User repository on GitHub
给任意的模型增加decision决策接口
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev — Logits-Based Decision API
A small HTTP service that adds a decision interface to any chat model: POST /v1/decision evaluates a finite set of candidate labels in one forward pass, reads next-token logits and returns structured JSON without model.generate() or output decoding.
- Stars
- 0
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- siyehua
- Last verified
- Oct 2026
Used inJev for Software & DevTools
Highlights
- Adds a dedicated POST /v1/decision interface while keeping chat and generation separate.
- Evaluates a finite label set via one forward pass, reading next-token logits — no model.generate() and no decoded output tokens.
- Backend: Hugging Face Transformers causal LM, default Qwen/Qwen2.5-0.5B; strategy label_logits; FastAPI + Uvicorn.
- SPEC.md §7.6 limit: probabilities are softmax-normalized relative label scores, not calibrated probabilities of real-world correctness.
- Implements SPEC.md v0.1 as an MVP; the Chinese description says it adds a decision interface to arbitrary models.
Quickstart
bash
pip install -r requirements.txt
python run.py --host 127.0.0.1 --port 8000
# model loads once at startup; ready when you see "service ready"Watch out
README and GitHub metadata reviewed 2026-10-17; the digest truncates the curl example mid-request, so the full request and response schema is not captured. The project itself warns that returned probabilities are relative label scores, not calibrated probabilities.
Questions about this entry
Short answers drawn from the linked project or page.
A small HTTP service that adds a dedicated decision interface to chat/generation models: POST /v1/decision returns structured JSON decisions over a finite set of candidate labels.
It performs one forward pass, reads the next-token logits for the candidate labels, and returns a structured JSON result — it never calls model.generate() and never decodes output tokens.
A Hugging Face Transformers causal LM, Qwen/Qwen2.5-0.5B by default, with the label_logits strategy and a FastAPI + Uvicorn server.
No. SPEC.md §7.6 says the returned probabilities are softmax-normalized relative label scores over the labels in one request, not calibrated probabilities of real-world correctness.
Install requirements, then run python run.py --host 127.0.0.1 --port 8000; the model loads once at startup and the service is ready when 'service ready' appears.
It implements SPEC.md v0.1 as an MVP per the README; the digest shows a Python repository with 0 stars, created and pushed 2026-10-15, and no licence listed.
No, it is an independent implementation of the decision-interface idea on local Hugging Face models; the digest shows no TypeSafe API usage, and SPEC.md §7.6 documents its interpretation limit.
More Jev resources like this
A 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-opencvPythonAn 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 proof-of-concept HTTP service for constrained decisions with a local MLX model: send text or JSON plus typed choice, score, or noul questions and it returns probabilities over the answers you declared — no generated text, no parsed model JSON.
Jev Tools & CLIs#community#github#python
Community
snellingio/system-onePython
snellingio/system-onePythonA 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 System One layout optimizer for decision-model workloads: it reorders rows and fields so a prefix-caching backend reuses input computation, then returns per-record decisions in original order. Its bundled backend serves AutoTrust JEV-9B through vLLM.
Jev Repos & SDKs#community#github#python
Community
0814wdwd/solo_jevPython
0814wdwd/solo_jevPythonA drop-in proxy that distills a repeated Jev classification call into a local model as it runs: it learns from Jev's own probability distributions, then answers most requests itself while uncertain ones and a 2% audit slice keep going to Jev.
Jev Tools & CLIs#community#python#distillation
Community
tomerglick57/JevstillerPython
tomerglick57/JevstillerPythonFrom 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.