sqlite-vectordb - SKILL.md Agent Skill

name: sqlite-vectordb description: SQLite vector DB for work log storage and semantic search. Use for indexing work logs, generating embeddings, semantic search, and DB maintenance.

SQLITE VECTOR DB SKILL

Store work logs in a searchable vector database and provide semantic search infrastructure.

When to use:

Add entry: After /log-work execution (integrated with work-logger)
Search: When previous work context is needed (called by work-context-finder)
Delete: To clean up incorrect/outdated logs

Note: DB initializes automatically. No need to run init_db.py manually.

Add Entry

Index a markdown log into the vector DB.

uv run .claude/skills/sqlite-vectordb/scripts/add_entry.py \
  --file "private-docs/work-logs/YYYY-MM-DD-slug.md" \
  --summary "One-line summary" \
  --tags "tag1,tag2"

--file, -f (required): Work log file path
--summary, -s (required): One-line summary for search indexing
--tags, -t (required): Comma-separated tags

Search

Semantic similarity search in work logs.

uv run .claude/skills/sqlite-vectordb/scripts/search.py \
  --query "search terms" \
  --limit 5

--query, -q (required): Search query
--limit, -l: Max results (default: 5)
--tag, -t: Filter by tag
--type, -T: Filter by log type
--json, -j: JSON output

Delete Entry

Remove a work log entry from the database.

uv run .claude/skills/sqlite-vectordb/scripts/delete_entry.py \
  --file "private-docs/work-logs/YYYY-MM-DD-slug.md"

Initialize DB (Optional)

Manual schema creation. Usually not needed - other scripts auto-initialize.

uv run .claude/skills/sqlite-vectordb/scripts/init_db.py

DB location: private-docs/work-logs/.vector-db/work-logs.db
Engine: SQLite + sqlite-vec extension
Embedding model: all-MiniLM-L6-v2 (384-dim)
Chunk types: summary, details, challenges, other
Execution: uv run with PEP 723 inline deps
Schema reference: references/schema.md

Language requirement: All data stored in the vector DB MUST be written in English.

Summary: English only (for consistent embedding quality)
Tags: English only (e.g., auth, refactor, not 인증, 리팩토링)
Query: English preferred for optimal search accuracy
Rationale: Embedding model (all-MiniLM-L6-v2) performs best with English text

Exit codes:

0: Success
1: File not found
2: DB error

Common fixes:

DB corrupted: Delete private-docs/work-logs/.vector-db/work-logs.db and re-run (auto-recreates)
Missing deps: Run uv sync

Valid usage:

# Add entry (auto-initializes DB if missing)
uv run .claude/skills/sqlite-vectordb/scripts/add_entry.py \
  --file "private-docs/work-logs/2026-01-16-auth-feature.md" \
  --summary "Implement OAuth2 authentication" \
  --tags "auth,oauth,security"

# Search with filters
uv run .claude/skills/sqlite-vectordb/scripts/search.py \
  --query "authentication login" --limit 3 --tag auth

Invalid usage:

# WRONG: Non-existent file → Exit code 1
uv run .../add_entry.py --file "missing.md" ...

# WRONG: Empty summary → Poor search quality
uv run .../add_entry.py --file "..." --summary "" --tags "..."