Skip to content

Prerequisites¶

Required¶

Python 3.12, 3.13, or 3.14
UV package manager — for dependency management
Node.js 22.x + npm — for UI clients
At least one search database: Elasticsearch or OpenSearch (or BM25, built-in)
At least one vector database: Qdrant (recommended) or another supported vector DB
At least one property graph database: Neo4j (recommended) — unless using vector-only RAG
OpenAI API key (recommended) or Ollama running locally

Tip

The docker/docker-compose.yaml can provide all databases as Docker containers — no manual installs needed.

Optional (depending on features)¶

LangChain integration — RDF QA fusion retriever and property graph retrievers:
```
uv pip install -e ".[langchain]"
```
ArcadeDB embedded mode — in-process ArcadeDB with bundled JVM (no separate Java needed):
```
uv pip install arcadedb>=26.3.2
```
Enterprise Repositories: Alfresco, SharePoint, Box, CMIS-compliant repository
Cloud Storage: Amazon S3, Google Cloud Storage, Google Drive, Azure Blob, Microsoft OneDrive