RAGFlow
RAGFlow is an open-source retrieval-augmented generation engine from InfiniFlow that parses source documents with deep layout and structure understanding before feeding retrieved passages to an LLM. It layers in agent-style orchestration so answers can be traced back to cited excerpts from the original documents.
What gets deployed with it
- Qdrant — required Vector storage and retrieval dependency for the initial one-hop dependency model.
Baseline topologies
| Topology | Minimum nodes | Roles |
|---|---|---|
| compact-rag-app | 1 | ragflow |
| app-plus-processing | 2 | ragflow-api, ragflow-worker |
What drives the cost
- Managed documents (millions)
- Approximate source documents managed by the deployment.
- Daily content ingest (GB)
- Daily volume of new or reprocessed content entering the pipeline.
- Peak concurrent queries
- Peak simultaneous retrieval or answer-generation requests.
Where it runs
bare_metal, dedicated, cloud_vm
Size RAGFlow and get a cost estimate