Skip navigation
introducing Jevbox: an open-source document drive that organizes itself, answers questions, and…
0:00 / 0:40

introducing Jevbox: an open-source document drive that organizes itself, answers questions, and…

Kushal Byatnal

@kushalbyatnal8K followers

Visit on X
2,404 likesDownloadOriginal post

180,401 views

introducing Jevbox: an open-source document drive that organizes itself, answers questions, and cites everything (all powered by Jev). TLDR; - upload anything and Jev categorizes and files it into the folder tree - search is hierarchical: Jev picks the folder, then the document, then the section. No embeddings, no vector DB - every answer cites the page and section it came from - permissions are enforced at retrieval. If someone loses access to a source, they lose access to answers built on it too - ships with an MCP server, so your agents get the same permission-aware context Andrew Luo has been experimenting with Jev internally and wanted to see how far it could go on a real document library. It's built on Extend Parse + Jev, and fully open-source The GitHub has instructions to self-host, or you can deploy it to Render in one click

31 Comments

Comments from the original post on X (Twitter)

  • Kushal Byatnal's profile picture

    @kushalbyatnal

    open source repo here:

    21 likes

  • Michael's profile picture

    @michael_chomsky

    where can i try this?

    3 likes

  • Yechan Do's profile picture

    @doyc_1

    Cool. Jev is solving retrieval problems

    3 likes

  • Kushal Byatnal's profile picture

    @kushalbyatnal

    if you squint, everything is a classification problem

    7 likes

  • Saad's profile picture

    @eSaadster

    gh repo?

    2 likes

  • Kushal Byatnal's profile picture

    @kushalbyatnal

    here you go:

    1 like

  • brexton's profile picture

    @brexton

    Insane

    2 likes

  • Kushal Byatnal's profile picture

    @kushalbyatnal

    @andrewlu0 is insane

    2 likes

  • Darren Chow's profile picture

    @dchow_

    @ameyaajoshi the use cases with extend + jev are crazy

    2 likes

  • Josh Kaplan's profile picture

    @JoshKaplanc

    cool

    1 like

  • Fred Dogan's profile picture

    @firatcand

    this is cool, organizing folders is a pet peeve of mine. I really don't enjoy it. Notion could use a similar approach honestly. Cool use case for Jev, I have been building a tool router for machine payments with Jev it's quite interesting. Will launch soon too.

    1 like

  • Sael's profile picture

    @0xsael

    no vector db is the real flex here, cant wait to stress test the hierarchy

    1 like

  • Aniruddha Ganesh's profile picture

    @IndicGames

    @theo I recall you doing something about file organisation and files not needing to be at one exact path. This seems like the most natural evolution of that ideao.

    1 like

  • Sahibzada Allahyar's profile picture

    @singularity_bly

    I’d use GLiDE for the folder → document → section choices. It beats Jev on Decision Index retrieval, 60.9 vs 55.4, and can reason further when two paths look plausible. That’s worth testing on the same document library.

  • Jack's profile picture

    @salesoptecs

    The filing is the easy half. What actually breaks document search in a real company is that the drive holds three versions of the same policy and nothing marks which one is current. A citation only helps if somebody decided that page was still true.

  • flare on's profile picture

    @highrthanthesun

    Looks interesting

  • Mildo 🥽's profile picture

    @Mildo_ai

    Is this a different kind of vector database? What’s the benefit over graphrag?

  • kartik bhardwaj's profile picture

    @kartikb753

    Retrieval-time permission checks are the part most RAG stacks get wrong, cached chunks serve after access is revoked. The failure I would eval first is miscategorization, a misfiled doc goes invisible to hierarchical search. AI engineer working on agent harnesses and evals.

  • Farzan Ansari's profile picture

    @imfarzanansari

    people here are doing crazy things!

  • Johannes Laslo's profile picture

    @JohannesLaslo

    what problem does this solve? Is this advantageous to a local wiki style knowledgebase with vector database?