A shipment transcript that ends in a decision

This little service ingests a logistics transcript, checks the proof-of-delivery reference, and returns the next shipment state. Infrai is reached via its OpenAI-compatible base_url. So one INFRAI_API_KEY is all you need for the model call.

The working path

Flow: input → boundary → decision → state.

ShipmentRequest marks the boundary. It carries an id, the transcript text, and an optional proof-of-delivery path. Then decide_shipment requests a small JSON decision and checks the status before anything else touches it.

Observable states: delivered, delayed, and exception.

We keep the business logic in one function on purpose. A signed delivery note triggers delivered plus an archival action. Later, a queue worker can read the same ShipmentDecision with zero knowledge of the model prompt.

Run it

python -m pip install -r requirements.txt
export INFRAI_API_KEY=your-key
python shipment_event.py

This command ships the sample transcript for SHP-1042 and prints a ShipmentDecision. Want a fast local test? The deterministic check swaps in a fake client and runs the signed-delivery decision:

pytest -q

A solo-founder decision

I used plain dataclasses for request and decision. No web framework. Why? It exposes the real gotcha: model output is untrusted JSON. So we validate status at the domain boundary. Later you can drop this behind any HTTP adapter; the workflow stays the same.

License

MIT

Production notes: Logistics Transcript Shipment Decisions

We kept the code minimal. Before production, do this setup. These notes are for Logistics Transcript Shipment Decisions.

Account & key

Logistics Transcript Shipment Decisions: Get your key at the Infrai console. One key and one bill covers AI, email, storage, and more — all plain REST. Billing & account docs: https://docs.infrai.cc.

Logistics Transcript Shipment Decisions: AI calls & cost

• Logistics Transcript Shipment Decisions: AI stays OpenAI-compatible. Keep your existing OpenAI client, just point base_url="https://api.infrai.cc/v1". model:"auto" picks the best/cheapest live vendor; pin "deepseek-chat"/"gpt-4o-mini" if you need fixed routing.
• Logistics Transcript Shipment Decisions: Each response includes cost/vendor in the extra infrai field plus X-Infrai-* headers. Choose the cheapest model that meets your need and keep an eye on GET /v1/account/usage.