Jev Graph Search
github.com/emlembow/jev-graph-search- Category
- Developer Tools
- Rank
- No. 1652Tools index
Previous survey · No. 1568 ·
- Pricing
- Open Source
- Type
- TOOL
- Use case
- Data, Retrieval & Knowledge
- Interfaces
- CLI · Agent Skill / Plugin
- GitHub
- 13 stars
- Latest release
- v0.2.2
- Date
About
Jev Graph Search helps an AI agent locate relevant notes and passages in a local Obsidian vault or a folder of Logseq Markdown files. It first builds a local shortlist of candidates, then uses the Jev ranking service to reorder them against the query, returning the original passage and source reference; on one 8,851-node test graph it raised top-five source recall from 63.9% to 81.8%.
At a glance
Jev Graph Search finds passages in a local Obsidian or Logseq folder and returns them with their source references. Try it if an agent needs to cite your notes and keyword matches regularly miss the passage you want.
Good fit
People connecting an agent to a Markdown knowledge base and comparing retrieval quality on questions with known answers.
Consider something else when
Skip hosted ranking if your notes cannot leave your machine. The documented offline mode uses local keyword ranking.
Start here
Use the included sample graph or a synthetic folder. Ask a question with a known source and check whether that passage appears in the first few results. Compare offline and hosted rankings on the same query before using real notes.
Cost and requirements
MIT-licensed CLI with Node.js 20 or newer. Hosted ranking needs a TypeSafe or OpenRouter API key and can incur provider charges. The offline mode works without a key.
Limits to know
The shortlist is made locally. Hosted reranking sends candidate material to the configured provider. The published recall result comes from one tax-code graph, so measure whether reranking helps on your own questions.
Related work
- JevMan (context)
Sources checked 2026-11-05. Source-based assessment.
Why it made the leaderboard
Gives agents a way to find and cite the right passage in a personal knowledge vault by reranking a local shortlist, improving retrieval accuracy over plain search without sending notes to a third party.
Tags
Tech Stack
Comments (0)
No comments yet
Editorially curated, with community endorsements as a secondary signal. Corrections welcome.