Skip to content

aider.md

Latest commit

 

History

History
77 lines (53 loc) · 1.91 KB

aider.md

File metadata and controls

77 lines (53 loc) · 1.91 KB

Aider Integration

engram generates a .aider-context.md snapshot from your knowledge graph. Aider reads it as static context at session start — no plugins or special config required.

Setup

1. Generate context file

engram gen-aider -p .

This writes .aider-context.md to your project root with five sections: Architecture (god nodes), Hot Files, Known Issues, Decisions, and Key Patterns.

2. Add to .aider.conf.yml

read:
  - .aider-context.md

Aider loads this file before every chat session. Commit .aider.conf.yml to the repo.

3. Keep it fresh

# One-shot refresh after re-indexing
engram gen-aider -p .

# Auto-refresh on graph changes (watch mode)
engram gen-aider --watch -p .

Watch mode stays alive and regenerates whenever engram detects graph changes.

Typical workflow

# Index your project (first time or after large changes)
engram init .
engram index .

# Generate context
engram gen-aider -p .

# Start Aider — it picks up .aider-context.md automatically
aider

What gets included

Section Source Limit
Architecture God nodes (highest connectivity) Top 10
Hot Files Pattern nodes with type: hot_file Top 10
Known Issues Mistake nodes, ranked by query frequency Top 5
Decisions Decision nodes from last 30 days All
Key Patterns Pattern nodes with confidence >= 0.8 All

HTTP bridge (future)

Once engram ships its HTTP server, you can fetch context dynamically in pre-session scripts:

curl "localhost:7337/query?q=auth+flow" >> .aider-context.md

This will let you prime Aider with task-specific context before starting a session.

Positioning

Aider builds per-file repo maps from AST analysis. engram adds the persistent layer: architectural decisions, known failure modes, and recurring patterns that survive across sessions. They complement each other.