voice conversation first.

Disclaimer: This is a research / demo assistant. It does not diagnose or replace professional mental-health care. Configure the emergency contact number in apps/backend/.env.

Current version: 1.3.3

• Voice-first UI — full-screen voice sanctuary on launch; switch to chat anytime
• Bilingual UI — Persian and English with locale-locked system prompts
• OpenAI-compatible TTS / STT — HTTP speech synthesis and transcription
• Rhubarb lip-sync — mouth cues from the WAV (A–H / X → VRM / GLB morphs)
• VRM / GLB avatar — Three.js + @pixiv/three-vrm; named catalog (Kira, Lucien)
• FAQ roadmap — data/faq_dataset.json steers short, question-led replies (no vector RAG)
• Safety layer — high-risk detection and escalation replies
• Desktop app — Tauri + PyInstaller sidecar (Windows NSIS; Linux/macOS via CI releases)
• Themeable UI — themes, voice picker, and face-age controls

Animated preview: assets/media/preview.gif. Full installers: GitHub Releases.

git clone https://github.com/Satan2049/persona-ai.git
cd persona-ai/apps/backend
python -m venv .venv

Windows: .venv\Scripts\activate · Linux/macOS: source .venv/bin/activate

pip install -r requirements.txt
cp .env.example .env    # Windows: copy .env.example .env

Edit apps/backend/.env — set MODEL_* and optionally TTS_* / STT_*. Then from the repo root:

scripts\start-backend.bat        # Windows
./scripts/start-backend.ps1      # PowerShell

Open http://127.0.0.1:8000/ · Health check: http://127.0.0.1:8000/health

Download the latest installer from GitHub Releases.

On first launch the app:

1. Starts the bundled persona-backend sidecar on a local port
2. Waits for /health, then loads the UI in voice mode
3. Shows setup tips if the LLM or TTS API is misconfigured

Verify downloads with docs/TRUST.md.

persona-ai/
├── apps/
│   ├── backend/          # FastAPI, TTS, STT, Rhubarb, FAQ guidance
│   └── desktop/          # Tauri + sidecar packaging
├── assets/               # Icons, media (screenshots/GIF), avatars source, default.env
├── data/                 # FAQ corpus
├── docs/                 # Architecture, trust, voice, release notes, data layout
├── scripts/              # Dev and release helpers
└── ui/                   # Static avatar + voice UI

Not in git: API keys, generated apps/audio/ cache, downloaded tools/rhubarb/.

Windows:

npm run sidecar:build

Linux / macOS:

chmod +x scripts/build-sidecar.sh
./scripts/build-sidecar.sh

npm install
.\scripts\sync-desktop-ui.ps1
npm run sidecar:build
npm run desktop:build

• Windows: apps/desktop/src-tauri/target/release/bundle/nsis/
• Linux / macOS: GitHub Actions uploads AppImage/deb/DMG (+ SHA256-*.txt) on v* tags

.\scripts\generate-sha256.ps1 -ReleaseDir "dist\release"

See docs/TRUST.md and docs/release-notes-1.3.3.md.

• docs/architecture/overview.md — system overview
• docs/voice-conversation.md — voice-first flow
• docs/desktop-data-layout.md — install / AppData layout
• docs/TRUST.md — verify release hashes
• docs/release-notes-1.3.3.md — v1.3.3 release notes
• assets/avatars/README.md — replace avatars
• assets/media/README.md — screenshots / preview GIF

MIT — see LICENSE.