JevModel

What JevModel Does

JevModel is a structured decision API powered by the Jev model from TypeSafe AI. Unlike traditional text-generation models, JevModel returns typed, structured answers — Choice (select from your options), Score (assign discrete levels), or Noul (yes/no judgments) — each with confidence probabilities your application can inspect and act on. It's designed for developers who need reliable, deterministic decisions rather than free-form text.

Who It's For & Key Problems Solved

JevModel targets engineering teams building customer-facing applications that require automated triage, content moderation, routing, and classification. Whether you're filtering spam, routing support tickets to the right queue, scoring lead priority, or detecting prompt injection attempts, JevModel provides a single API endpoint that returns only the options you define — with probabilities you can branch on in your code.

Key Features & Capabilities

• Three Decision Types: Choice (multi-class classification), Score (ordinal ranking), Noul (binary judgment)
• Decision Flows: Chain multiple questions on a visual canvas; one API call executes the entire flow, billing only the questions on the path taken
• Batch Processing: Upload CSV, TXT, or JSONL files (up to 1,000 rows) to evaluate model performance on your own data before production deployment
• Web Console: Manage API keys, test questions in the playground, design flows, review recent runs, and monitor your token balance from a unified dashboard
• Flexible Pricing: Free tier with 10 daily runs after email verification; subscription plans refresh tokens monthly, and token packs never expire
• Custom Criteria: Define your own labels, questions, and routing logic to match internal workflows, product categories, or policy rules

Typical Workflows & Use Cases

Common applications include content moderation (toxic/safe), spam filtering, support ticket intent classification and routing, lead scoring, language detection, personal data screening, prompt injection detection, and sentiment analysis. Teams typically prototype decisions in the playground, validate thresholds using batch evaluation, then integrate via REST API.

Integrations

JevModel exposes a straightforward REST API with official examples in Python and JavaScript. It also integrates with Codex through an official TypeSafe skill. The same account balance powers playground experiments, batch jobs, and production API calls.

What Differentiates JevModel

Unlike general-purpose LLMs, JevModel enforces structured output and returns probabilities for every answer, enabling confident automation and human-in-the-loop branching when certainty is low. Decision flows let you compose complex logic without writing orchestration code, and batch tooling ensures you can measure accuracy on real data before committing to thresholds.