Integration setup failures: six recorded examples
The downloads contain six original inputs, their expected labels and one recorded API observation per input revision. Disagreements stay visible.
Provide the error reason, retry state and precomputed duration. Compare reauthentication, a changed integration and a transient failure under the displayed four-week rule.
The same timeout text can lead to different review categories depending on whether retries continue and how long the failure has been observed. The duration bucket is computed outside the model.
Changed is a review label under this disclosed policy, not proof that hardware is permanently broken. A setup-error state means automatic retries stopped; a restart or reload may still recover it. Reauth identifies reported sign-in trouble without guaranteeing a fix.
This original error-reading exercise adapts the setup-failure categories in ha-gutcheck, linked by the source post. The errors and duration buckets are synthetic. The demo does not retry an integration, sign in, remove anything or diagnose physical equipment.

Use the sample to understand the rule, then test your own input for integration setup failures.

Provide the error reason, retry state and precomputed duration.

Open the decision rules and keep the intended distinction between Reauth, Changed, Transient, Unclear. Edit the choices only when testing your own rule.

Use a sample first, then try your own input. Compare the selected label with the other returned probabilities.

Review Unclear before using a result. The label is a decision about the supplied information, not an executed action.
| Original input | Observed label | Leading probability |
|---|---|---|
| Short timeout retry | Transient | 100% |
| Revoked token | Reauth | 100% |
| Service retired | Changed | 100% |
| Long timeout | Changed | 95.0% |
| Stopped trying | Changed | 83.0% |
| Unexplained failure | Unclear | 93.0% |
These are six original synthetic inputs, tested once per request revision. The percentage is the leading option probability, not measured accuracy. Authored expectations and any disagreements remain in the downloadable evidence. 6 of 6 results matched the authored fixture expectations.Verified
The downloads contain six original inputs, their expected labels and one recorded API observation per input revision. Disagreements stay visible.
The single-request SDK runner uses your private TYPESAFE_API_KEY. The offline consumer reads saved observations, checks request hashes and prints a pending output. It does not call the model or change any external system.
# Offline: inspect the downloaded saved observations. No API key is needed.
python preview_choice_workflow.py samples.json saved-api-results.json
# Optional live request: install the SDK and set TYPESAFE_API_KEY privately.
npm install @typesafe-ai/sdk@0.6.0
node run-choice-sample.mjs samples.json case-1
# The live response prints to the terminal; it does not update saved-api-results.json.The input contract and these limits define what the result means.
No. It calls for separate review of the stated setup. The application should preserve configuration and require an explicit decision for any change.
No. The error category is enough. Use redacted error text and a symbolic credential description; a secret value adds no useful evidence.
It describes the distribution over the supplied choices for this request. These six synthetic examples are functional tests with authored expectations, not a production accuracy benchmark.
No. The demo displays a label and distribution. The downloadable consumer produces a pending preview packet. Application code and a separate authorization policy would control any real action.
Edit the example above to see how a changed fact affects the decision, or open the Playground for another task.