dsh-jev-advisor
DeepSeek Harness plugin that gives the human a Jev second opinion beside the agent's multiple-choice questions.
Overview
dsh-jev-advisor takes the opposite approach from most Jev plugins for DeepSeek Harness: it gives Jev to the human, not the agent. When the coding agent stops and asks you a multiple-choice question, the plugin turns it into a structured TypeSafe request and floats a recommendation beside the options, showing the full probability spread, Jev's confidence, the exact JSON sent, and a one-click button to adopt it.
It changes nothing about the model: no prompt edits and no tool the model has to remember to call. It uses an overlay and a settings section and never replaces the question card. Setup is a single API key in settings. The README is explicit that it is not an agent-loop gate and points to other plugins for tool restriction or routing. Tests cover request building, client wiring and a round trip against a mock endpoint.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which of the options the agent offered is the best choice? | — |
How it works
- Agent asks the user a multiple-choice question
- Plugin builds a Jev Choice request from the question and options
- Overlay shows recommendation, distribution and confidence
- User adopts the recommendation or chooses otherwise
Strengths
- Useful from the first question with one setting
- No model or prompt changes
- Shows raw request JSON for transparency
Caveats
- Advisory only; does not gate agent actions
- Requires a TypeSafe key
Getting started
dsh plugin --profile <name> add github:gordan-code/dsh-jev-advisor, then paste an API key in Settings → Jev.At a glance
- Type
- Project
- Status
- beta
- Language
- TypeScript
- Runtime
- DeepSeek Harness plugin (UI overlay)
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Jev (TypeSafe System One) gives the human a second opinion, not the agent.
wh000wh000/awesome-jev-live
Frequently asked questions
Who receives Jev's answer?
The human user, as a recommendation beside the agent's question.
Does it change agent behaviour?
No, it only adds an overlay; the agent and prompts are untouched.
How do I configure it?
Paste a TypeSafe API key in Settings → Jev.
Related projects
DeepSeek Harness plugin registering a jev_decide tool so the agent can ask Noul, Choice or Score questions and gate on probabilities.
PluginbetaTypeScriptSelf-hosted Yao Agents platform for agents, workspaces and task boards across devices, which announces Jev support.
ProjectbetaDeepSeek Harness jev_decide tool that brings the Jev decision paradigm to any OpenAI-compatible LLM, or to real Jev with a key.
ProjectprototypeTypeScript- jev-chat-windows★ 802
Windows WeChat reply assistant: local offline OCR reads the chat, Jev judges intent and tension and ranks three drafted replies to fill in.
ProjectproductionPython AstrBot plugin where Jev decides whether the bot should join a group conversation; the main model still writes the reply.
ProjectproductionPythonOh My Pi coding agent (a Pi fork with IDE wiring) with an optional TypeSafe judgment provider for bounded decisions.
ProjectproductionRust