OpenDecision

At a glance

OpenDecision runs local models that answer predefined questions about text, documents or images. It is worth evaluating if your application needs a structured decision, such as which queue receives a request, and you can test the possible answers in advance.

Good fit

Developers classifying inputs into known options who want local inference.

Consider something else when

Skip it if the task needs an open-ended answer or depends on details that do not fit the chosen input mode.

Start here

Read the routing example and create a few synthetic requests, including one with no suitable category. Check the selected value and decide when the application should ask a person.

Cost and requirements

The docs describe a local runtime with default models. Provision memory for your input length and concurrency, and check the selected model’s requirements.

Limits to know

The docs give different context limits for different question types. Long documents are split into passages and images are resized. Measure errors on inputs that resemble the ones your application will receive.

Related work

• Snap (alternative)