OpenAdapt
1,648 stars on OpenAdapt, 258 forksEvaluate a workflow

← All workflow templates

Workflow templateMeasured reference

Automate health insurance claim intake in openIMIS

The insurance reference: a health-facility claim entered once in openIMIS — the open-source system used by national health-insurance schemes — compiled, and replayed with a fresh claim number, with success established only by a direct SQL read of the claim row.

This reference workflow has published run evidence. The trial count, oracle, exact version, and limits appear below.

Published results

Trials
3
Verified / expected halts
3 / 0
Silent incorrect successes
0
Over-halts
0
Model API calls per run
Not measured

Source: Reference environment replays (deliberately not a benchmark). Measured on compiled arm on openadapt-flow 1.11.0 (evidence committed 2026-07-17); agent arm on openadapt-flow 1.19.0 (2026-07-21).

Separate small-N paid-agent run on a different flow build: 3/3 correct, 0/3 over-halt, 0/3 silent incorrect, $0.4793 in estimated model API charges per run; no matched timing matrix, so no comparative claim is made

Provenance

Runs on

openIMIS 25.10, run locally from digest-pinned images with the upstream synthetic demo dataset — fictional insurees, facilities, and tariffs; everything binds to localhost. openIMIS is a browser UI over a supported GraphQL API, so a real openIMIS deployment should prefer the API; the browser demonstration stands in for the many commercial claims platforms that expose no such API.

Source in openadapt-flow →

Demonstrated steps

  1. 1On the blank Health Facility Claim form, enter the insuree number — which resolves the policyholder and her in-force policy
  2. 2Enter the claim number — a parameter, substituted fresh at every replay
  3. 3Enter the main diagnosis code
  4. 4Enter the explanation note — a parameter
  5. 5Add the service (auto-tariffed by the system)
  6. 6Save the claim into status “Entered” — the step before checking and adjudication

Parameters: insurance_no · claim_no · explanation — recorded values are the defaults; every replay can override them.

How the outcome is verified

A direct SQL oracle on the claims database: the run is accepted only when exactly one non-voided claim row exists with the replayed claim number, in status Entered, for the demonstrated insuree and health facility. The costly failure in claims operations is not a crash — it is the claim silently entered twice, or against the wrong policyholder, surfacing weeks later in reconciliation. A duplicate or missing row fails this run loudly instead of reporting success.

  • direct SQL claim-row oracle (exactly one non-voided row, status Entered, right insuree and facility)

Evidence and scope

A reference environment, deliberately not a benchmark: 3/3 compiled replays (measured on Flow 1.11.0, evidence committed 2026-07-17) and, in a separate small-N run, 3/3 paid-agent trials (measured on Flow 1.19.0, 2026-07-21) were correct, with 0/3 over-halt, 0/3 silent incorrect, and $0.4793 in estimated model API charges per run. There is no matched timing matrix or publication protocol.

Questions about this template

What failure does the SQL oracle catch?

The costly one: the claim silently entered twice, or against the wrong policyholder, surfacing weeks later in reconciliation. A duplicate or missing claim row fails the run loudly instead of reporting success.

Source: benchmark/openimis_claims

Is this a benchmark result?

No. The repository describes it as a reference environment, deliberately not a benchmark: there is no matched timing matrix or publication protocol behind the 3/3 compiled replays.

Source: openIMIS claims reference environment

Try it from the command line

The compiler and runtime are open source and MIT licensed. The default healthy path runs locally and makes no generative-model API calls.

pip install openadapt

Install the OpenAdapt launcher, compiler, and included browser capability.

openadapt flow record --url https://your.app --out rec

Record one browser tab and same-origin route set; Ctrl-C to finish. A task that crosses an origin, tab, application, or backend needs separately qualified bundles and external orchestration.

openadapt flow compile rec --out bundle --name my-task

Compile the recording into a deterministic bundle with auto-classified risk per step.

openadapt flow lint bundle --strict

Report coverage gaps. Complete the required identity, effect, idempotency, and postcondition contracts before governed use.

openadapt flow replay bundle --url https://your.app

Rehearse locally. This command does not certify the bundle. Governed certification and execution also require a deployment configuration with the declared policy, identity, effect, and runtime bindings.

Qualify this workflow

Bring your version of this workflow and the record that proves its outcome. We'll map the deployment, verification, shadow run, and supervised rollout.

Evaluate a workflowRead docs