Skip to content
Twelve manila index cards in a grid on an oak desk
Twelve schemas. One catalog.

Jev Patterns catalog

Jev Patterns catalog: twelve schemas, closed answers

Each Jev Patterns entry is a paste-ready System One request: state, typed questions, and the thresholds that belong in your code. Filter by category or primitive.

The Jev Patterns catalog is the index of every reusable decision schema on this site. It is not a prompt gallery. It is not a chatbot template pack. Each card is a TypeSafe Jev request you can copy into production and a fixture you can run in Jev Studio. If you needed “JSON classifier prompts for tickets,” you are in the right problem space and the wrong output type — start with support ticket routing.

12 patterns

SupportCHOICENOUL

Support ticket routing

Send a ticket to the right department and flag whether it should jump the queue.

Open pattern
SupportSCORE

Urgency scoring

Place a ticket on a four-level urgency rubric your SLA code already understands.

Open pattern
AgentsCHOICE

Agent tool selection

Given the current agent state, pick the next tool from a closed toolbox — including stop.

Open pattern
AgentsCHOICE

Agent continue or stop

After a tool result, decide whether to continue, retry, ask the user, or halt.

Open pattern
GuardrailsNOULCHOICE

LLM output guardrail

Decide whether a generated answer is grounded in the sources you actually retrieved.

Open pattern
ModerationNOUL

Content moderation screen

Run independent Nouls for spam, harassment, sexual content, self-harm, and malware.

Open pattern
GrowthCHOICE

Lead intent classification

Route an inbound message to demo, pricing, support, partnership, or spam.

Open pattern
OpsNOUL

Refund eligibility

Ask two Nouls: did they request a refund, and does the policy you passed in support one?

Open pattern
OpsNOUL

Spam or duplicate detect

Two Nouls: is this a duplicate of recent tickets, and is it spam?

Open pattern
SupportNOUL

Escalate to human

A single Noul: should a person take this over from the bot or the junior queue?

Open pattern
GuardrailsSCORE

RAG chunk relevance

Score each retrieved chunk on a 0–3 rubric before it is stuffed into the writer prompt.

Open pattern
GuardrailsCHOICE

Eval judge

Given a claim and a passage, choose supports, contradicts, or says nothing.

Open pattern

How to choose from the Jev Patterns catalog

If the next line of code is a switch on a queue, tool, or intent, start with a Choice pattern. If you already have an ordered rubric (SLA, relevance 0–3), start with a Score. If the question is a boolean you will threshold — urgent, grounded, needs a human — start with a Noul. Many production jobs mix two primitives in one call; the catalog marks those on the card.

Support: routing, urgency, escalate to human. Agents: tool selection, continue or stop. Guardrails: LLM output, RAG chunks, eval judge. Moderation: parallel harm Nouls. Growth: lead intent. Ops: refunds and duplicates. Read the Jev guides before you auto-apply. Compare Jev vs JSON mode if you are migrating off a parser. Jev Patterns is independent and not affiliated with TypeSafe AI.

What “starting point” means in the Jev Patterns catalog

Criteria text is ours, not a legal taxonomy. Your queues will differ. Rename keys. Add other. Delete sales if you have no sales queue. Keep the primitive. The Jev Patterns catalog is useful when the shape is right and the nouns are wrong — that is a five-minute edit, not a research project. What you should not edit casually: mixing urgency into a department name, or asking Jev to write the reply in the same request. Those are the rails.

Every card links to Studio with ?pattern=slug. Use that before you bikeshed JSON. The fixture bars are canned; the request JSON is the thing you copy. Official endpoint and model names stay on the pattern page so you do not have to memorize them.

Next

After you pick a schema

Jev Patterns catalog FAQ

What is in the Jev Patterns catalog?
Twelve paste-ready TypeSafe Jev schemas: support ticket routing, urgency scoring, agent tool selection, continue or stop, LLM output guardrail, content moderation screen, lead intent classification, refund eligibility, spam or duplicate detect, escalate to human, RAG chunk relevance, and eval judge.
Are Jev Patterns official TypeSafe examples?
No. The Jev Patterns catalog is independent and not affiliated with TypeSafe AI. Official docs: docs.typesafe.ai.
How do I run a schema from the Jev Patterns catalog?
Open the pattern, copy the request JSON, POST to api.typesafe.ai/v1/systemone with jev-latest — or load it in Jev Studio and run the fixture first.
Can I filter the Jev Patterns catalog?
Yes. Filter by category (support, agents, guardrails, moderation, growth, ops) or by primitive (Choice, Score, Noul).
Do Jev Patterns generate text?
No. Every schema in the Jev Patterns catalog returns typed decisions. Use a writer model if you need a paragraph.

Last reviewed 21 September 2026. Independent of TypeSafe AI.