User repository on GitHub
24/7 market-making system around Jev (TypeSafe System One) decisions: deterministic state, calibrated judgments, hard risk vetoes.
- Forks
- 9
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Repos & SDKs
jev-trader (market-making system)
A 24/7 market-making system built around Jev: deterministic code computes the market state, Jev answers six atomic judgments with calibrated probabilities, and the policy engine applies thresholds while hard risk vetoes stay in code.
- Stars
- 44
- Language
- Python
- License
- MIT
- Reading time
- 3 min
- Published by
- Community
- Author
- buberlo
- Last verified
- Oct 2026
This is one of several Jev listings for the same project. The canonical entry is Jev Trader — jarrodwatts.
Highlights
- One Jev call per block answers six atomic judgments: regime, direction, toxic_flow, liquidity_stressed, quote_environment, inventory_pressure.
- Paper runs are deterministic for a given --seed and need no key; triples (state, decision, outcome) log to data/calibration.jsonl.
- Hard risk rules are an absolute veto and are never delegated to the model; a kill switch flattens and stops when limits break.
- Calibration reports Brier score, log loss, expected calibration error and a reliability curve, then fits Platt scaling to your venue's data.
- MIT, Python, 44 stars; created and last pushed 2026-09-19, with Phase 0 (foundation and paper loop) marked complete in the roadmap.
Quickstart
bash
uv sync
uv run jev-trader paper --blocks 2000
export TYPESAFE_API_KEY="..."
uv run jev-trader paper --blocks 10000
uv run jev-trader calibrateWatch out
README reviewed 2026-10-17; the repo is for research and education, not financial advice, and the README warns to run in paper mode and on a testnet before risking capital. Only Phase 0 is complete, no live venue adapter ships, and the digest shows no performance numbers.
Questions about this entry
Short answers drawn from the linked project or page.
A 24/7 market-making trading system built around Jev, TypeSafe AI's System One decision model. Its stated principle: Jev should not own the trading system but selected judgments inside it, so code calculates the state, Jev interprets it, code applies policy, and execution places orders.
Each block builds a deterministic state snapshot under 400 tokens and sends one Jev call with six atomic judgments. The typed probability vector goes to a policy engine whose thresholds live in code, then hard risk rules can absolutely veto, and execution places post-only quotes.
Yes. uv run jev-trader paper --blocks 2000 runs the paper loop in fallback mode with no key, using a deterministic fallback judge. Live Jev judgments need TYPESAFE_API_KEY exported, and a --realtime flag paces runs at 300 ms blocks instead of accelerated simulation.
Every block logs the triple (state, decision, outcome) to JSONL, and jev-trader calibrate reports Brier score, log loss, expected calibration error and a reliability curve, then fits Platt scaling. The README notes RLCD calibrates against TypeSafe's distribution, not your venue's.
Hard-coded limits with an absolute veto and a kill switch that flattens and stops, plus a fallback ladder: high confidence operates normally, low confidence reduces size, a missed block deadline holds rather than quoting stale state, and an unavailable Jev falls back to deterministic judgments.
MIT, written in Python, with Phase 0 (foundation and paper loop) complete per the roadmap. The digest shows 44 stars, 9 forks, and creation and last push both on 2026-09-19, so activity since then is not visible in the material reviewed.
No. Several unrelated projects share the name, including jarrodwatts/jev-trader, a Monad block-trading bot, and renatosousa/jev-trader, an FX market-analysis bridge. This one is a market-making system with an Avellaneda-Stoikov pricing module, a six-question judgment battery and a calibration loop.
Reactions & coverage
Posts, threads, and videos about this entry from around the web.
X: A trading loop with Jev and GPT-6 Astra
X: News sentiment in a trading terminal
X: A Jev trading bot on Telegram
X: A Jev and Grok copy-trading bot
This post couldn't be loaded here — read it on X.https://x.com/Mnilax/status/2102760642872508510More Jev resources like this
A reproducible PyTorch stack for cross-sectional factor research: 213 mask-aware factors, cost-aware backtests and auditable reports, plus a preview typed portfolio-decision adapter in the System-One/Jev style.
Jev Repos & SDKs#community#github#python
Community
initial-d/ml-quant-tradingPython
initial-d/ml-quant-tradingPythonA Python library that turns Jev-style decision models into ordinary function calls: check returns a bool, classify picks one option, label picks every option that applies, and score rates on a scale, with probabilities behind every answer.
Jev Repos & SDKs#community#github#python
Community
jlowin/vibecheckPython
jlowin/vibecheckPythonA Bangla-focused adaptation of Laya, the non-autoregressive typed-decision model: targeted data preparation, fine-tuning and evaluation for classification, routing and scoring on Bangla, Banglish and code-switched text, with the checkpoint still to come.
Jev Repos & SDKs#community#github#python
Community
shamspias/laya-banglaPython
shamspias/laya-banglaPythonA frozen decision server that serves Gemma 4 31B — quantized to FP8 at startup — in TypeSafe's /v1/systemone format, reading the next-token distribution instead of generating tokens. It needs one FP8 NVIDIA GPU with 80 GB and has no authentication.
Jev Repos & SDKs#community#github#python
Community
krishna-gogineni-765/deck31bPython
krishna-gogineni-765/deck31bPythonPython with a smart if: .jev files add the ~ question operator and jev/case blocks that Jev answers with a number or a label, never text. Files are rewritten to plain Python at import time; every ~ is one paid TypeSafe API call.
Jev Repos & SDKs#community#catalog-import#logicrw
Community
sumanmichael/jevlangPython
sumanmichael/jevlangPythonA pre-release Python framework for predictions on mixed text and tabular data: Jev Noul and Score questions turn text columns into numeric features, then a tabular foundation model such as TabPFN learns from labeled rows in context. Offers hosted and fully local backends.
Jev Repos & SDKs#community#tabular#predictions
Community
edamame-labs/tab-jevPython
edamame-labs/tab-jevPythonFrom 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.