jev-system-one
Terminal Q&A app where Jev sets response policy and reviews drafts while OpenAI only writes, orchestrated with LangGraph.
Overview
Jev System One is a terminal interface in which OpenAI generates answers and TypeSafe Jev acts as the authority on response policy and quality review. For each plain-English question, Jev first decides the response mode (Choice), whether clarification is required and whether an uncertainty notice is required (Nouls), and the response depth (Score). OpenAI drafts from that policy, resolving ambiguity itself if needed.
Jev then reviews the draft: does it answer the question, does it contain unsupported claims, is a revision required (Nouls), and an answer quality Score. If revision is needed, LangGraph loops OpenAI through revisions until approval or a maximum number of iterations, then OpenAI finalises. The TUI shows the answer beside a decision report of every typed judgment, so the user sees why the answer has its shape. The human is involved only once. It has CI and an MIT licence.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which response mode should the answer use? | — |
| Yes / no | Is clarification required before answering? | — |
| Yes / no | Does the draft contain unsupported claims? | — |
| Score | How good is this draft answer? | — |
How it works
- Jev decides response mode, clarification, uncertainty notice and depth
- OpenAI drafts according to that policy
- Jev reviews the draft; revision loop until approved or max iterations
- OpenAI finalises; TUI shows answer and decision trail
Strengths
- Transparent decision trail beside each answer
- Clear split: Jev decides, OpenAI writes
- Automated review loop with no extra user input
Caveats
- Requires both OpenAI and TypeSafe keys
- Demo app; review quality not benchmarked
At a glance
- Type
- App
- Status
- prototype
- Language
- Python
- Runtime
- Terminal UI (LangGraph)
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
A terminal Q&A app where OpenAI writes answers and Jev sets response policy and reviews drafts.
logicrw/awesome-jev-projectsA polished OpenAI + TypeSafe Jev terminal interface for answers with transparent decision reports.
hellogumbo/awesome-jev
Frequently asked questions
What does Jev decide?
Response mode, clarification, uncertainty notices and depth before drafting, then whether the draft answers the question, has unsupported claims and needs revision.
What does OpenAI do?
Only drafting, revising and finalising wording from Jev's decisions.
Can I see the reasoning?
The TUI shows a decision report of every typed judgment beside the answer.
Related projects
Framework-independent participation gate letting group-chat agents answer untagged questions only when Jev finds a useful reason to join.
ProjectbetaTypeScriptWeChat reply assistant fork: DeepSeek drafts three replies, Jev pre-judges, reviews and ranks them; auto-send is cancellable.
ProjectbetaPythonPilot that puts Jev into LangGraph routing and Deep Agents guardrails and claim verification, with executed notebooks and caveats.
ProjectprototypePythonUnofficial Python adapter that brings Jev verification, classification and scoring into the Mellea framework as Requirements.
ProjectbetaPythonCompanion code comparing Google ADK and Microsoft Agent Framework for structured video timelines, with Jev as a shared per-segment QC gate.
LibraryprototypePythonUiPath bring-your-own-guardrail connector that runs AI Trust Layer llm_as_judge checks as Jev noul questions.
ProjectprototypeJavaScript