jev-carbone-factor-fit
French library ranking ADEME emission factors for an activity label, splitting code rules from Jev's semantic match.
Overview
Jev Carbone Factor Fit, documented in French, turns the matching of an activity or expense label to a sourced carbon emission factor from France's ADEME database into an explicit, reviewable category. It separates rules that can be checked in code from the semantic comparison delegated to Jev, and returns outcomes such as an exact match, a plausible match requiring human review, or no candidate.
Three demos use only synthetic data and probabilities, make no network calls and do not measure real Jev quality. The main scenario ends in correspondance_exacte with an assertion that fails if the contract changes; a deterministic edge case applies a business rule before any semantic call and returns aucun_candidat with zero Jev calls; and a review scenario shows synthetic confidence of 0.79 requiring review while 0.80 does not, and an incomplete record at 0.62 producing review true. It exposes a matchEmissionFactor function for library use, requires Node 22+, is at version 0.2.0 with tests, and is MIT licensed.
What the decision model decides
| Type | Question | Answers |
|---|---|---|
| Choice | Which candidate ADEME emission factor best matches this activity or expense label?(inferred) | correspondance_exacte · correspondance_plausible · aucun_candidat |
How it works
- Apply deterministic business rules to the label and candidates
- If candidates remain, ask Jev for the semantic match
- Map confidence to a category; below threshold flag human review
- Return a reviewable result
Strengths
- Clear split between verifiable rules and semantic judgment
- Explicit human-review threshold
- Offline demos with contract assertions
Caveats
- Demos are synthetic and do not measure real quality
- French-language documentation
Getting started
`git clone https://github.com/gbesse/jev-carbone-factor-fit.git && cd jev-carbone-factor-fit && npm install && npm run demo`.At a glance
- Type
- Project
- Status
- prototype
- Language
- JavaScript
- Runtime
- Node library (22+)
- Has code
- yes
- Has tests
- yes
- License
- MIT
Use cases and integrations
How curated lists describe it
Classe les facteurs d.
wh000wh000/awesome-jev-live
Frequently asked questions
What does Jev decide?
Which candidate emission factor semantically matches an activity or expense label.
When is human review required?
When confidence is below the threshold, e.g. 0.79 versus 0.80 in the demo.
Do the demos call Jev?
No, they use synthetic data and make no network calls.
Related projects
Mines timestamped driving scenes from vehicle logs by asking Jev lateral and longitudinal manoeuvre questions per moment.
ProjectbetaPythonResearch pipeline that scores NYC 311 complaint descriptions with Jev and maps severity and safety signals on a city grid.
ServiceresearchPythonResume screening where an LLM writes the questions once per job and Jev answers them for every resume, quoting the evidence line.
ProjectproductionTypeScriptClassifies tax PDF pages into 261 IRS forms and 7 page kinds with one Jev request per page and a confidence gate.
ProjectproductionTypeScriptRust tool that asks Jev which PDF pages genuinely need OCR, extracting text pages locally and sending only the rest to OCR.
ProjectbetaRustFills Japanese government PDF forms by asking Jev which profile item each unlabeled field is; personal values stay in the browser.
ServicebetaJavaScript