Cases & corpora

A case is a corpus indexed in VOL, with its own RAGFlow dataset, chat assistant and knowledge map. Every chat, search and graph call takes a case identifier.

case	Corpus	Nodes	Graph
papers	Scientific papers	7,656	ragflow_v2
cordis	EU-funded research projects (CORDIS)	83,699	ragflow_cordis
patents	Patents	—	pending
trials	Clinical trials	—	pending

Cases without a published graph are returned with available: false; they cannot be queried yet.

Discovering cases

GET /api/v1/cases is public and returns the live list, so your UI never has to hard-code it:

[
  {
    "id": "papers",
    "name": "Scientific papers",
    "description": "Peer-reviewed literature on AI, knowledge graphs and retrieval.",
    "nodes": 7656,
    "available": true,
    "graphVersion": "ragflow_v2"
  }
]

graphVersion changes whenever the layout is recomputed. Use it as a cache key for the binary graph files.

Node indices

Within a case each document is a node with a stable integer nodeIndex (its position in positions.bin and labels.json) and a string nodeId such as paper_123. Citations, search results, subgraphs and clusters all refer to nodes by nodeIndex.