Varina
Chinese multi-seat exploration engine for game and rule-system design; Jev only gates whether to start deep exploration or ask.
Overview
Varina is a Chinese-language engine for exploring complex design problems, such as game mechanics, product rule systems, numeric balance and architecture trade-offs, where the solution space is large. It runs repeated rounds of divergence, fact-checking and consolidation: it locks the question and constraints, reads workspace documents, splits sub-questions, lets multiple seats propose mechanisms and counterexamples using 40 cognitive operators, checks claims against a fact ledger, and produces comparable final options.
Jev plays a narrow role. With an optional TYPESAFE_API_KEY, it is used only as a post-hoc START or ASK decision gate; it does not take part in chat answers, long-form generation or the seats' reasoning. Per the list summary, Jev decides whether an answer deserves deeper exploration and whether a new idea is already on the board. The project uses a Business Source License.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Should deep exploration START, or should the system ASK the user first? | START · ASK |
| Yes / no | Is this new idea already on the board? | yes · no |
How it works
- Lock the question and read workspace materials.
- Jev gates whether to start deep exploration or ask.
- Seats diverge with cognitive operators; claims are checked and deduplicated.
- Produce comparable final design options.
Strengths
- Jev gates cheaply without affecting generated content.
- Structured, fact-checked exploration rounds.
Caveats
- Jev is a minor component; most work uses generative LLMs.
- Business Source License, not open source.
- Documentation in Chinese.
At a glance
- Type
- Project
- Status
- beta
- Language
- JavaScript
- Runtime
- Node web app
- Has code
- yes
- License
- Business Source License
Use cases and integrations
How curated lists describe it
Multi-seat exploration engine for game mechanics and rule systems that diverges, fact-checks and repeats; Jev decides whether an answer deserves deeper exploration and whether a new idea is already on the board. Business Source License, in Chinese.
MrJev/awesome-jev
Frequently asked questions
What does Jev do in Varina?
Only a post-hoc START or ASK gate, plus duplicate checks per the list; it does not generate content.
Is Jev required?
No, the TypeSafe key is optional.
What is Varina for?
Exploring large design spaces for game mechanics and rule systems.
Related projects
AutoGPT agent platform, listed as adding Jev decision blocks for typed routing, filtering, scoring and confidence-gated next actions.
ProjectproductionPythonMulti-agent orchestration layer for Claude Code with opt-in Jev hooks for fast judgment points, decision caching and egress controls.
PluginproductionTypeScriptOh My Pi coding agent (a Pi fork with IDE wiring) with an optional TypeSafe judgment provider for bounded decisions.
ProjectproductionRust- SystemOneHarness★ 193
Agent harness that compiles an environment's finite actions into typed questions for Jev and gates each step by confidence.
CLIbetaPython ComfyUI nodes that branch workflows on typed Choice, Yes/No and Score judgments from Laya, Jev or Cloudflare Clef.
ServicebetaPythonAgent plugin that asks Jev whether the session is at a safe boundary to /compact and can trigger it on Pi and Claude Code.
PluginbetaTypeScript