Classifiers
Reusable decision classifiers for decision models such as TypeSafe's Jev and the open-weight Kev. Each classifier asks one constrained question about a state, and your code acts on the answer. Chains show how classifiers combine, with branches driven by outcomes.
Educational catalog. Nothing here calls a model. Example outcomes are illustrative judgments written for teaching, not model outputs, scores or accuracy claims. Sourced documentation is labelled separately.
Classifiers
12 classifiersIn-scope request check
Decide whether a message is about something your team supports before trusting a routing choice.
Support ticket routing
Send each support message to the team that can resolve it.
Ticket urgency
Rate how soon a message needs a response, and send unclear cases to a person.
Refund request detection
Tell whether a customer is actually asking for money back.
Refund information check
Find out what a refund request is missing before anyone decides on it.
Refund policy eligibility
Read a refund request against your written policy and recommend a decision.
Community policy category
Name the part of your community policy a post most likely breaks, if any.
Harm severity
Rate how much harm content could cause if it stays visible.
Prompt injection signal
Flag content that tries to give instructions to an AI assistant, as one layer among several.
Agent tool routing
Choose which tool an assistant should use next, from a fixed list.
Retrieved passage relevance
Score how useful each retrieved passage is for answering the question.
Answer groundedness
Check whether the sources support each claim in a drafted answer.
Decision chains
4 chainsComplete flows that reuse the classifiers above, with branches driven by outcomes.
Support inbox triage
Screen, route and prioritize every incoming support message, and hand refund requests to their own flow.
- In-scope request check
- Support ticket routing
- Refund request detection
- Ticket urgency
Refund request resolution
Confirm a refund is requested, check the request is complete, then recommend a decision under your policy.
- Refund request detection
- Refund information check
- Refund policy eligibility
Community post moderation
Sort posts by policy area, then decide how visible they stay based on potential harm.
- Community policy category
- Harm severity
Grounded support assistant turn
Screen a message, choose a tool, keep relevant passages and check the drafted answer before it is sent.
- Prompt injection signal
- Agent tool routing
- Retrieved passage relevance
- Answer groundedness