beliciobcardoso/systemone_gate
User repository on GitHub
Motor de Decisão Ultrarrápido e Local para Agentes de IA baseado em Ollama System One (JEV)
- Forks
- 0
- Last pushed
- Updated Oct 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
SystemOne Gate
A local decision engine for AI agents and software developers, built on the System One architecture in TypeSafe's Jev style and running fully offline through Ollama 0.35+. Ships a CLI, an MCP server and a Python library for error triage, diff review, shell guardrails and routing.
- Stars
- 1
- Language
- Python
- License
- MIT
- Reading time
- 2 min
- Published by
- Community
- Author
- beliciobcardoso
- Last verified
- Oct 2026
Used inJev for Software & DevTools
- Languages
- Python
- Capabilities
- Open sourceRuns locally
Highlights
- Runs decision models fully offline through Ollama 0.35+, with no cloud token cost and no telemetry.
- Ships as a CLI (diff, triage, guard, install-hook), a zero-dependency MCP server with five tools, and a Python library.
- Measured on one RTX 3060: tev1:0.8b at 162.6 ms p50, nimble at 391.6 ms p50 for guard and diff rubrics.
- Deterministic guard rules answer in about 40-50 microseconds and stay the fast path; model verdicts are additional heuristics.
- MIT, Python 3.9+, with pattern-based secret redaction before any model call.
Quickstart
bash
git clone https://github.com/beliciobcardoso/systemone_gate.git
cd systemone_gate
pip install -e .Watch out
README and GitHub metadata reviewed 2026-10-17 from the round-eleven capture; the digest truncates the feature list mid-sentence, and the README's latency numbers are explicitly measured on one RTX 3060 and not re-measured since the rubrics moved to English.
Questions about this entry
Short answers drawn from the linked project or page.
A local decision engine for AI agents and software developers. It evaluates structured input in parallel, returning 1 to 3 output tokens with the model's probabilities, and runs fully offline through Ollama 0.35 or newer instead of a cloud text model.
The README says it is based on the System One architecture in TypeSafe AI's JEV style. It does not call the hosted Jev API: it runs decision models locally through Ollama's /v1/systemone endpoint, with a note that its probabilities are not calibrated.
Four documented uses: error triage for compile, syntax, linker, memory-leak and timeout failures; diff risk review before a commit; a shell command guardrail; and subagent routing for development tasks.
Install Ollama 0.35 or newer, pull tev1:0.8b and nimble, then clone the repository and run pip install -e . That gives you the systemone-gate CLI and the systemone-mcp server; systemone-gate doctor diagnoses the backend.
The README's measurements come from an RTX 3060 with 12 GB, where nimble occupied 8.9 GB while loaded and took 12 to 72 seconds for a cold load. On a modest machine it recommends using only the 811 MB tev1:0.8b model for every task.
Probabilities are explicitly not calibrated, and a calibration over 84 commands with two labelers found no tev1:0.8b or nimble threshold reaching 90% recall at 5% false positives, so the deterministic rules layer was widened instead of trusting the model verdict.
No. Requests go to localhost:11434, remote endpoints are rejected unless SYSTEMONE_ALLOW_REMOTE=1 is set, secrets are pattern-redacted before any model call, and the project states it collects no telemetry; redaction is risk reduction, not a guarantee.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: Jev in front of the HeyGen MCP
X: Playwright CLI + Jev vs Playwright MCP
X: A local MCP server that cuts computer use from 23s to 0.8s. Cartwright built Cortex, a local MCP server where…
X: A side panel that drives any site's WebMCP tools
More Jev resources like this
An explainable intent router that cascades regex, TF-IDF with CatBoost, System One decision models (Laya, Jev, CLM) and a generative LLM fallback, answering from the cheapest engine able to decide and leaving a replayable decision record. Runs fully local.
Jev Tools & CLIs#community#github#python
A safety guardrail for coding agents (Claude Code, Codex, Cursor, Antigravity, Grok) that intercepts shell commands, script executions and file writes before they run. Its neural tier can use Liquid AI d1-3B, CLM, TypeSafe Jev cloud or pure rules; a deterministic tier blocks destructive patterns in under 0.1 ms.
Jev Tools & CLIs#community#github#security
Community
Bl4ckk01/agent-guardrailPython
Bl4ckk01/agent-guardrailPythonA local, open-source personal AI assistant for your job: mail, chats and tickets become tasks, agents you already use such as Claude Code, Codex and Gemini do the work, and nothing is sent until you approve. The current release is v0.3.8.3.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ldbumble/taskuaryPython
ldbumble/taskuaryPythonA 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-onePythonMIT-licensed Python tool router for coding agents that keeps obvious calls on a fast path, asks Jev (default typesafe/jev-1.13 via OpenRouter) for genuinely ambiguous choices, and can run bounded MCTS when multi-step consequences matter. Ships a CLI, MCP server and hook integrations.
Jev Tools & CLIs#community#routing#mcp
An 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/ollajevPythonFrom 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.