AIHub

Self-hostable AI backend for document Q&A — runs fully local with Ollama, or plug in OpenAI / Gemini API keys. Switch providers at runtime, no code changes.

AIHub gives you a complete RAG (Retrieval-Augmented Generation) stack you can run on your own machine: ingest documents, search them semantically, chat with sources, and link directories that auto-sync as files change. Designed for teams that want the convenience of a hosted AI assistant without sending their documents to a third party.

• Local-first — defaults to Ollama, no API keys needed
• Bring Your Own Model (BYOM) — swap in OpenAI or Gemini per request
• Full RAG pipeline — parsing, chunking, embeddings, vector search, answer synthesis with source attribution
• Persistent chat sessions — multi-turn conversations with history
• Directory linking — point at a folder, files are auto-ingested as they change
• Interactive CLI — drive everything from the terminal without writing curl by hand
• Encrypted credentials — provider API keys are encrypted at rest

For a deep dive on the design, see ARCHITECTURE.md.

Stack

Node.js · TypeScript · Fastify · PostgreSQL (Drizzle ORM) · Redis (BullMQ) · Qdrant · Ollama / OpenAI / Gemini

Client → Fastify API → Provider Registry → Ollama / OpenAI / Gemini
                     → BullMQ Queue → Ingestion Worker → Qdrant
                     → RAG Engine → Qdrant + AI Provider
                     → Chat Manager → History + RAG + AI Provider

Quick Start

1. Start infrastructure

docker compose up -d

This starts PostgreSQL, Redis, Qdrant, and Ollama.

2. Pull an Ollama model

docker exec aihub-ollama ollama pull llama3.2
docker exec aihub-ollama ollama pull nomic-embed-text

3. Install and run

cp .env.example .env
# Generate an encryption key and paste it into .env:
node -e "console.log(require('crypto').randomBytes(32).toString('hex'))"

npm install
npm run db:push
npm run dev

Server starts at http://localhost:3001.

4. Use the CLI (recommended)

npm run cli

Interactive menu for ingesting files, querying, chatting, and managing providers — no curl required.

5. Or use the HTTP API

# Ingest a document
curl -X POST http://localhost:3001/api/ingest \
  -F 'file=@./my-document.pdf'

# Query
curl -X POST http://localhost:3001/api/query \
  -H 'Content-Type: application/json' \
  -d '{"query":"What is this document about?"}'

CLI

The CLI (npm run cli) wraps the HTTP API in an interactive menu:

• Link directory — point at a folder and auto-ingest everything in it
• Upload files — pick files to ingest individually
• Query — single-shot RAG question
• Chat — multi-turn conversation with history
• Sources — list and manage ingested documents
• Providers — configure Ollama / OpenAI / Gemini at runtime

Provider configuration

Ollama is used by default. To switch to a cloud provider at runtime:

OpenAI

curl -X POST http://localhost:3001/api/providers/config \
  -H 'Content-Type: application/json' \
  -d '{"providerType":"openai","capability":"generation","apiKey":"sk-...","isActive":true}'

Google Gemini

curl -X POST http://localhost:3001/api/providers/config \
  -H 'Content-Type: application/json' \
  -d '{"providerType":"gemini","capability":"generation","apiKey":"AI...","isActive":true}'

API keys are encrypted with AES-256-GCM before being written to Postgres, using the ENCRYPTION_KEY from your env.

API Reference

Method	Endpoint	Description
GET	/api/health	Health check
POST	/api/ingest	Upload a document (multipart)
POST	/api/query	Single-shot RAG query
GET	/api/sources	List ingested documents
GET	/api/sources/:id	Get document details
DELETE	/api/sources/:id	Delete document + vectors
POST	/api/directories/link	Link a directory for ingestion
GET	/api/directories	List linked directories
POST	/api/directories/:id/sync	Re-sync a directory
DELETE	/api/directories/:id	Unlink directory
POST	/api/chat	Create chat session
POST	/api/chat/:id/message	Send chat message
GET	/api/chat/:id/history	Get chat history
GET	/api/chat	List chat sessions
DELETE	/api/chat/:id	Delete chat session
POST	/api/providers/config	Configure AI provider
GET	/api/providers	List provider configs
DELETE	/api/providers/:id	Delete provider config

Environment Variables

Variable	Default	Description
PORT	3001	Server port
DATABASE_URL	postgresql://...localhost:5433/aihub	PostgreSQL connection
REDIS_URL	redis://localhost:6380	Redis connection
QDRANT_URL	http://localhost:6333	Qdrant connection
OLLAMA_URL	http://localhost:11434	Ollama connection
ENCRYPTION_KEY	(required)	32-byte hex key for encrypting provider API keys at rest

License

MIT — see LICENSE.