docs-notes-retrieval

Notes Retrieval

Quick Start — Concrete Commands

Three common scenarios using the scripts in scripts/.

Scenario 1: Vault inventory

Before indexing or packaging, understand what is in the vault.

# Full inventory as JSON — one record per note with title, frontmatter, wikilinks, tags, word count, mtime
python scripts/scan_vault.py inventory /path/to/vault > vault-inventory.json

# Tag summary sorted by note count
python scripts/scan_vault.py tags /path/to/vault

# CSV for spreadsheet review
python scripts/scan_vault.py inventory /path/to/vault --format csv > vault-inventory.csv

Scenario 2: Packaged context for an LLM session

Build a markdown bundle ready to paste into a Claude or GPT session.

# Keyword search — include all notes matching "weekly review", budget 40k chars
python scripts/build_context_pack.py /path/to/vault \
    --query "weekly review" \
    --max-chars 40000 > context-pack.md

# Specific notes with heading-level chunking
python scripts/build_context_pack.py /path/to/vault \
    --notes "Projects/Alpha.md" "Projects/Beta.md" \
    --chunk-strategy heading \
    --max-chars 60000 > project-alpha-pack.md

# Most recent notes first, paragraph chunking, written to file
python scripts/build_context_pack.py /path/to/vault \
    --query "architecture decision" \
    --order-by recency \
    --chunk-strategy paragraph \
    --out session-context.md

# Filter on status before recency: drop superseded and archived notes
python scripts/build_context_pack.py /path/to/vault \
    --query "pricing" --exclude-status superseded archived --order-by recency

The pack keeps conflicting notes side by side under a conflict block (path, status, mtime for each) when they declare conflicts_with: [[Other Note]] in frontmatter or share a title under different paths; a group is included or omitted whole. It does not detect contradictions in prose, so still read for them. The footer lists wikilinks in included notes that resolve to no file.