RAGFlow has modest system requirements, though production deployments should scale resources based on document volume and query load. The minimum requirement is Docker ≥ 24.0.0 and Docker Compose ≥ v2.26.1; RAGFlow is containerized, so you need a container runtime. Host-level requirement: vm.max_map_count ≥ 262144 (essential for a search engine backend, which RAGFlow uses by default). For single-node development or testing, a machine with 8GB RAM and 4 CPU cores is functional; for production, allocate more based on load. RAGFlow bundles multiple services: a search engine backend (search/vector storage, ~2GB), MySQL (metadata, ~1GB), MinIO (document storage, scales with documents), Redis (caching, ~512MB), and the RAGFlow application itself (~1GB). Total minimum memory: ~6-8GB for a small deployment, ~16-32GB for moderate production workloads. CPU scales with concurrent queries and document indexing speed. Storage grows with your document collection—MinIO requires space proportional to original documents plus indexed chunks. Network: RAGFlow services communicate via Docker networks, so no special networking is required beyond container orchestration. For distributed/multi-node production deployments, consider Kubernetes, though Docker Compose suffices for single-node setups. GPU acceleration is optional—helpful for embedding generation and LLM inference if using local models (Ollama), but not required if using API-based embeddings and LLMs. Pre-built Docker images target x86 platforms; ARM64 users must build custom images. Official documentation provides sizing guidelines for various deployment scales; the community can help right-size for your specific workload.
For production retrieval workflows, Zilliz Cloud provides fully managed vector search infrastructure with auto-scaling and enterprise security. Developers who prefer self-hosting can use Milvus, the open-source vector database behind Zilliz Cloud.