helpme-Jev
Self-hosted helper for tricky chats and everyday decisions: Jev makes typed checks, an LLM handles language, the user decides.
Overview
HelpMe JEV is a small self-hosted assistant for questions that are small but strangely hard: is this message actually negative, should I reply now or ask another question, what is the safest next step at work, can I say this more clearly. You paste a message, screenshot or situation, and it turns it into a few concrete checks and possible next moves, with drafts you can edit before anything is sent.
Its rule is that Jev decides, the LLM explains and you decide. Jev answers typed questions (yes/no probabilities, option probabilities and scores) and never writes the final message; an LLM translates input for Jev and clarifies, drafts or polishes text on request, using OpenAI Chat Completions, OpenAI Responses or Anthropic Messages protocols. A chat helper analyses intent, risk, emotion, needs and next actions in one-on-one or group chats, and a decision workbench handles general choices. It is version 0.1.0 and AGPL-3.0, with documentation in English and Chinese.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Yes / no | Is this message actually negative? | yes · no |
| Choice | What is the best next move: reply now, ask another question, or leave it? | — |
| Score | How risky is this situation?(inferred) | — |
How it works
- Paste a message, screenshot or situation
- LLM converts input into state for Jev
- Jev answers typed checks on intent, risk, emotion and next actions
- LLM drafts text on request; user chooses and sends
Strengths
- Clear split between judgment and wording
- User stays in control of what is sent
- Supports several LLM protocols
Caveats
- Early 0.1.0 release
- AGPL-3.0 licence
- Requires both Jev and LLM endpoints
Getting started
Self-host per the Quick start in the README and configure JEV and LLM endpoints via environment variables.At a glance
- Type
- Project
- Status
- prototype
- Language
- TypeScript
- Runtime
- self-hosted web app
- Has code
- yes
- License
- AGPL-3.0
Use cases and integrations
How curated lists describe it
HelpMe JEV —— JEV 决策工作台：JEV 出决策、LLM 做表达、用户拍板（聊天副驾 + 通用决策）.
hellogumbo/awesome-jev
Frequently asked questions
Does Jev write replies?
No, Jev only answers typed questions; an LLM drafts text when asked.
Who sends the message?
You do; the app stops before anything is sent.
Which LLM APIs are supported?
OpenAI Chat Completions, OpenAI Responses and Anthropic Messages.
Related projects
- 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 Android chat co-pilot that reads the active chat via Accessibility and returns typed verdicts on intent, emotion and risk; never replies.
AppbetaKotlinAndroid chat copilot: Jev judges intent, needs and risk in visible chats, then ranks drafted replies to fill, never send.
HookDeepSeek Harness plugin that classifies each user message's emotion, urgency and department with Jev and injects it as context.
PluginbetaTypeScriptZero-dependency Android app that judges the visible WeChat message's intent, emotion, urgency and reply posture with Jev.
AppbetaJavaAndroid WeChat assistant where Jev judges the chat and scores the relationship, and a chat model drafts three replies.
AppbetaKotlin