How latchkey-dev/gusset works
Gusset indexes a repository into a dependency graph, then uses that graph to produce verified impact reports, architecture documentation, and documentation-drift findings. In autonomous mode, events pass through invariant guards and workflows; oracle scores govern the autonomy ladder before scoped actions are taken. The local server exposes graph and run information through a browser UI. The README documents optional tracing and self-healing integrations; their internal wiring is only represented where supported by the sampled source.
Main entry points: runner.py Custodian runner · cli.py CLI · impact.py Impact analysis · store.py Dependency graph
Components 22
Repository graph
- Graph indexer
src/gusset/graph/indexer.py - Code extractor
src/gusset/graph/extract.py - Dependency graph
src/gusset/graph/store.py
Upkeep workflows
- Impact analysis
src/gusset/workflows/impact.py - Architecture atlas
src/gusset/workflows/atlas.py - Docs drift
src/gusset/workflows/docsdrift.py - Seed resolution
src/gusset/workflows/seeds.py
Autonomous supervision
- Oracle scoring
src/gusset/oracle/scores.py - Custodian runner
src/gusset/supervisor/runner.py - Autonomy ladder
src/gusset/supervisor/ladder.py - Scoped actions
src/gusset/supervisor/actions.py
Local interface
- HTTP server
src/gusset/serve/server.py - Serve API state
src/gusset/serve/api.py - Live events
src/gusset/serve/heartbeat.py - Browser UI
src/gusset/serve/static/app.js
Optional observability
- PandaProbe tracing
src/gusset/probe/tracing.py - Self-healing
src/gusset/probe/selfheal.py
Other components
- Maintainer
- CI events
- Target repository
- CLI
src/gusset/cli.py - Anthropic model
src/gusset/llm.py
Connections 29
- Maintainer → CLI: runs commands
- CLI → Graph indexer: indexes repo
- Graph indexer → Code extractor: extracts symbols
- Graph indexer → Target repository: reads source
- Graph indexer → Dependency graph: writes graph
- CLI → Impact analysis: runs analysis
- CLI → Architecture atlas: generates docs
- CLI → Docs drift: checks drift
- Impact analysis → Seed resolution: resolves seeds
- Impact analysis → Dependency graph: queries dependencies
- Impact analysis → Anthropic model: requests explanations
- Architecture atlas → Dependency graph: uses graph
- Docs drift → Dependency graph: checks references
- CI events → Custodian runner: dispatches event
- Custodian runner → Impact analysis: runs workflow
- Custodian runner → Architecture atlas: runs workflow
- Custodian runner → Docs drift: runs workflow
- Custodian runner → Oracle scoring: scores run
- Custodian runner → Autonomy ladder: records scores
- Custodian runner → Scoped actions: applies allowed action
- Scoped actions → Target repository: writes artifacts
- Scoped actions → CI events: posts PR output
- CLI → HTTP server: starts server
- HTTP server → Serve API state: routes requests
- Serve API state → Dependency graph: reads graph
- HTTP server → Live events: streams events
- Browser UI → HTTP server: requests data
- Custodian runner → PandaProbe tracing: emits traces
- Custodian runner → Self-healing: uses repair harness