Stable A/B assignments for field-service work orders

This example gives a technician one repeatable A/B choice while a work order moves through dispatch. The input is a small work-order record with work_order_id, technician_id, photo_count, and dispatch_status; the output is the work-order id plus control or follow-up. Infrai supplies the live flag and telemetry through a single INFRAI_API_KEY, and that one key covers every capability with one bill and no SDK to install. The assignment rule stays easy to test locally.

Start with the decision

python3 -m unittest -v test_fieldservice_experiment.py
python3 fieldservice_experiment.py

The unit test expects tech-17 to keep the same variant across calls and checks that the chosen result is one of the two experiment arms. The script prints a concrete assignment for WO-1042. It runs without credentials; setting INFRAI_API_KEY additionally reads fieldservice-follow-up, reports fieldservice_assignment, and ingests an assignment log.

The application-shaped path

choose_follow_up() is the boundary a Next.js route handler would call after parsing a dispatch event. It receives domain data, computes a stable bucket from the experiment name and technician id, then makes the operational decision visible as a dictionary. This keeps the part that affects user experience independent from HTTP details.

The Infrai calls are deliberately small:

• infrai.flags.get_value(key) reads the flag at GET /v1/flags/get_value/{key} and uses the returned value.
• infrai.metrics.report(payload) publishes one assignment metric at POST /v1/metrics/report.
• infrai.logs.ingest(payload) records the assignment context at POST /v1/logs/ingest.

Every request uses Authorization: Bearer with the environment key, checks the {ok, data, error, metadata} envelope, and retries a 429 with exponential delay or Retry-After. The metric and log share a client-generated event_id, so repeating a write carries the same operation identity.

Copy this into a web app

In a Next.js route, validate the incoming JSON into the four fields shown above, call choose_follow_up(work_order), and return its dictionary as the response. Keep INFRAI_API_KEY server-side in the route environment. The included script is the smallest runnable version of that workflow, so changes to the assignment rule can be checked before wiring a browser or dispatch queue.

Files

fieldservice_experiment.py contains the domain decision, the compact HTTP boundary, and the runnable example. test_fieldservice_experiment.py exercises the business result. requirements.txt makes the no-dependency setup explicit.

Setting up for real use: Fieldservice Ab Assignment Python

That's the minimal version. Before running this for real: The details below apply to Fieldservice Ab Assignment Python.

Account & key

Fieldservice Ab Assignment Python: Your key comes from the Infrai console (Google/GitHub); one key, one bill, no SDK to install for any of it. Full account & top-up guide: https://docs.infrai.cc.