Choice
Choice: route to a defined option
Select a team, category or handler from options you define. Ask MiniMax Code to add an other route and a review queue so an incomplete label set does not silently force a match.
Try Jev
TypeSafe Jev: Build AI Workflows with MiniMax Code
Turn natural-language inputs into decisions your code can use. Build with the TypeSafe / Jev plugin in MiniMax Code for ticket routing, document ranking, value selection and evidence checks.
Start a TypeSafe Jev Workflow
Choose a starting prompt, adapt it to your project, and continue in MiniMax Code.
Workflow starters
Choose a workflow
Four Decision Workflows to Build with Jev
From ticket routing to evidence checks, choose a starting point for your project in MiniMax Code. Images are illustrative concepts, not verified execution results.
Use Choice to route requests to defined queues. Keep the original ticket and send ambiguous cases to a reviewer.
Three Building Blocks for AI Decisions
Jev is TypeSafe AI's System One model: it returns bounded judgments, not prose. Ask MiniMax Code to compose those judgments into your application.
Score
Score: rank with an explicit rubric
Judge relevance or priority on described, ordered levels. Keep weights and sorting in code so you can change product priorities without redesigning the entire workflow.
Try Jev
Noul
Noul: check a focused condition
Get a probability for a yes-or-no question, such as whether a source supports a claim. Use domain-tested rules to flag uncertain cases; a probability is not permission to publish or execute an action.
Try Jev
Where Jev Fits in Your AI Stack
Jev and MiniMax Code have different jobs. Keep generation, decisions and execution boundaries explicit.
MiniMax Code
- Primary role
- Help plan, write and revise the integration.
- Output
- Code, implementation guidance and test drafts for review.
- Uncertainty
- Help design tests and review paths for your domain.
- Execution and safety
- Help implement authentication, validation and approval gates.
Jev
- Primary role
- Evaluate focused questions about supplied state.
- Output
- Choice, Score or Noul values consumed by code.
- Uncertainty
- Return probability signals; Choice and Score also expose confidence.
- Execution and safety
- Supply judgments, not authorization or a guarantee of correctness.
Get Started with Jev in Four Steps
Use the Jev skill in MiniMax Code to move from a clear task to a tested integration. Live TypeSafe API calls require separate access.
- Step.01
Describe the decision
Provide sample inputs, allowed outcomes and review requirements. The launcher passes your task with a Jev skill instruction to MiniMax Code; confirm the skill or plugin in the main app before live use.
- Step.02
Choose the right primitive
Ask MiniMax Code to read the current docs. Use Choice for categories, Score for ordered ratings and Noul for focused yes-or-no questions. Keep exact rules in code.
- Step.03
Build and test the integration
Implement with the official JavaScript/TypeScript or Python SDK, using synthetic fixtures and mocks first. Keep API credentials server-side when you authorize live calls.
- Step.04
Review before connecting
Evaluate normal, ambiguous, no-match and timeout cases. Tune thresholds, permissions and fallbacks before connecting downstream actions, with human approval where needed.
TypeSafe Jev FAQs
Model basics, plugin setup and practical integration boundaries.