Organization repository on GitHub
Jev-powered traffic signal control on CityFlow with structured phase and green-time decisions.
- Forks
- 2
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
JevLight
Traffic signal control with Jev applied to SUMO simulation: each control step packs the road network into one state with a choice question per intersection, and the model's chosen phase, distribution and confidence drive the signal with local-ranking fallbacks.
- Stars
- 10
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- usail-hkust
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceSelf-hostedRequires API key
Highlights
- One shared state with one choice question per intersection means the whole network decides in a single API call per step.
- Maps a junction snapshot to JSON and 'which phase next?' to a choice over ETWT / NTST / ELWL / NLSL, plus a congestion noul.
- Empty junctions never call the API; failed or low-confidence answers keep the previous signal or use the local ranking.
- Three transports: the Jev community MCP server, official TypeSafe /v1/systemone, or a self-hosted Jev-compatible server.
- Ships a 12-intersection SUMO scenario under data/Jinan/3_4, a three.js visualizer and unit tests with no network.
Quickstart
bash
pip install -r requirements.txt
export SUMO_HOME=/path/to/sumoWatch out
Grounded on a truncated README and GitHub metadata captured 2026-10-15; the GitHub description says CityFlow while the README documents SUMO and SUMO setup, an inconsistency worth checking before adopting it. No license is listed and the README excerpt carries no benchmark results.
Questions about this entry
Short answers drawn from the linked project or page.
It applies Jev, the System One decision model, to traffic signal control in SUMO simulation, building on the observation and control-flow conventions of LLMLight / CoLLMLight. One request sends a state plus typed questions and returns one typed answer per question.
A junction snapshot — queue, approaching vehicles by segment, waiting times — becomes the state. 'Which phase next?' becomes one choice question with the phases ETWT, NTST, ELWL and NLSL as the options. A speculative noul logs whether the junction is congested.
The chosen phase plus its full distribution and confidence come back as answers.<id>.choice, .probabilities and .confidence. Failed, rate-limited or unparsable requests keep the previous signal by default, or use the local waiting-time ranking with --fallback ranking.
By default no: every active intersection is packed into one shared state with one question each, so the whole network decides in a single call per control step. With --packaging per_intersection each junction gets its own request, matching LLMLight's pattern.
Junctions with no vehicles never call the API; they use the exact CoLLMLight local waiting-time ranking. Answers with confidence below --min_confidence are rejected the same way as failures.
Python 3.10+, SUMO 1.20 or newer with traci and sumolib on PYTHONPATH or installed via the eclipse-sumo pip package, plus the Python dependencies from requirements.txt. Three transports are supported: a community MCP endpoint, the official TypeSafe API, or a local server.
The repo includes a 3D SUMO visualizer bridge, a three.js frontend under viz/, and optional W&B logging. Unit tests run without network or SUMO, and a 12-intersection scenario ships under data/Jinan/3_4.
More Jev resources like this
A governed execution layer and drop-in proxy between typed-decision models like Jev and callers such as Claude Code, Cursor/MCP and SDKs: raw probabilities become deterministic policy actions from ALLOW to DENY with a DuckDB audit trail.
Jev Integrations & Adapters#community#catalog-import#logicrw
Community
0xshikhar/jev-fusePython
0xshikhar/jev-fusePythonModeration for communities and apps, powered by Jev: every message gets a probability for spam, scam, harassment, nsfw, doxxing and more, plus rules in plain English. Thresholds and actions are yours, and flag-only is the default.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ohernandezdev/jevmodPython
ohernandezdev/jevmodPythonLatitude is an open-source observability platform for AI agents: one line of telemetry captures traces, failing traces group into tracked signals, and fixes are dispatched to coding agents and replayed against real traces.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
latitude-dev/latitude-llmTypeScript
latitude-dev/latitude-llmTypeScriptA 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-assistantPythonToolkit, knowledge base and local dashboard for building with Jev via OpenRouter Alpha Decisions or the TypeSafe Direct API: a zero-dependency SQLite request log, a localhost REST server and Python CLI demos.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
rileycarney/jevtoolsHTML
rileycarney/jevtoolsHTMLAn 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.