Issue evidence · Run · infiniflow/ragflow

MCP server: first retrieval on large datasets hangs for minutes / times out (v0.27.2)

GitHub issue: https://github.com/infiniflow/ragflow/issues/20919

Not run

The first ragflow_retrieval call after MCP server start paginates every document's metadata, so on an 11,701-document dataset it exceeds 300 s.

What Badgr ran

Nothing was run: reproducing needs a full RAGFlow v0.27.2 deployment (document store, database, MCP server) plus a synthetic dataset, and no container runtime was available in the test environment.

What came back

  • The report gives ~45.7 s cold against ~4 to 6 s warm on smaller datasets, and attaches a diff.

Per the test plan, the 11,701-document case would not be claimed even if a smaller dataset were run.

Checked 2026-10-14 in Badgr’s local development environment.

← All issue evidence