Measuring what Jev does in a RAG pipeline
Chinese X post measuring Jev in a RAG pipeline over TypeSafe's docs: correct top-ranked answers rose from 7 to 17 of 20.
This summary is based on how curated lists describe the project; its README was not available to review.
Overview
This X post, written in Chinese, sets out to measure rather than repeat claims about Jev in retrieval-augmented generation. The author built a RAG pipeline over 42 pages of TypeSafe's official documentation and, per the listing, reports that for 20 answerable questions the correct answer ranked first went from 7 to 17 with Jev in the pipeline, and that none of the unanswerable questions were handed to the generating model, meaning Jev's answerability check filtered them out.
No post text or code was available here, so the method and pipeline details are not confirmed beyond the summary. It is a useful small, concrete measurement of Jev as a reranker and answerability gate.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How relevant is this retrieved passage to the question? (reranking)(inferred) | — |
| Yes / no | Can this question be answered from the retrieved documentation?(inferred) | — |
How it works
- Index 42 pages of TypeSafe documentation
- Retrieve candidates for each question
- Jev reranks and checks answerability
- Only answerable questions go to the generator
- Compare first-ranked accuracy with and without Jev
Strengths
- Concrete before/after measurement
- Shows an answerability gate preventing hallucination
Caveats
- Small sample (20 answerable questions)
- Post only; code not available here
- Single document set
At a glance
- Type
- Discussion
- Status
- discussion
Use cases and integrations
How curated lists describe it
X (Chinese): instead of repeating the unsupported claims about Jev in RAG, builds a pipeline over 42 pages of the official docs and reports that the answer ranked first among 20 answerable questions went from 7 to 17, with none of the unanswerable ones handed to the model.
yibie/awesome-jev
Frequently asked questions
What improved?
Correct answers ranked first rose from 7 to 17 of 20, per the listing.
What happened to unanswerable questions?
None were passed to the generating model.
Is the code available?
Not in this listing.
Related projects
MCP server giving coding agents typed judgment tools (verify, screen, rerank, gate, score) backed by Jev, with auto/review/escalate policy.
MCP serverbetaPythonTyped-decision accuracy layer for RAG: six checks for intent, relevance, clarification, sufficiency, grounding and caveats.
ServicebetaPythonSmall stdio MCP server with ask, rank and verify tools backed by Jev, a local event log and installers for several agent harnesses.
MCP serverbetaTypeScriptPermission-aware document library using Jev or Cloudflare Clef for hierarchical search, passage scoring and automatic filing.
LibrarybetaTypeScriptPython library that replaces hard-coded RAG thresholds with calibrated decisions for stopping, chunking, selection, abstention and caching.
ServicebetaPythonGraph-database-agnostic GraphRAG framework where swappable Laya or Jev decisions score edges, route intents and gate hallucinations.
ServicebetaPython