# Classify

> A public directory of Jev use cases. Browse real TypeSafe Jev examples for classification, browser agents, guardrails, extraction, and retrieval, with cost, latency, and pipeline notes for each case.

Host: https://classify.st

## Jev directory

- [Jev use cases directory](https://classify.st/): Home of the Jev directory.
- [Classify a job with Jev](https://classify.st/classify): Paste a job and Jev returns one pattern plus confidence.
- [About the Jev directory](https://classify.st/about): What gets listed and why.
- [Sponsor the Jev directory](https://classify.st/sponsor): Book a slot beside the catalogue.

## Jev use case patterns

- [Jev classification use cases](https://classify.st/patterns/classification): Jev Choice and Score questions that assign a label, topic, or hierarchy node. Used for papers, RAG passages, taxonomies, and industry groups.
- [Jev browser agent use cases](https://classify.st/patterns/browser-agent): Browser agents that ask Jev for the next UI action instead of generating a long plan. Fast enough to click through live sites.
- [Jev multimodal use cases](https://classify.st/patterns/multimodal): Structured questions over images: scene, subject, objects, and visibility. Jev returns typed answers instead of free text.
- [Jev extraction use cases](https://classify.st/patterns/extraction): Turn free text into dates, values, function calls, or numeric features. Code keeps the schema; Jev answers the narrow questions.
- [Jev guardrail use cases](https://classify.st/patterns/guardrails): Screen prompts, tool calls, and model output with Noul and Score questions. Use confidence to pass, review, or block.
- [Jev retrieval use cases](https://classify.st/patterns/retrieval): Re-rank shortlists, score passages, and search line by line. Jev sits after cheap lexical search and before the answering model.
- [Jev function calling use cases](https://classify.st/patterns/function-calling): Map a sentence onto typed function names and closed-set arguments. Confidence decides whether code should call the function.
- [Jev research use cases](https://classify.st/patterns/research): Autoresearch loops, citation checks, and feature discovery. Jev turns notes and papers into numbers a model or reviewer can use.

## Jev use cases

- [Classify 1,018 AI research papers](https://classify.st/use-cases/ai-paper-classification): Assign each paper one of 24 topics after a cheap summary step.
- [Zurich to London flights in 7.1 seconds](https://classify.st/use-cases/browser-use-flights): Drive a live flight-search UI with Jev choosing the next browser action.
- [Structured questions over images](https://classify.st/use-cases/multimodal-image-questions): Ask a closed set of image questions and get typed yes/no and category answers.
- [Autoresearch feature discovery](https://classify.st/use-cases/autoresearch-feature-discovery): Propose Jev questions that turn tasting notes into numeric features for CatBoost.
- [Double-checking citations](https://classify.st/use-cases/citation-check): Decide whether a quoted passage actually supports a claim.
- [Classifying RAG passages](https://classify.st/use-cases/classifying-rag-passages): Score each retrieved passage before it reaches the answering model.
- [Date extraction from documents](https://classify.st/use-cases/date-extraction): Ask Jev for the date parts named in a document, then resolve them in code.
- [Function calling from natural language](https://classify.st/use-cases/function-calling): Turn a trading request into a typed function name and closed-set arguments.
- [Hierarchical classification](https://classify.st/use-cases/hierarchical-classification): Walk a deep taxonomy to the right leaf with parallel Choice questions.
- [Re-ranking legal passages](https://classify.st/use-cases/legal-reranking): Re-rank a 30-passage BM25 shortlist with one Jev question per pair.
- [Guardrails for LLMs](https://classify.st/use-cases/llm-guardrails): Screen every inbound and outbound message for hazards and harm.
- [13 questions in one Jev call](https://classify.st/use-cases/parallel-questions): Run a regulatory briefing as one batched Jev request instead of 13 singles.
- [Pick one skill from 182](https://classify.st/use-cases/skill-suggestion): Choose at most one Hermes skill for an agent turn.