I gave an AI assistant a memory of PubMed - and learned why "forgetting" is the hard part

Description and use

The author built a PubMed connector for cognee to automatically ingest and update literature on KRAS G12C resistance, addressing the challenge of remembering new papers.

Problem

A frozen LLM cannot keep up with rapidly evolving cancer research literature, such as resistance mechanisms to KRAS G12C inhibitors.

Outcome

The connector ingests 155 records in 7 s, performs zero re‑fetch on the second sync, and correctly handles 10,000‑result limits, time‑zone offsets, and retractions.

Workflow

• 1. Define PubMed query string
• 2. Call cognee.remember with pubmed_source(term, mindate)
• 3. Cognee ingests XML, parses to Markdown, extracts entities
• 4. Run cognee.search with GRAPH_COMPLETION query
• 5. Agent answers using updated memory

Evidence

The query matches 253 PubMed records in total. Limiting it to records added since 2024 left 155.

Source

The first sync stored all 155 in about 7 seconds, using just two EFetch batches.

Source

The second sync made zero EFetch calls and finished in about 2.4 seconds. Nothing had changed, so nothing was downloaded or sent to the LLM again.

Source

ESearch never returns more than 10,000 PMIDs for a query.

Source

NCBI assigns those dates in US Eastern time, though, so a run just after midnight UTC can miss records added late on the previous Eastern day.

Source