Introduction

Airbeeps is a self-hosted, local-first assistant-based Retrieval-Augmented Generation (RAG) system for individuals and small teams who want to build AI assistants on top of their own documents.

You run it yourself, upload your data, and interact with multiple AI assistants backed by your private knowledge base. All data stays on your machine or server.

Who is this for?

Airbeeps is well suited for:

• Individuals running personal or local AI assistants
• Small teams sharing internal documents
• Internal tools, labs, and early production setups
• Developers who want a hackable, self-hosted RAG system

Quick install

pip install airbeeps
airbeeps run

Open http://localhost:8500 and sign up — the first user becomes admin.

Key features

Multi-Assistant Platform

Create multiple assistants with configurable system prompts, model parameters, and knowledge base access.

RAG Knowledge Base

Upload PDFs, DOCX, PPTX, TXT, Markdown, Excel, and CSV. Smart chunking (semantic, hierarchical, or sentence-based), vector embedding, and semantic search with citations.

Pluggable LLM Providers

Uses LiteLLM to support OpenAI, Anthropic, Gemini, and any OpenAI-compatible API.

Agent Capabilities

Tool-calling agents powered by LangGraph with built-in tools for knowledge base search, web search, code execution, and file operations. MCP server integration and multi-agent orchestration.

Advanced Retrieval

Hybrid search (dense + BM25 with RRF fusion), multi-query expansion (HyDE, step-back), and cross-encoder reranking (BGE, Cohere, ColBERT).

Local-First

All data (database, files, vectors) stored locally with no external dependencies. SQLite + Qdrant by default, with optional PostgreSQL, S3/MinIO, and Redis for production.

Next steps

• Quickstart — Get up and running in 5 minutes
• Configuration — Environment variables and settings
• Knowledge Bases — Learn how document storage works
• Retrieval — Understand dense, hybrid, and rerank strategies
• Agents — AI agents with tool calling and MCP
• Docker Deployment — Production Docker Compose setup

Tech stack

Layer	Technology
Backend	FastAPI, SQLAlchemy, Alembic, LlamaIndex, LiteLLM, LangGraph
Vector Store	Qdrant (default), ChromaDB, PGVector, Milvus
Frontend	Nuxt 4, Vue 3, TailwindCSS, shadcn-nuxt, Reka UI
Auth	FastAPI Users (JWT with refresh tokens, OAuth)
Optional	Redis (caching), Celery (task queue), S3/MinIO (file storage)

License

Airbeeps is open source under the MIT License.