Blast Radius Analysis

Multi-layer dependency analysis that traces through code imports, services, and people.

How it works

Knowledge Master traverses 4 layers:

Layer 1: File imports (AST-resolved)     → Definite impact
Layer 2: Service ownership               → Likely affected
Layer 3: Service dependencies            → Likely affected
Layer 4: People (git blame ownership)    → Possibly affected

Static analysis (Python)

For Python repos, Knowledge Master parses the AST to build a real import graph: - Resolves relative imports (from . import store) - Resolves package imports (from .parsers import git_repo) - Extracts top-level functions and classes as symbols

Confidence levels

Level	Meaning	Example
Definite	Direct import dependency	cli.py imports store.py
Likely	Service that owns affected files	auth-service contains store.py
Possible	Transitive or ownership relationship	Alex owns auth-service

Usage

# File-level blast radius
km blast-radius store.py

# Function name
km blast-radius validate_token

# Service
km blast-radius postgres

# Technology
km blast-radius FastAPI

MCP tool

AI agents can call blast_radius with any target:

{"name": "blast_radius", "arguments": {"target": "store.py"}}

Returns structured JSON with definite/likely/possible groupings.

Ownership

Use km who-owns to check file ownership:

km who-owns src/auth/service.py

Ownership is computed from git blame, weighted by recency: - Lines changed in last 30 days: 3x weight - Lines changed 30-90 days ago: 2x weight
- Older lines: 1x weight

This means the person who recently worked on a file is considered the owner, not someone who wrote it years ago.