User repository on GitHub
LLM-authored decision SOP generator for Jev System One model
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
CommunityJev Tools & CLIs
Jevtree
A Python CLI that takes a CSV, JSON file or directory plus a plain-language goal, has an LLM normalise it into a FeatureTable, grows an information-gain decision tree, and exports a runnable DecisionSOP with a trace.
- Stars
- 4
- Language
- Python
- License
- MIT
- Published by
- Community
- Author
- qzqdz
- Last verified
- Oct 2026
Works forJev for Classification
Used inJev for Software & DevTools
- Languages
- Python
Highlights
- The pipeline is LLM normalize, then information gain grow, producing feature_table.json, tree.json, sop.json, story.md and optional traces.json.
- `jevtree run --sop ... --trace` replays an existing SOP against one observation and records the path the decision took.
- Reported on MiniBooNE, Acc@budget beats random and sequential at a budget of 5, while at 10 the two information-gain variants effectively tie.
- The backend is configured through JEVTREE_LLM_* variables against any OpenAI-style base URL, model and key.
Watch out
The README documents the language model only as an OpenAI-style endpoint configured with JEVTREE_LLM_*; it never states that TypeSafe's Jev is the default or required, so check what you are pointing it at before assuming the Jev path is wired up. The published numbers cover MiniBooNE and Cube only, and on Cube ig_static and sequential both score 1.000 at Acc@3, so that row separates nothing from the baseline. The caption under the pipeline diagram notes that training-free means no gradient training, not no data statistics or tree building. Needs Python 3.11 or newer. Copy reviewed against the repository README on 2026-10-04, not from running it.
Questions about this entry
Short answers drawn from the linked project or page.
The README does not say so. It configures the language model through JEVTREE_LLM_* variables and describes them as compatible with an OpenAI-style interface taking a base URL, model and key, which is how you would point it at Jev or at anything else that speaks that shape. The Jev name appears in the project name and the variable prefix rather than in a stated dependency, so verify the endpoint yourself.
Treat them as directional. The Acc@budget table shows the discriminative variant at 0.824 against 0.746 random and 0.724 sequential at a budget of five, which is a real margin, but at a budget of ten it scores 0.820 against IG_static's 0.822, so the two variants tie. On Cube, ig_static and sequential both hit 1.000 at Acc@3 while random gets 0.766, which tells you the task is easy rather than that the method won.
It is the right move when you need to read the rule back, which is what the DecisionSOP export is for, since the tree and SOP are JSON you can inspect and run. It is the wrong move if you want accuracy on data where information gain has nothing to bite on. The README's own caption is the fair summary: training-free means no gradient training, not that no data statistics or tree building are involved.
The LLM-normalised feature table as JSON and CSV, tree.json for the information-gain tree, sop.json as the runnable DecisionSOP, a bilingual story.md, and traces.json if you pass --trace-examples. Output goes wherever --out points, defaulting to results/decide_<timestamp>/, and decide, run, validate, eval-afa and version are the documented subcommands, with run-job and from-data as aliases.
Create a Python 3.11+ virtualenv, pip install -e ., copy .env.example to .env and fill JEVTREE_LLM_API_KEY, then run scripts/quickstart.sh or jevtree decide with --data, --goal and --out.
Yes. jevtree run --sop results/.../sop.json --input an observation JSON replays one observation, and adding --trace records the path taken; validate checks a saved SOP.
It runs hard-budget feature acquisition, the evaluation behind the MiniBooNE Acc@budget and Cube Acc@3 tables, configured through a file in configs/.
More Jev resources like this
34GitHub stars
Pure-HTTP automation for the TypeSafe register, magic-link login, onboarding, and API-key creation flow, with pluggable temp-mail backends, concurrent batch runs, and key verification exports.
Jev Tools & CLIs#community#awesomejev#catalog-import
Community
2951461586/Jev-Register-ToolPython
2951461586/Jev-Register-ToolPython36GitHub stars
Python toolkit plus 0.8B multimodal model that takes a message, document, or image and returns calibrated choice, yes/no, and score answers in one forward pass for local decision workloads.
Jev Tools & CLIs#community#awesomejev#catalog-import
Community
PsiACE/dohnutsPython
PsiACE/dohnutsPython39GitHub stars
Video tool where Jev scores subtitle segments and code assembles a short highlights cut plus a de-watered full cut with timestamped summaries.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
cclank/jevclipPython
cclank/jevclipPython30GitHub stars
Forkable weekday radar that asks Jev one yes-or-no question per interest over each new arXiv announcement and publishes a page plus RSS feed.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
Eliot5566/JEV-Paper-RadarPython
Eliot5566/JEV-Paper-RadarPython38GitHub stars
Async supervisory layer for Hermes Agent that locks Definitions of Done, proves mechanical criteria locally, and calls Jev only for unresolved semantic checks under a token budget.
Jev Tools & CLIs#community#jevbest#catalog-import
Community
keeltrace/hermes-nervePython
keeltrace/hermes-nervePython15GitHub stars
Two-tier decision engine pairing a 22.6M-parameter MiniLM encoder on the Apple Neural Engine with the on-device Apple Foundation Model as fallback for uncertain inputs.
Jev Tools & CLIs#community#awesomejev#catalog-import
Community
kerryrm/systemANEPython
kerryrm/systemANEPythonFrom the Jev guides
Original write-ups on what this entry is for.
The request shape barely changes across surfaces; the surrounding contract does. When to call the API directly and when a gateway, a Worker binding, or a router entry earns its extra hop.
Jev Tools & CLIs#gateway#vercel#cloudflare
Read article
Taking a decision checkpoint out of Python: what the GGUF and browser WebGPU exports preserve, what they change, and the parity evidence the model cards publish.
Jev Tools & CLIs#open-models#local#onnx
Read article
Jev labels a record that already exists. Code moves, flags, or holds it. How the document, mail, and transfer tools keep the action out of the model.
Jev Tools & CLIs#classification#email#security
Read article
From the Jev community
Posts from builders shipping with Jev right now.