User repository on GitHub
decision-first data cleaning system powered by Jev
- Forks
- 1
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Clean
Decision-first CSV cleaning: local code builds bounded repair candidates, Jev returns action, applicability and semantic-risk distributions, and a fixed policy decides what may be applied — or abstains. The model never writes repairs or mutates the source CSV.
- Stars
- 2
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- kunyanli230
- Last verified
- Oct 2026
Used inJev for Data & Analytics
Highlights
- Repairs are proposed by local code; Jev supplies action, applicability and semantic-risk distributions, and the fixed policy decides apply or abstain.
- The full distribution is kept, so a 0.20 high-risk tail can force abstention even when expected risk matches a passing option.
- A recorded 2026-09-22 live demo produced 10 candidates, 6 committed decisions and 4 abstentions with a partial outcome.
- v0.2.0 is labelled experimental and supports Python 3.11–3.13; the package and CLI keep the name idac.
- Without an API key, baseline mode runs the same builders and validator with no Jev calls and no probability decision records.
Quickstart
bash
git clone https://github.com/Kunyanli230/jev-clean.git
cd jev-clean
uv sync --locked
uv run idac --helpWatch out
Grounded in the truncated repo README and GitHub metadata read on 2026-10-15. Its risk examples are explicitly synthetic policy illustrations, not live Jev measurements, and the README stresses the demo shows inspectability, not calibrated accuracy or superiority; live decisions can vary between runs.
Questions about this entry
Short answers drawn from the linked project or page.
It cleans CSVs with a decision-first loop: local code builds bounded repair candidates, Jev answers targeted questions about action, applicability and semantic risk, and a fixed policy decides which repairs may be committed or whether to abstain. It calls itself an independent experimental project, currently v0.2.0.
No. The model neither writes executable repairs nor mutates the source CSV; local code proposes candidates and Jev only returns explicit action, applicability and semantic-risk distributions. cleaned.csv is an export of the committed snapshot, the input file is never overwritten, and each run needs a fresh output path.
Because an expected score can hide the tail. The README walks through two synthetic distributions with the same expected risk of 0.40: one passes, while the other has P(high risk) = 0.20 above the 0.10 limit and the policy abstains. Jev Clean saves the distribution, the recomputed expectation, each threshold comparison and the final status.
Install uv and use Python 3.11, 3.12 or 3.13: git clone the repo, run uv sync --locked, then uv run idac --help. The public name is Jev Clean but the Python package and CLI remain idac; offline inspection starts with uv run idac profile against the example CSV and config.
Yes, in baseline mode: uv run idac baseline uses the local candidate builders and validator but makes no Jev calls and has no probability decision records. Baseline exports include a hashed snapshot that preserves row identities and nulls for evaluation.
A recorded run on 2026-09-22 produced 10 candidates, 6 committed decisions and 4 abstentions with an overall partial outcome, and the rule baseline repaired more of that demo dataset. The README says these observations show inspectability, not calibrated accuracy or superiority, and live model decisions can vary between runs.
A TypeSafe API key, which the README suggests reading with read -rsp so it never enters shell history, then exporting before uv run idac clean with input, config and output arguments. Results land in runs/first-run/report.md and decision_cards.md; live API calls consume provider quota and may incur charges.
More Jev resources like this
A dependency-free Python pipeline that rebuilds a site's internal link map with Jev: code picks candidate targets and in-copy anchor phrases, Jev judges them, and only links that clear the threshold are placed.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
stas4000/jev-linkmapHTML
stas4000/jev-linkmapHTMLWatches X for the updates that matter to you, has Jev judge each post against a YAML profile, groups posts about the same story, and alerts you in Slack early enough to act.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ethan-ab/xscout-jevTypeScript
ethan-ab/xscout-jevTypeScriptA Windows desktop assistant that OCRs a selected chat area, lets you correct the transcript, then has Jev analyze the conversation and rank three candidate replies from an optional reply model; nothing is auto-sent.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
qcjlchina/jev-chat-assistantPython
qcjlchina/jev-chat-assistantPythonA C/C++ compiler driver that combines Clang's optimiser with Jev for function-call inlining at every discretionary call site. The project reports 58.2% code-size reduction on Statemate from the Embench benchmark; Linux with pinned LLVM 21.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ramneet-singh/jevoptPython
ramneet-singh/jevoptPythonAn auto-clipper that turns long videos into short standalone clips: code enumerates every plausible cut point, Jev judges the clip each candidate edge would make, and you get ranked mp4s for Shorts, Reels and TikTok.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
vbs2004/jevcutPython
vbs2004/jevcutPythonAn Android automation agent that decides from the accessibility tree instead of screenshots: one TypeSafe Jev request per step picks the operation and target, ADB executes, and an optional vision fallback takes over when the fast path gets stuck.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
xinwang-nwpu/jev-mobilePython
xinwang-nwpu/jev-mobilePythonFrom the Jev guides
Original write-ups on what this entry is for.
From the Jev AI model community
Posts from builders shipping with Jev right now.