How seif-douida/patch works
PatchPulse is an analyst-agent foundation for explaining Steam review-sentiment changes and estimating patch effects; the README says the project is still in Phase 0. Its implemented workflow is a nightly run that fetches and classifies Steam data, stores it, invokes dbt, exports dashboard JSON, and publishes it. The graph retains the documented analytics stages and source-indexed modules; unsampled dbt model internals and some runner calls are represented cautiously rather than inferred in detail.
Main entry points: runner.py Nightly orchestration · http.py Steam HTTP client · engine.py SQL database · cli.py Pipeline commands
Components 23
Steam ingestion
- Tracked games
config/games.yaml - Game configuration
src/patchpulse/ingest/games.py - Steam HTTP client
src/patchpulse/ingest/http.py - Review paging
src/patchpulse/ingest/reviews.py - News fetching
src/patchpulse/ingest/news.py - Price fetching
src/patchpulse/ingest/prices.py - Patch classification
src/patchpulse/ingest/patches.py - Review coverage
src/patchpulse/ingest/coverage.py - Typed records and privacy
src/patchpulse/ingest/models.py
Nightly pipeline
- Pipeline commands
src/patchpulse/pipeline/cli.py - Nightly orchestration
src/patchpulse/pipeline/runner.py - dbt execution
src/patchpulse/pipeline/dbt.py
Persistence
- SQL database
src/patchpulse/db/engine.py - Raw and ops tables
src/patchpulse/db/tables.py - Ingest persistence
src/patchpulse/ingest/load.py
Analytics and export
- Staging, core, marts
dbt/models - Dashboard JSON export
src/patchpulse/export/site.py - Export publisher
src/patchpulse/pipeline/cli.py
Configuration and operations
- Pipeline settings
src/patchpulse/config.py - Cost guard
src/patchpulse/monitoring/cost_guard.py
Other components
- Studio analyst
- Steam
- Published dashboard data
Connections 30
- Studio analyst → Published dashboard data: inspects
- Tracked games → Game configuration: loaded by
src/patchpulse/ingest/games.py - Pipeline commands → Nightly orchestration: starts nightly
src/patchpulse/pipeline/cli.py - Pipeline commands → SQL database: connects through
src/patchpulse/pipeline/cli.py - Pipeline commands → Pipeline settings: loads settings
src/patchpulse/pipeline/cli.py - Nightly orchestration → Game configuration: loads games
src/patchpulse/pipeline/runner.py - Nightly orchestration → Steam HTTP client: uses client
src/patchpulse/pipeline/runner.py - Nightly orchestration → Review paging: fetches reviews
src/patchpulse/pipeline/runner.py - Nightly orchestration → News fetching: fetches news
src/patchpulse/pipeline/runner.py - Nightly orchestration → Price fetching: fetches prices
src/patchpulse/pipeline/runner.py - Nightly orchestration → Review coverage: plans coverage
src/patchpulse/pipeline/runner.py - Nightly orchestration → Patch classification: classifies news
src/patchpulse/pipeline/runner.py - Nightly orchestration → Ingest persistence: stores fetched data
src/patchpulse/pipeline/runner.py - Review paging → Steam HTTP client: requests pages
src/patchpulse/ingest/reviews.py - Review paging → Typed records and privacy: creates records
src/patchpulse/ingest/reviews.py - News fetching → Steam HTTP client: requests news
src/patchpulse/ingest/news.py - News fetching → Typed records and privacy: creates records
src/patchpulse/ingest/news.py - Price fetching → Steam HTTP client: requests prices
src/patchpulse/ingest/prices.py - Price fetching → Typed records and privacy: creates records
src/patchpulse/ingest/prices.py - Steam HTTP client → Steam: sends requests
src/patchpulse/ingest/http.py - Ingest persistence → Raw and ops tables: writes tables
src/patchpulse/ingest/load.py - Ingest persistence → SQL database: writes to
- Nightly orchestration → dbt execution: runs dbt
src/patchpulse/pipeline/runner.py - dbt execution → Staging, core, marts: builds models
- Dashboard JSON export → SQL database: reads marts
src/patchpulse/export/site.py - Nightly orchestration → Dashboard JSON export: builds export
src/patchpulse/pipeline/runner.py - Nightly orchestration → Export publisher: publishes files
src/patchpulse/pipeline/runner.py - Export publisher → Published dashboard data: writes JSON
- SQL database → Pipeline settings: configured by
src/patchpulse/db/engine.py - Cost guard → SQL database: guards spend