dsh-jev-adapter
DeepSeek Harness jev_decide tool that brings the Jev decision paradigm to any OpenAI-compatible LLM, or to real Jev with a key.
Overview
dsh-jev-adapter registers a jev_decide tool on DeepSeek Harness that lets the agent ask typed questions in the Jev style: choice over up to 255 options, score over 2 to 10 ordered levels, and boolean statements returning P(true). Many questions can be asked in one call at roughly the cost of one.
Its point is reach: by default it uses an openai channel that works with any OpenAI-compatible endpoint, such as DeepSeek's own API, Ollama, vLLM, OpenRouter, SiliconFlow, LM Studio or an air-gapped gateway, so no TypeSafe key is needed. On that channel probabilities are the model's self-reported estimates and not calibrated, which every result notes. A typesafe channel talks to the real Jev API for those who have a key. The README points to dsh-jev for official API integration.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Any choice question with up to 255 options | — |
| Score | Any rating on 2 to 10 ordered levels | — |
| Yes / no | Is this statement true? | true · false |
How it works
- Agent calls jev_decide with state and questions
- Plugin sends them to the configured channel (openai or typesafe)
- Receives picks, distributions and confidence
- Agent acts on the typed answers
Strengths
- Works without a TypeSafe key
- Supports air-gapped or local endpoints
- Can switch to real Jev when available
Caveats
- Probabilities on the openai channel are self-reported, not calibrated
- Specific to DeepSeek Harness
Getting started
dsh plugin --profile <name> add github:BetterZflyee/dsh-jev-adapter, then export JEV_OPENAI_API_KEY and restart dsh.At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- DeepSeek Harness plugin
- Has code
- yes
Use cases and integrations
How curated lists describe it
Use the Jev (System One) decision-model paradigm with any OpenAI-compatible LLM — no TypeSafe key required. A jev_decide tool for DeepSeek Harness (dsh).
hellogumbo/awesome-jev
Frequently asked questions
Do I need a TypeSafe key?
No, the default channel works with any OpenAI-compatible LLM.
Are the probabilities calibrated?
Only on the typesafe channel; the openai channel returns self-reported estimates.
What tool does it add?
jev_decide, supporting choice, score and boolean questions.
Related projects
opencode plugin that routes the agent's small multiple-answer decisions to Jev and labels each answer auto or escalate by confidence.
PluginbetaTypeScriptDeepSeek Harness plugin that gives the human a Jev second opinion beside the agent's multiple-choice questions.
ProjectbetaTypeScriptDeepSeek Harness plugin registering a jev_decide tool so the agent can ask Noul, Choice or Score questions and gate on probabilities.
PluginbetaTypeScriptDeepSeek Harness plugin with 12 opt-in Jev or Luna Decisions judgments: skill and file selection, supervision, filtering and approval help.
PluginbetaTypeScriptSelf-hosted Yao Agents platform for agents, workspaces and task boards across devices, which announces Jev support.
ProjectbetaReads candidate logits from existing MLX-LM/VLM models on Apple Silicon to map state and a request to an allowed action.
ProjectprototypePython