Support triage

Support ticket classification: triage every ticket automatically

How to use AI to classify support tickets by category, urgency and sentiment, route them to the right queue in Zendesk, Intercom or Freshdesk, and escalate the ones that can't wait.

Content moderation

AI content moderation with your own rules

How to moderate comments, reviews, listings and messages with AI using your own community rules: allow, send to review or remove, with probabilities you can threshold.

Document classification

AI document classification: know what every file is

How to classify documents such as invoices, receipts, contracts, resumes and purchase orders with AI before extraction and routing, including OCR tips and handling unknown document types.

Bulk classification

Classify spreadsheet rows with AI: Excel and CSV

How to categorize thousands of rows in Excel, Google Sheets or CSV with AI: product feedback, survey answers, transactions, leads or listings, with a confidence score per row.

Sentiment analysis

Sentiment analysis API with your own scale

How to run AI sentiment analysis on reviews, tickets, surveys and social posts with a custom scale, aspect-level sentiment and confidence, via a simple API.

AI email classification

AI email classification that routes itself

Classify incoming email into sales, support, billing, spam and newsletters, flag urgent messages, and route them automatically with a fast AI decision layer.

AI lead qualification

AI lead qualification in real time

Qualify and score inbound leads the moment they arrive: fit score, industry and buying intent, routed to sales, nurture or disqualified automatically.

AI agent guardrails

Guardrails for AI agents, in milliseconds

Add a fast, low-cost decision layer in front of your AI agent's tool calls: allow safe actions, block dangerous ones, and send risky ones to human review.

Classification API

AI classification API: classify any text in one request

How an AI classification API works, when to use one instead of prompting an LLM or training a model, and how to classify text and JSON into your own labels with calibrated confidence.

LLM routing

LLM routing: pick the right model for every request

What LLM routing (model routing) is, the main routing strategies, and how to build a router that sends simple requests to cheap models and hard ones to strong models without hurting quality.

LLM cost reduction

How to reduce LLM costs without hurting quality

Practical ways to cut your OpenAI, Anthropic or OpenRouter bill: filter before you generate, route to smaller models, trim context, cache, batch and replace generation with decisions where possible.

Intent detection

Intent classification: detect what users want, reliably

How AI intent classification works for chatbots, support and voice apps, how to design an intent list, handle low confidence and out-of-scope messages, and implement it in a few lines of Python.

Zero-shot classification

Zero-shot classification: labels without training

What zero-shot text classification is, how it works with NLI models, LLMs and decision models, how to write labels that classify well, and how to evaluate it before production.

LLM-as-a-judge

LLM-as-a-judge: automated evaluation of AI outputs

How to evaluate AI answers automatically with a judge model: grounding checks, rubric scores, pass/fail gates, calibration against human labels, and when a fast decision model makes a better judge.

Comparison

Jev vs. GPT-style LLMs for classification

A practical comparison of a decision model (TypeSafe's Jev) and general-purpose chat LLMs such as GPT models for classification, scoring and routing: output format, cost, latency and calibration.

Comparison

Description-based AI classification vs. a fine-tuned classifier

When to classify text with a description-based model (LLM or decision model) and when to train or fine-tune your own classifier, compared on data needs, time to ship, accuracy, cost and maintenance.

Jev AI

Jev: TypeSafe's decision model, explained

What Jev is, how TypeSafe's Jev decision model works, what it's good at, its limits, and how to use Jev AI for classification, routing and yes/no decisions.

TypeSafe

TypeSafe AI and decision models

An overview of TypeSafe AI's approach to decision models, the System One idea behind Jev, and when a decision model beats a large language model.

Jev API

The Jev API: request format, question types and responses

A developer guide to the Jev API on OpenRouter's Decisions endpoint: authentication, state, questions, choice/noul/score criteria, the response format and error handling.

Jev + OpenRouter

Using Jev on OpenRouter

How to access TypeSafe's Jev model through OpenRouter: model ids, the Decisions endpoint, credits and billing, and how it differs from chat completions.

Jev pricing

Jev pricing and how to estimate cost

How Jev is priced on OpenRouter (input tokens only), how to estimate cost per decision, and how ClassifierHub's credit plans compare.

Jev examples

Jev examples: decision patterns that work

Practical Jev examples for ticket routing, lead scoring, spam detection, content moderation and AI agent guardrails, with question definitions you can copy.

Jev tutorial

Jev tutorial: build a ticket router

Step-by-step Jev tutorial: define questions, call the Decisions API, set thresholds from labeled examples and handle failures safely.

Jev + TypeScript

Using Jev with TypeScript

Call TypeSafe's Jev from TypeScript with fetch and Zod: typed requests, validated responses, timeouts and threshold logic.

Jev + Python

Using Jev with Python

Call TypeSafe's Jev from Python with requests: choice and score questions, timeouts, and batch processing patterns for data pipelines.