Your question.
Its decision.
Define a classifier in one request. Start with an example or describe your own workflow — without training a model.
Your result will appear here.
Run a request to see the model's decision and answer probabilities.
Real inference · no precomputed resultsThe model is making a decision.
Sending request…
The first call or a GPU queue can increase waiting time.
Expected action costs
Answer probabilities
Source excerpts evidence
Highlights are the model's pointers into the original context. They do not guarantee that the decision is correct.
facts: auto is enabled. Before model evaluation, code calculates recognised dates and amounts and adds the results to the context as supporting facts.
Probabilities describe the model's assessment. Validate results on examples from your own process.
Input request JSON
Model response JSON
This form exposes five schemas and up to 10 options per question. The full basal-2.0 API also supports the tool type and up to 255 options. One model, many workflows. Change the question and answer descriptions to build another classifier. The demo accepts text only; images and audio are supported by experimental basal-omni, running on your own GPU.