dsh-jev-plugin
DeepSeek Harness plugin with 12 opt-in Jev or Luna Decisions judgments: skill and file selection, supervision, filtering and approval help.
Overview
deepseek-harness-jev connects DeepSeek Harness to Jev, or OpenAI's Luna Decisions, as a judgment layer. Its 12 features are individually configurable from one settings page and all disabled by default. They cover agent skill and file selection (asking a Noul for each skill summary or glob path and ordering candidates by probability), task supervision, completion checks, shared-finding corrections, tool-output filtering, single-operation approval assistance and historical stage navigation.
The main DSH model still plans, answers and calls tools; enabled features apply Jev results at public Cordis and DSH extension points without modifying the host. Jev and Luna Decisions are reachable through OpenRouter or OpenAI, with separate saved connections per profile. A Chinese website explains each integration point, what data is sent, and observed test cases and limits, and a public test report compares the two decision providers. It is an early-stage community project with a bounded DSH compatibility range.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this skill relevant to the current task? | yes · no |
| Yes / no | Is this glob path relevant to the request? | yes · no |
| Yes / no | Should this single operation be approved?(inferred) | yes · no |
How it works
- Enable chosen features in the Jev settings page
- At DSH extension points, send candidates or context to the decision provider
- Order skills or paths by probability, or assist approvals
- Main model continues planning with the filtered inputs
Strengths
- Everything opt-in and individually configurable
- Supports Jev and Luna Decisions
- Detailed documentation of data sent and limits
Caveats
- Early-stage; bounded DSH compatibility
- Judgments are not a correctness guarantee
Getting started
Install through the DSH web UI as described in the README, then configure a Jev or Luna Decisions connection on the settings page.At a glance
- Type
- Plugin
- Status
- beta
- Language
- TypeScript
- Runtime
- DeepSeek Harness plugin
- Has code
- yes
Use cases and integrations
How curated lists describe it
Coding agents: asks Jev a `Noul` for each skill summary or original glob path, orders candidates by the returned relevance probability, and supplies the selected skills or ranked paths to DeepSeek Harness, with all features disabled by default.
yibie/awesome-jevNative DeepSeek Harness (DSH) plugin integrating TypeSafe Jev or Decision api…
wh000wh000/awesome-jev-liveDeepSeek Harness plugin where Jev Noul probabilities rank skill summaries and glob path results before DSH shows them, with opt-in supervision, completion-check, tool-output filtering and single-operation approval judgments; every feature is off by default and the main DSH model still plans and calls tools.
cobanov/awesome-jevNative DeepSeek Harness (DSH) plugin integrating TypeSafe Jev as a System One decision layer for agent selection, supervision, corrections, and approvals.
hellogumbo/awesome-jevEleven DeepSeek Harness judgments - skill and file selection, supervision, tool-output filtering, approval help - each configurable and all off by default. The main model still plans and answers.
MrJev/awesome-jev
Frequently asked questions
Is anything on by default?
No, all 12 features are disabled by default.
Which decision providers work?
Jev and Luna Decisions, via OpenRouter or OpenAI.
Does Jev replace the main model?
No, the main model still plans and answers.
Related projects
Verified computer-use agent: a Jev guard checks claimed completions and gates irreversible actions; Jev also takes simple steps.
ProjectbetaPythonClaude Code plugin adding typed Jev decisions to hooks: task kind, command risk, failure class, completion; never grants permissions.
PluginbetaJavaScriptopencode plugin that routes the agent's small multiple-answer decisions to Jev and labels each answer auto or escalate by confidence.
PluginbetaTypeScriptSix-file DSPy coding agent where Jev answers closed questions (safety, tool fit, history, delegation) and Python owns policy.
ProjectprototypePythonCoding agent where Jev handles context selection, tool routing, plan gating and verification, and the big model sees only key context.
LibraryprototypePythonSkill and CLI that compares a shortlist of agents, MCP servers or skills against a task and permission ceiling, abstaining on weak leads.
MCP serverprototypeTypeScript