User repository on GitHub
Evidence-backed knowledge graph construction with typed Jev relation decisions
- Forks
- 5
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
JevGraph
A schema-guided document-to-graph pipeline: local parsing feeds entity mentions and bounded candidate pairs to Jev, which chooses among closed-set relations, and every edge lands as proposed, review, or rejected with page-level evidence.
- Stars
- 42
- Language
- Python
- License
- APACHE-2.0
- Reading time
- 2 min
- Published by
- Community
- Author
- chenmingtang830
- Last verified
- Oct 2026
Used inJev for Data & Analytics
- Languages
- Python
- Platforms
- CLIVercel AI Gateway
- Capabilities
- Open sourceRuns locallyRequires API key
Highlights
- Replaces open-ended prompt-to-triples extraction with bounded stages ending in closed-set Jev Choice questions over blocked candidate pairs.
- On the v0.5.1 FewRel benchmark Jev scored 87.50% planned-case accuracy with 313 ms p50 and 431 ms p95 latency at an illustrative 0.410¢ per 100 decisions.
- Every edge keeps source and canonical hashes, parser identity, page spans, and character-level evidence offsets; results export as JSON, CSV, or Neo4j Cypher.
- Quickstart: uv sync --extra documents --extra dev, then uv run jevgraph build with a YAML ontology and entity gazetteer.
- The authors say the benchmark is one narrow operation, not a general model ranking or an end-to-end knowledge-graph score.
Quickstart
bash
git clone https://github.com/chenmingtang830/jevgraph.git
cd jevgraph
uv sync --extra documents --extra dev
uv run jevgraph build examples/company_events.txt \
--ontology examples/ontology.yml \
--entities examples/entities.yml \
--provider keyword \
--out runs/demo.jsonWatch out
Copy reviewed against the repo README on 2026-10-15. Jev's cost figures are an illustrative list-price equivalent rather than a provider receipt, and the README limits the benchmark to direct relation selection. Output is explicitly not claimed to be true, approved, or safe for consequential reuse.
Questions about this entry
Short answers drawn from the linked project or page.
An open-source, schema-guided pipeline that turns PDF, DOCX, PPTX, and text files into candidate knowledge graphs, where each relation is a typed Jev decision and every edge keeps its evidence. It targets fixed or slowly changing ontologies.
After local parsing and candidate blocking, Jev answers closed-set Choice questions over eligible entity pairs, for example picking one of 16 explicit FewRel relations. Deterministic schema checks then route edges to proposed, review, or rejected.
Python 3.11+ and uv are required: git clone the repo, then uv sync --extra documents --extra dev, then uv run jevgraph build with a YAML ontology and entity gazetteer. DOCX and PPTX rendering also needs LibreOffice.
Yes. The keyword provider is a labeled deterministic baseline that makes no model call, and document parsing is local. Jev relation choices need access through the TypeSafe API or a Vercel AI Gateway key.
On the repo's v0.5.1 FewRel run Jev scored 87.50% planned-case accuracy, below GPT-5.6 Luna at 89.375% and DeepSeek V4.1 Flash at 93.125%, but with lower latency and a far lower illustrative cost of 0.410¢ per 100 decisions.
Every edge keeps source and canonical hashes, parser identity, page spans, and character-level evidence offsets, plus probabilities and request receipts. Results export as JSON, CSV, or Neo4j Cypher.
It does not claim that model output is true, approved, or safe for consequential reuse. The README also calls the benchmark one narrow operation, not a general model ranking or an end-to-end knowledge-graph score.
More Jev resources like this
A local app that reads a few lines of notes about your product and lets Jev pick its way to a finished landing page: Jev never writes text, it only chooses between options, and code builds the page from those picks.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
sagapranav/jev-pagesHTML
sagapranav/jev-pagesHTMLClassifies and splits PDFs, DOCX, and PPTX with local page text from LiteParse and category decisions from Jev.
Jev Tools & CLIs#community#python#document
Community
jerryjliu/docjevPython
jerryjliu/docjevPythonA Rust CLI that streams Kubernetes logs, triages them with local offline rules or typed Jev questions online, and presents an interactive terminal UI plus JSON reports. Collection is read-only and every judgment is advisory.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
sunil-sadasivan/jevernetesGo
sunil-sadasivan/jevernetesGoAn Ollama-style local runner for System One decision models from Hugging Face: pull, serve and query them behind the same /v1/systemone API as TypeSafe's Jev, with a terminal model manager and a browser playground.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
nvkudva/ollajevPython
nvkudva/ollajevPythonA Python library shaped like the re module that finds, replaces, splits and extracts text by meaning: loose finders propose spans and Jev judges each one with a probability. Every match stays a verbatim slice of the input with its offsets, so it can be wrong but never invented.
Jev Repos & SDKs#community#github#python
Community
nothans/ex-regexPython
nothans/ex-regexPythonA 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-PythonFrom 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.