codex-jev
Retrieval layer for Codex that finds evidence locally, optionally ranks sanitized excerpts with Jev, and returns bounded, cited evidence.
Overview
Codex Jev adds a retrieval layer to Codex for large repositories and noisy logs. It finds candidate evidence locally, optionally ranks sanitized excerpts with Jev, and returns bounded, source-addressed evidence with previews, source hashes, line ranges and exact follow-up reads. It also lists retained, omitted and unscored candidates so omissions can be recovered.
It does not replace your coding model, proxy inference, rewrite conversation history or bypass Codex usage limits, and it works without a Jev key in local-only mode, with paid activation explicit. It reports Jev, cache, bypass and fallback statistics separately and leaves missing usage unknown. The README labels it a public beta: offline safety and packaging tests pass, but generalized token, cost and speed savings are not yet established. The adapter is MIT and the hosted Jev service is billed separately.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Score | How relevant is this sanitized excerpt to the investigation query?(inferred) | — |
How it works
- Search code, tests, config and logs locally
- Optionally send sanitized excerpts to Jev for ranking
- Return bounded evidence with hashes and line ranges
- Codex reads more exactly as needed
Strengths
- Local-only mode without a key
- Recoverable omissions
- Honest, separated statistics
Caveats
- Public beta; savings not established
- Jev billed separately
Getting started
git clone https://github.com/Hyper-AI-Lab/codex-jev.git and follow docs/INSTALLATION.md.At a glance
- Type
- Project
- Status
- beta
- Runtime
- Codex adapter / MCP
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Evidence selection and quota-aware recovery for Codex. Optional Jev reranking, local privacy controls, exact source reads. Keep your coding model.
hellogumbo/awesome-jev
Frequently asked questions
Does it need a Jev key?
No, local-only mode works without one.
Does it change my model?
No, it only supplies evidence.
Are savings proven?
Not yet; the README says they are not established.
Related projects
Codex search plugin that runs ripgrep and keeps only passages Jev judges relevant to the question.
PluginprototypeTypeScriptPipe grep or file lists into Jev to get every candidate scored, sorted and cut at a calibrated threshold, so agents read only what matters.
ProjectbetaTypeScriptMCP server that returns only the JS/TS code a task needs by following the TypeScript compiler graph and asking Jev about relevance.
MCP serverbetaTypeScriptnpm CLI that screens large agent tool outputs with Jev, returning compact typed decisions and IDs while originals stay recoverable.
CLIbetaTypeScriptCLI and MCP server giving coding agents Jev-powered semantic search, focused reads and list picking in about a second for about a cent.
BenchmarkbetaPythonDeterministic code-graph context for coding agents on Postgres, AGE and pgvector, with optional Jev or local re-selection of symbols.
MCP serverbetaPython