What shipped
Freight Exception Triage Studio is a runnable local Python CLI that creates a synthetic freight exception workspace, scores the queue, generates a triage pack, and validates the generated artifacts.
Implemented commands:
python -m freight_exception_triage_studio sample --path /tmp/freight-exception-triage-studio-demopython -m freight_exception_triage_studio triage --path /tmp/freight-exception-triage-studio-demopython -m freight_exception_triage_studio generate-pack --path /tmp/freight-exception-triage-studio-demopython -m freight_exception_triage_studio validate-pack --path /tmp/freight-exception-triage-studio-demo
Generated artifacts include a triage board CSV, exception brief, workflow map, ROI model, human approval policy, pilot rollout plan, and consolidated JSON pack.
Architecture
- Standard-library Python only, with
argparse,dataclasses,json,csv,pathlib, and simple deterministic scoring rules. - Seeded synthetic data is the source of truth for the MVP.
- Scoring separates priority, cost at risk, automation readiness, owner routing, evidence needs, and approval boundaries.
- Validation is local and strict so the demo can be trusted without external services.
- Human approval gates are modeled as first-class output fields and policy language, not as a footnote.
Trimmed scope
- No web UI.
- No database.
- No auth or billing.
- No live TMS, tracking, ELD, carrier, compliance, email, SMS, EDI, browser, or provider integration.
- No model calls or external APIs.
- No real freight records.
Limitations
- Scores are deterministic pilot heuristics, not learned predictions.
- ROI assumptions are simple category-level estimates for discovery conversations.
- The input contract is JSON only.
- Human approvals are represented as policy and validation state; there is no workflow engine or signature capture.
Suggested next steps
- Add an importer for a sanitized CSV export from a TMS or exception queue.
- Add configurable scoring weights for a specific brokerage, shipper, or carrier operations team.
- Add approval audit fields if the pack becomes part of a real pilot.
- Add side-by-side baseline versus pilot reporting after operators review real queue samples.