Titan AI
Local AI code generation platform built for private, GPU-backed development on your own hardware.
Live Demo
- Repository:
https://github.com/tylerdotai/titan-ai - Local app URL:
http://localhost:3000
About
Titan AI is a local-first coding assistant that keeps inference on your own hardware. The current repo combines a FastAPI backend, a static web interface, lightweight auth, and streaming chat endpoints that connect to a local model server for code generation and related tasks.
Tech Stack
| Layer | Technology |
|---|---|
| Backend | FastAPI |
| Database | SQLite |
| Auth | OAuth2-style token flow |
| Inference | Local model server over HTTP |
| Frontend | Static HTML in app/static/ |
Features
Core Product
- Local code generation with private inference flow
- Streaming chat endpoint in
/chat/stream - Text-to-speech endpoint for audio output
- Lightweight auth and user/task storage
Repo Surface
- Single FastAPI app in
app/main.py - Static frontend bundled in
app/static/ - SQLite-backed local state in
tasks.db
Project Structure
app/main.py FastAPI app, auth, chat, and TTS endpoints
app/static/index.html Local web interface
tasks.db SQLite data store
LICENSE MIT license
README.md Project overview
Getting Started
Prerequisites
- Python 3.10+
- A reachable local model server for inference
pip
Installation
git clone https://github.com/tylerdotai/titan-ai.git
cd titan-ai
pip install fastapi uvicorn sqlalchemy requests httpx python-multipart
Deployment
Titan AI is currently optimized for local/self-hosted use.
- Repository:
https://github.com/tylerdotai/titan-ai - Local app URL:
http://localhost:3000
Usage
uvicorn app.main:app --reload --port 3000
Current Limitations
- Local model server endpoints are hard-coded in the app
- Dependency management is implicit rather than fully packaged
- The current frontend is lightweight and not split into a separate modern web app
Roadmap
- Externalize model server configuration cleanly
- Add a formal dependency file and reproducible setup flow
- Improve the frontend experience for multi-step coding workflows
- Expand model and deployment flexibility across local hardware setups
License
MIT License - see LICENSE for details.
