How gokulkrisnaa-n/courier-dispatch-prediction-and-optimization works
Courier Dispatch predicts courier-to-order pickup times, then minimizes total pickup time with one-to-one assignment. Its main workflow is event replay or live event consumption, snapshot construction, batch scoring through the API, and assignment; a separate offline pipeline cleans LaDe data, builds shared features, trains the model, and creates a monitoring baseline. Monitoring capabilities include operational metrics, prediction performance, and feature drift. Paths and wiring follow the README and provided source index; unsampled subsystem details are represented only where documented.
Main entry points: dispatcher.py Stream dispatcher · main.py Scoring API · data.py Clean pickup data · train.py Model training
Components 17
Offline modeling
- Clean pickup data
src/dispatch/data.py - Shared features
src/dispatch/features.py - Model training
src/dispatch/train.py - Shared configuration
src/dispatch/config.py
Scoring service
- Prediction model
src/dispatch/model.py - API contracts
src/dispatch/api/schemas.py - Scoring API
src/dispatch/api/main.py
Live dispatch
- Stream dispatcher
src/dispatch/stream/dispatcher.py - Dispatch snapshot
src/dispatch/snapshot.py - Assignment optimizer
src/dispatch/assignment.py
Event transport
- Event replay producer
src/dispatch/stream/producer.py - Event bus
src/dispatch/stream/bus.py
Monitoring
- Monitoring baseline
src/dispatch/monitoring.py - Monitoring dashboard
src/dispatch/dashboard.py
Other components
- LaDe pickup data
- Scoring clients
- Redpanda broker
Connections 28
- LaDe pickup data → Clean pickup data: loads and cleans
README.md - Clean pickup data → Shared features: builds inputs
README.md - Shared features → Model training: transforms data
README.md - Clean pickup data → Model training: supplies cleaned rows
src/dispatch/train.py - Model training → Prediction model: saves artifact
README.md - Model training → Monitoring baseline: creates baseline
README.md - Prediction model → Shared features: builds serving features
src/dispatch/model.py - Scoring clients → Scoring API: requests scores
README.md - Scoring API → API contracts: uses contracts
src/dispatch/api/main.py - Scoring API → Prediction model: predicts pairs
src/dispatch/api/main.py - Scoring API → Dispatch snapshot: builds snapshot
src/dispatch/api/main.py - Scoring API → Assignment optimizer: scores pairs
src/dispatch/api/main.py - Event replay producer → Clean pickup data: loads historic data
src/dispatch/stream/producer.py - Event replay producer → Event bus: publishes events
src/dispatch/stream/producer.py - Event bus → Redpanda broker: transports events
README.md - Stream dispatcher → Event bus: consumes events
README.md - Stream dispatcher → Dispatch snapshot: maintains state
src/dispatch/stream/dispatcher.py - Stream dispatcher → Scoring API: requests scoring
README.md - Stream dispatcher → Assignment optimizer: assigns orders
src/dispatch/stream/dispatcher.py - Stream dispatcher → Monitoring baseline: records performance
src/dispatch/stream/dispatcher.py - Monitoring dashboard → Monitoring baseline: shows drift metrics
README.md - Monitoring dashboard → Stream dispatcher: shows operations
src/dispatch/dashboard.py - Monitoring dashboard → Prediction model: shows model status
src/dispatch/dashboard.py - Clean pickup data → Shared configuration: reads settings
src/dispatch/data.py - Shared features → Shared configuration: reads feature settings
src/dispatch/features.py - Model training → Shared configuration: loads settings
src/dispatch/train.py - Stream dispatcher → Shared configuration: loads settings
src/dispatch/stream/dispatcher.py - Event bus → Shared configuration: selects transport
src/dispatch/stream/bus.py