Seed from Git History¶
Populate your distill knowledge base from an existing repository's git history. This extracts decisions, patterns, failures, and context from commits.
Prerequisites¶
- distill MCP server running
- Inside a git repository with commits
Quick start¶
Option 1: CLI (recommended)¶
distill seed
This prints the seeding workflow for your AI assistant to follow. Pipe it into your agent or copy-paste it into a conversation.
Option 2: Claude Code skill¶
If you have the distill skills installed in Claude Code, use the /seed command directly in your conversation.
How it works¶
The seed workflow:
- Surveys your repo — counts commits, shows first and last 20
- Processes commits in batches of 10
- Classifies each commit as decision, pattern, failure, dependency, or context
- Stores distilled knowledge via the
remembertool - Verifies entries after each batch
What gets extracted¶
| Type | What to extract | Example |
|---|---|---|
decision |
Chose X over Y because Z | "Replaced SQLite with asyncpg for stateless GKE deployment" |
pattern |
Recurring convention | "All services use structlog with JSON output to stderr" |
failure |
Tried X, abandoned because Y | "Tried Redis for persistence, reverted — too complex for single-node" |
dependency |
Service A depends on B | "Parser requires pgvector extension >= 0.5" |
context |
Migration/state at a point in time | "Auth middleware rewrite started, driven by compliance requirements" |
What gets skipped¶
chore:formatting, linting, dep bumps without rationale- Merge commits with no original content
fix:typos, whitespace, trivial CI tweaks
Tips¶
- Quality over quantity — skip noise, capture reasoning
- Capture the WHY — "added postgres_store.py" is useless; "chose PostgreSQL for stateless deployment" is valuable
- Connect related commits — a decision and its later reversal should reference each other
- Use date ranges — for large repos, seed in chunks with
--since