Cognee: Knowledge-Graph Memory Over Your Documents

The Stack

Tested Against

Prerequisites

• Python 3.10+
• pip
• LLM_API_KEY for the ingest/recall steps

Steps

1. 1

   Install and verify the four-verb API surface

   pip install cognee, then confirm the public API the docs promise: remember, recall, forget, and improve all exist and the async API resolves. CI runs exactly this, no key.

   python3 -m venv .venv
   .venv/bin/pip install -q cognee
   .venv/bin/python - <<'EOF'
   import inspect
   import cognee
   
   verbs = ["remember", "recall", "forget", "improve"]
   missing = [v for v in verbs if not hasattr(cognee, v)]
   assert not missing, f"cognee is missing documented verbs: {missing}"
   assert all(callable(getattr(cognee, v)) for v in verbs)
   print("cognee API OK: remember, recall, forget, improve all present")
   EOF

2. 2

   Ingest and recall (the model step, not checked by CI)

   Set LLM_API_KEY, then await cognee.remember(...) to build the graph and cognee.recall(...) to query it. The graph build runs on an LLM, so CI never claims it.