Search query fidelity: six recorded examples
The downloads contain six original inputs, their expected labels and one recorded API observation per input revision. Disagreements stay visible.
Compare three supplied query strings with the user's request and separate control parameters. Select a faithful candidate or no match without generating new terms or searching the web.
This is selection from code-supplied candidates, not query generation. The model cannot repair a bad candidate by adding missing words; None is necessary when all available strings lose part of the request.
Parameter separation also needs evidence. A phrase such as this week can leave the query only if another input says the freshness constraint is already handled. Removing it without that context changes the user’s request.
This original text demonstration adapts a decision in a repository linked by the cited X post. The examples and explicit uncertainty outcomes are ours. The post is a directory reference, not proof that its author built every linked project. No third-party benchmark, business outcome or external action is reproduced.

Use the sample to understand the rule, then test your own input for search query fidelity.

Compare three supplied query strings with the user's request and separate control parameters.

Open the decision rules and keep the intended distinction between A, B, C, None. 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 uncertain or unexpected results before using a result. The label is a decision about the supplied information, not an executed action.
| Original input | Observed label | Leading probability |
|---|---|---|
| Filter already handled | B | 99.0% |
| Keep a relationship | A | 100% |
| Preserve a year | C | 54.0% |
| Source filter handled | A | 99.0% |
| Every option drops detail | None | 76.0% |
| Missing goal | None | 100% |
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.
Not in this Choice request. It returns an offered label. Generate another candidate in code or a separate writing step, then review it.
The shortest string may remove the actual subject or relationship. Fidelity to the user’s request is the selection rule.
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.