Self-hosted local AI platform with chat, agents, and a full code workspace

What is Pirarara AI?

Pirarara AI is a self-hosted alternative to cloud AI tools. Run powerful open-source models locally via Ollama, chat with them through a polished UI, build agents that can read and write your code, and keep every byte on your own hardware.

Named after the Pirarara (Phractocephalus hemioliopterus) — a magnificent red-tailed catfish native to the Amazon river. Powerful, local, and distinctly Brazilian. 🇧🇷

Features

• Chat — streaming chat with SSE, multi-conversation sidebar, rename/delete, chat history persisted in SQLite
• Model manager — pull models with live progress bar, load into VRAM, delete with confirmation
• Agent loop — coding agent with read_file, write_file, and run_shell tools
• Workspace — file tree, file viewer with diff tab, integrated terminal panel
• System stats — live CPU / RAM / GPU bar in the navbar
• OpenAI-compatible API — drop-in replacement for any OpenAI client (/v1/chat/completions, /v1/models)
• User auth — JWT login, admin role, default admin / pirarara credentials seeded on startup
• Settings — toggle visible tabs, set default model, change password

Quick start

git clone https://github.com/YOUR_USERNAME/pirarara-ai
cd pirarara-ai
cp .env.example .env
# Edit .env — set DB_PASSWORD and BACKEND_SECRET_KEY at minimum

# CPU only
docker compose up

# With Nvidia GPU
docker compose -f docker-compose.yml -f docker-compose.gpu.yml up

Open http://localhost:3000 and sign in with admin / pirarara.

Requirements

• Docker + Docker Compose
• 8 GB RAM minimum (16 GB recommended)
• Nvidia GPU with 6 GB+ VRAM (optional but recommended)
• Ollama reachable from the container (or running on the same host)

See docs/requirements.md for full details.

OpenAI-compatible API

Any OpenAI-compatible client can point at Pirarara. All inference runs on Ollama.

OPENAI_BASE_URL=http://localhost:8000/v1
OPENAI_API_KEY=pirarara   # any non-empty string

Backend endpoints

Local backend development

cd backend
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt
uvicorn app.main:app --reload --port 8000

Minimum .env for local dev:

OLLAMA_URL=http://192.168.15.180:11434
DB_PASSWORD=any_local_password
BACKEND_SECRET_KEY=any_local_secret

Unraid

See docs/unraid-setup.md for the full Unraid setup guide.

License

MIT