Blitz

Blitz helps choose among named options using the case information you provide. It returns a decision and a probability for each choice.

Use the web form

Enter the evidence in state. For one decision, enter a question and a choices JSON object with two to twenty names and descriptions. The form starts with a complete example you can run or edit.

To evaluate several questions about the same state, paste a JSON array into questions (up to eight items). Each item needs a question and its own choices object; an id helps identify each result. When questions is filled, the single question and choices fields are ignored.

API examples

The choices and questions fields are JSON strings in the Replicate input. These are complete input objects:

One decision

{
  "state": "The deployment failed twice with the same error after identical retries.",
  "question": "What should happen next?",
  "choices": "{\"inspect\":\"Inspect the logs and change the plan before retrying.\",\"retry\":\"Repeat the same deployment unchanged.\"}"
}

Test a hypothesis

{
  "state": "The hypothesis predicts HTTP 401. The latest trace shows HTTP 500, with no authentication step in the failing path.",
  "question": "Does the evidence still support the hypothesis?",
  "choices": "{\"viable\":\"The hypothesis is still consistent with the evidence.\",\"invalid\":\"A necessary prediction was contradicted.\"}"
}

Several questions about one case

{
  "state": "Three unchanged retries produced the same missing-file error. The expected file is absent.",
  "questions": "[{\"id\":\"retry_helpfulness\",\"question\":\"Would another unchanged retry be useful?\",\"choices\":{\"helpful\":\"A retry is likely to help.\",\"unlikely\":\"A retry is likely to repeat the failure.\"}},{\"id\":\"next_step\",\"question\":\"What is the better next step?\",\"choices\":{\"retry\":\"Repeat the same action.\",\"inspect\":\"Inspect the missing file path.\"}}]"
}

Output and limits

A single decision returns decision, probabilities, and truncated. Batch requests return one result per question in results. If truncated is true, some state text was omitted to fit the input limit. Blitz evaluates the supplied evidence; it does not independently verify it. Check decisions and probability thresholds against your workflow before consequential use.