openagent

Open-source LLM agent platform — knowledge bases, multi-channel agents, conversation flows, tool calling, embedding & semantic search.

Docker Hub · Self-hosting · Issues

✨ Features

• LLM-agnostic — built on LiteLLM; bring keys for OpenAI, Aliyun Bailian, Moonshot, Zhipu, Minimax, OpenRouter, or any OpenAI-compatible endpoint.
• Knowledge bases with semantic search — Git-backed document ingestion, pgvector-powered retrieval, BGE embedding + reranker out of the box.
• Visual agent builder — define tools, prompts, and conversation flows in the admin UI.
• Multi-channel chat — embed agents on websites, help centers, or via API.
• Help center hosting — Markdown-driven public documentation site, multi-tab navigation.
• Multi-tenant ready — single-tenant out of the box; the Enterprise edition adds platform-level tenant management.
• OpenTelemetry observability — vendor-neutral OTLP traces & logs (works with SigNoz, Honeycomb, Tempo, Grafana Cloud).

🚀 Quick Start

git clone https://github.com/DeepFlowAI/OpenAgent.git
cd openagent/docker
cp .env.example .env
python -c "import secrets; print('SECRET_KEY=' + secrets.token_hex(32))" >> .env

docker compose up -d        # bundles PostgreSQL (pgvector) + Redis by default

Open http://localhost:3000 and sign in:

Field	Default
Tenant	default
Username	admin
Password	Admin123456

⚠️ Change the password immediately after first login in any non-local deployment.

You'll also want to set at least one LLM provider API key in .env (e.g. OPENROUTER_API_KEY=sk-...) and SILICONFLOW_API_KEY=... for the knowledge-base embedding/reranker.

Bring your own database / Redis

Edit COMPOSE_PROFILES= in .env to drop the bundled service(s), then set DATABASE_URL / REDIS_URL. The command stays docker compose up -d. Full deployment guide → docker/README.md.

External PostgreSQL must have the pgvector extension installed (used for semantic search).

🛠️ Local Development

# Backend
cd server && python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env.dev   # set DATABASE_URL / REDIS_URL / LLM_API_KEY etc.
uvicorn app.main:app --reload --port 5001

# Frontend
cd web && npm install
cp .env.example .env.dev
npm run dev

Service	URL
Frontend	http://localhost:3000
Backend API	http://localhost:5001
API docs (Swagger)	http://localhost:5001/docs

🧩 Architecture

┌─────────────────────────┐    ┌─────────────────────────┐
│  Next.js + React        │◄──►│   FastAPI + LiteLLM     │
│  (web)                  │    │   (api)                 │
└─────────────────────────┘    └────────────┬────────────┘
                                            │
                          ┌─────────────────┼─────────────────┐
                          ▼                 ▼                 ▼
                  ┌──────────────┐  ┌──────────────┐  ┌──────────────┐
                  │  pgvector    │  │   Redis 7    │  │ Object store │
                  │ (vectors+RDS)│  │ (cache+queue)│  │ (uploads)    │
                  └──────────────┘  └──────────────┘  └──────────────┘

Backend follows a strict Router → Service → Repository → Model layering with Pydantic schemas for I/O. Closed-source extensions (e.g. multi-tenant management) plug in via a convention-based loader at app/extensions/.

🏢 Editions

Edition	Tenants	Tenant management API
Community	Single auto-provisioned default	—
Enterprise	Multi-tenant	/api/v1/tenants (closed)

The Community edition (this repository) is feature-complete for single-organization deployments. The Enterprise extension adds the platform-level tenant CRUD API for SaaS operators.

🤝 Contributing

Issues, discussions, and pull requests are welcome. Before opening a PR please make sure tests pass:

cd server && pytest

📜 License

openagent is released under the GNU Affero General Public License v3.0.