Ai Knowledge Base
Self-hosted AI knowledge base pairing Dify with Qdrant vector search and Postgres. Ingest documents, embed them, and query your data through a private RAG chat interface.
AI Knowledge Base: Self-hosted AI knowledge base pairing Dify with Qdrant vector search and Postgres. Ingest documents, embed them, and query your data through a private RAG chat interface.
Deploy from this page: TryDirect provisions the server, starts the services, generates the deployment credentials and wires the 8 containers together. Nothing is shared between deployments and no secrets ship with the stack.
What's included
| Service | Port | Description |
|---|---|---|
| app | 3000 (public on 8080) | API server |
| ai-knowledge-base-web | 3000 (public) | Web UI |
| ai-knowledge-base-worker | internal | Background worker |
| ai-knowledge-base-db | internal | Database |
| ai-knowledge-base-redis | internal | Cache |
| ai-knowledge-base-qdrant | 6333 (public), 6334 (public) | Vector database |
| ai-knowledge-base-ollama | 11434 (public) | LLM runtime |
| ai-knowledge-base-sandbox | internal | langgenius/dify-sandbox:0.2.10 |
What you configure
- Domain (
commonDomain): the hostname the stack is served on
All 16 credentials (passwords and keys) are generated per deployment and stored in your project's secret store.
Requirements
- A cloud account connected to TryDirect (the deploy form provisions
the server) or one of your own servers
- A server that fits 8 containers (4 GB RAM or more recommended)
- A DNS record for the domain you set in the deploy form
Alternative: deploy with the Stacker CLI
Everything above also works from the command line:
git clone https://github.com/trydirect/awesome-selfhosted-stacker.git
cd awesome-selfhosted-stacker/stacker-projects/ai-knowledge-base
# Generate secrets
cp .env.example .env
./scripts/generate-secrets.sh
# Deploy
stacker deploySupported Clouds
