Using an LLM as a selector to feed Claude the right notes from a 5k-note Obsidian vault
EXM7777 · x · 2026-09-26
The author shares a retrieval workflow for an Obsidian vault of 5,000+ atomic notes, each with a one-line summary in an index. Claude can't read everything each turn, and keyword search misses notes that make the same point in different words.
The fix: a model called Jev acts as a picker in four steps — read 28 vault descriptions to pick the right vault, judge every note summary in one call, return the top 5 with full text and drop irrelevant ones, then let code load only survivors into Claude's context. The candidate list is author-defined so Jev can't invent notes; each pick carries a confidence score that triggers a fallback to Claude when low.
More from coding & agent
- Grok now drives nearly 60% of agentic trading on Coinbase, dwarfing Claude and custom CLIs — XFreeze · 2026-09-26
- QuackIR: Jimmy Lin's EMNLP Paper Shows RDBMSes Match Vector DBs for RAG Retrieval — lintool · 2026-09-26
- Where to draw the determinism boundary in LLM pipelines — aronchick · 2026-09-26
- The only pipeline shape that survives production: LLM proposes, rules dispose — aronchick · 2026-09-26
- 100-slot Terminal-Bench 2.1 rerun shows Luna 6 far behind Luna 5.6 — s1lverkin · 2026-09-26
- Qwen3.8-Omni-Flash: natively multimodal agent model with 1M-token context — dair_ai · 2026-09-26