User repository on GitHub
Decision-vector profile retrieval with Clef and Jev, HNSW search, and BM25/BGE benchmark comparisons
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jevector
A Python toolkit that encodes profile criteria as 64- or 256-value decision vectors with Clef or Jev, retrieves candidates with HNSW search, then checks mandatory conditions in both directions and ranks the rest by preference coverage. A 52-profile benchmark compares the vectors against BM25 and BGE-small.
- Stars
- 5
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- lalalune
- Last verified
- Oct 2026
Used inJev for Software & DevTools
Highlights
- Encodes criteria with Clef or Jev into 64 or 256 numeric values; HNSW retrieves 32 candidates per query.
- The matcher checks mandatory conditions in both directions before ranking profiles by preference coverage.
- Benchmark on 52 synthetic profiles: Jev 64 and 256 reach 87.5% top-1; Clef 256 100%; BM25 12.5%; BGE-small 0%.
- Included vectors search without an API key; encoding new profiles needs a Cloudflare account or a Jev key.
- The README states the benchmark is small, pre-rule-filtering and built around the tested criteria.
Quickstart
bash
git clone https://github.com/lalalune/jevector.git
cd jevector
python3 -m venv .venv-search312
.venv-search312/bin/python -m pip install -r requirements.txt
scripts/python -m matching.cli rank --query-id family0-persona --gallery benchmarks/vectors/jev-256.json --method hnsw --candidates 32 --k 5Watch out
README reviewed 2026-10-17; the benchmark covers 52 synthetic profiles and eight queries from four profile-pair families excluded from development, and the README itself warns the results do not establish better performance on real profiles or general search.
Questions about this entry
Short answers drawn from the linked project or page.
Jevector is a Python toolkit for profile matching: it encodes each profile's criteria as numeric decision vectors with Clef or Jev, retrieves candidate profiles with HNSW search, checks mandatory conditions in both directions and ranks the remaining profiles by preference coverage.
Jev, or Cloudflare's Clef, encodes profile criteria as 64 or 256 numeric values, and the matcher compares these vectors: HNSW retrieves candidates, mandatory conditions are checked both ways, and the rest are ranked by coverage. The digest does not show the exact Jev question shapes used.
On 52 synthetic profiles with eight queries: Clef 256 reached 100% top-1, Jev 64 and Jev 256 87.5%, BM25 12.5%, and BGE-small and BM25 + BGE-small did not beat BM25. The README warns the measurement is before rule filtering and the test is small.
Yes, the included vectors search with no key: clone the repo, install requirements and run the matching.cli rank command against benchmarks/vectors/jev-256.json. Encoding a new profile needs either a Cloudflare account or a Jev API key entered at a hidden prompt.
HNSW retrieves candidate profiles (32 candidates by default) for the decision vectors, with exact search available via --method exact. The README notes HNSW can miss candidates, which is why the exact option exists.
The README says decision vectors performed better because the profiles were built around the selected criteria and the test is small, and that the results do not establish better performance on real profiles or general search.
The GitHub user lalalune publishes it under the MIT license; the repository was created 2026-10-08 and pushed 2026-10-09. Re-running the benchmark needs both provider accounts and can incur API charges, per the README.
More Jev resources like this
A proof-of-concept Python scanner that maps a backend codebase's attack surface and ranks likely vulnerabilities down to the suspicious line, using Jev as the classifier for about one cent per repo. It reads Python, JavaScript and TypeScript, and a demo site shows read-only results on PyGoat and NodeGoat.
Jev Tools & CLIs#community#github#security
A 2.4 GHz WiFi site-survey instrument built on a ZimaBoard 2 and an Alfa USB adapter: Python and Scapy turn captured 802.11 frames into structured RF telemetry, and Jev reads each measurement window through OpenRouter to return coverage, congestion and optimization decisions.
Jev Tools & CLIs#community#github#wifi
Community
ronibandini/jevwifianalyzer-Python
ronibandini/jevwifianalyzer-PythonA repository described as Jev use cases; the captured README documents an Austrian hotel ranking pipeline that merges Google Maps and Booking.com data, matches the ÖHV member list from meinhotel.at and computes a 0–100 guest score, with dashboards and Excel reports.
Jev Tools & CLIs#community#github#python
Community
schilfschneider/jevPython
schilfschneider/jevPythonA local deep-research tool that turns your own PDF, DOCX, PPTX, image, TXT, or Markdown files into a structured report, using TypeSafe's Jev to rerank Qdrant chunks and verify evidence coverage while an LLM plans and writes.
Jev Tools & CLIs#community#github#python
Community
oqura-ai/deepdocPython
oqura-ai/deepdocPythonAn independent open-source pipeline that gives Jev vision: it detects and tracks objects in images, video, and RGB-D camera feeds, then asks Jev a typed question about each relevant object and prints the chosen answer with its confidence.
Jev Tools & CLIs#community#github#python
Community
CharlesFeng0314/JEV_seesPython
CharlesFeng0314/JEV_seesPythonA local workbench that pairs OpenCV vision with Laya or Jev decision models: images are segmented and measured, hard rules settle the numbers, and the model gives the verdict — with one-click installers and an "open-source JEV" training path served at /v1/systemone.
Jev Tools & CLIs#community#github#python
Community
uu889/laya-opencvPython
uu889/laya-opencvPythonFrom 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.