How epochix-dev/epochix works
Epochix turns existing or live training output into normalized metrics, an interpreted training story, and a dashboard/report. The primary path is input → parsing/normalization → story grading and phases → persisted run data → live delivery or export. The Python pipeline’s ownership is explicit in its module documentation; the README supplies the broader capabilities, while sampled sources confirm key server, frontend, and VS Code paths. Some integration wiring is documented but not sampled, so those links are omitted or marked optional.
Main entry points: pipeline.py Run pipeline · main.js Web dashboard · routes_runs.py Run and snapshot routes · StandaloneEngine.ts Standalone engine
Components 22
Ingest and analysis
- Log ingesters
src/epochix/ingester - Framework parsers
src/epochix/parsers - Metric normalization
src/epochix/normalizer
Story interpretation
- Story and grading
src/epochix/story_engine
Run delivery
- CLI
src/epochix/cli.py - Run pipeline
src/epochix/pipeline.py - Run history
src/epochix/store/sqlite_store.py - Live message hub
src/epochix/server/hub.py - HTTP and stream API
src/epochix/server/app.py - Run and snapshot routes
src/epochix/server/routes_runs.py
Dashboards
- Web dashboard
frontend/src/main.js - Live stream client
frontend/src/ws-client.js - Run comparison
frontend/src/visualizations/CompareView.js - VS Code extension
epochix-vscode/src/extension.ts - Terminal detection
epochix-vscode/src/terminal/TerminalWatcher.ts - Standalone engine
epochix-vscode/src/webview/StandaloneEngine.ts
Integrations and exports
- Export routes
src/epochix/server/routes_export.py - Report exporters
src/epochix/exporters - Framework integrations
src/epochix/integrations
Other components
- ML engineer
- Training output
- LLM service
Connections 29
- ML engineer → CLI: runs command
- Training output → Log ingesters: supplies logs
- CLI → Run pipeline: starts processing
- Log ingesters → Run pipeline: yields lines
- Run pipeline → Framework parsers: selects and parses
- Framework parsers → Metric normalization: produces metrics
- Run pipeline → Metric normalization: normalizes metrics
- Run pipeline → Story and grading: processes events
- Run pipeline → Run history: writes run data
- Run pipeline → Live message hub: publishes updates
- ML engineer → HTTP and stream API: opens dashboard
- HTTP and stream API → Run and snapshot routes: dispatches requests
- Run and snapshot routes → Run history: reads and updates
- Web dashboard → Run and snapshot routes: fetches run data
- Web dashboard → Live stream client: starts live feed
- Live stream client → HTTP and stream API: opens WebSocket
- Live message hub → Live stream client: streams updates
- Web dashboard → Run comparison: renders comparison
- Run comparison → Run and snapshot routes: requests comparison
- Web dashboard → Export routes: requests exports
- Export routes → Run history: loads run data
- Export routes → Report exporters: generates reports
- ML engineer → VS Code extension: uses extension
- VS Code extension → Terminal detection: watches terminal
- Terminal detection → Standalone engine: feeds terminal text
- Standalone engine → Framework parsers: parses framework logs
- Standalone engine → Story and grading: interprets metrics
- Framework integrations → Run pipeline: reports training metrics
- Run pipeline → LLM service: requests fallback extraction