Is any of the P&L in here real money?

No. Execution is paper: it records fills, marks them and reaches no exchange. There is no wallet, no key, no live orders and no --live flag, and the README says going live means replacing one file with a real adapter that is deliberately not included. Any performance number is simulated over generated data.

Am I getting real Jev decisions by default, or a stand-in?

A stand-in. The default engine is a small scoring function over the state features (momentum, 24h change, volume, news, regime) with every contribution readable, and it is labelled offline on every decision so it is never mistaken for the model. The real TypeSafe call needs --engine jev plus TYPESAFE_API_KEY, and that endpoint is in gated early access.

What stops a confident wrong call from reaching the market?

The risk gate, deliberately kept in a separate file from the decision layer. It refuses when confidence is below the floor, when probability is below the floor, when the action is HOLD or AVOID, and when more than five positions are already open, and it names the single rule that bound on every skip. Acting on a decision stays the developer's job.

Where does the market data actually come from?

An offline generator across stocks, crypto and memes by default, which is why the single-symbol card shows synthetic fields like momentum and regime next to the price. Live feeds plug in behind the same state shape, and structured decisions over a live feed plus historical replay are listed on the roadmap rather than shipped.

How do I install and run it?

Python 3.10 or newer. Clone the repository and optionally pip install -e . to get the jev-bot command; python -m jev_bot decisions runs a cycle, run executes the approved paper trades, and card BTC unpacks one decision. There is a Makefile (make install, make test, make decisions) and a Docker path that needs no local Python.

What does a decision look like as data?

A small typed object: symbol, action (BUY, SELL, HOLD or AVOID), a calibrated probability, a confidence, and a source naming which engine made it, either offline or jev. The model produces the decision and the execution layer separately decides whether it may reach the market.

What are the gate's default refusals?

It refuses when confidence is below 70%, when probability is below 60%, when the action is HOLD or AVOID (which never execute), and when more than five positions are already open, naming the single binding rule on every skip.

CommunityJev Practices & Patterns

Jev Bot — bl888m

A market decision loop that feeds simulated stock, crypto and meme states to Jev or a local engine, then gates each answer. Execution is paper only and there is no live flag.

See highlights

Open github.com

Stars: 13
Language: Python
Published by: Community
Author: bl888m
Last verified: Oct 2026

bl888m/jev-bot

User repository on GitHub

View on GitHub

JEV-powered market decision bot for stocks, crypto and memes. State in, BUY/SELL/HOLD/AVOID out, paper by default

Forks: 5
Last pushed: Updated Oct 2026

Repo stats from the GitHub API, cached Oct 2026.Homepage

This is one of several Jev listings for the same project. The canonical entry is Jev Bot — nssmd.

About this project

JEV-powered market decision bot for stocks, crypto and memes. You watch the market. JEV makes the decision.

Quoted from the project's README. Read the full README.

Tagscommunity awesomejev catalog-import

Works forJev for Classification

Used inJev for Software & DevTools

Languages: Python

Highlights

Standard library only: zero runtime dependencies, 17 offline checks, and a Docker path if you would rather not install Python.
Two engines, labelled on every decision: a transparent local scoring function by default, or the real model with --engine jev.
The risk gate refuses on confidence, probability, action type and open exposure, and names the single binding rule.
HOLD and AVOID never execute, which is why flat and bearish rows come back as skips in the decisions table.

Watch out

Market data is simulated by default by an offline generator across stocks, crypto and memes, so any P&L shown is paper over generated data, and the README makes no claim that the model predicts markets or produces returns. The default engine is a local scoring function rather than Jev; the real model is off unless you pass --engine jev and set TYPESAFE_API_KEY, and it is in gated early access with the endpoint and key coming from the environment. Execution is paper by default and the only mode: no wallet, no key, no live orders, no --live flag, and the live adapter is deliberately not shipped. Live feed support is roadmap work. Copy reviewed against the repository README on 2026-10-04, not from running it.

Questions about this entry

Short answers drawn from the linked project or page.

More Jev resources like this

jev_local

Local unofficial /v1/systemone adapter that serves Choice, Score, and Noul probabilities via first-token LLM judgments or a fine-tuned ModernBERT scorer, including Japanese docs.

Jev Practices & Patterns#community#awesomejev#catalog-import

Community

Argos1111/jev_localPython

jev-on-a-laptop

Unofficial laptop study of parallel constrained decoding: prefill once, share the KV cache across fields, and decide all schema fields in one batched forward pass without writing JSON.