What stops this from placing trades on its own?

It cannot place, cancel, or mutate trades, has no embedded LLM in the core, and review states never trigger execution. The review assistant is opt-in and only sends redacted structured fields to your configured provider.

Where does Jev fit if the control plane is deterministic?

Jev is an optional typed-judgment layer with TypeSafe direct and OpenRouter backends for noul, choice, and score questions. Arithmetic, date comparison, and temporal boundary checks stay in deterministic Python.

What does the bundled finance benchmark actually test?

The repo ships finance-jev-v1, a frozen offline suite of 240 cases across trading-review, financial-filings, industry-events, and macro-policy, with a published reference run and replay commands.

How do I verify the loop works before trusting it?

Install with pip install -e .[dev], run smcub doctor, then capture-run, build-outcome, build-evidence-pack, and replay-evidence-pack over the toy sample. The package is not on PyPI and releases ship wheels from GitHub.

Which two backends can Jev use inside SmartMoney-Cub?

Typesafe direct and OpenRouter. The OpenRouter backend is reported as not_run in the published reference run because its credentials were unconfigured.

How accurate is Jev against the deterministic baseline?

On the 240-case finance-jev-v1 suite the deterministic rule baseline reaches 83.33% overall accuracy and the Jev typesafe_direct backend 78.43%. Jev leads on financial-filings at 77.00% versus 73.33% and has better calibration, an ECE of 0.1464 against 0.1667.

What does the sandbox flag actually do?

The --sandbox flag only selects a disposable tmp/sandbox output namespace; it does not isolate the process. Run untrusted commands inside an OS or container sandbox.

CommunityJev Tools & CLIs

SmartMoney-Cub

Local-first trading journal and review harness that is read-only over markets, captures runs into frozen envelopes and evidence packs, and requires explicit human promotion before any rule returns.

See highlights

Open github.com

Stars: 27
Language: Python
License: MIT
Published by: Community
Author: myc0576
Last verified: Oct 2026

myc0576/SmartMoney-Cub

User repository on GitHub

View on GitHub

Read-only trading journal and review harness: Jev typed judgments, agent integration, and a reproducible finance benchmark. No orders, no advice.

Forks: 1
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

smartmoney-cub-harness is a local-first, agent-agnostic trading journal and review harness: read-only over markets and execution, writable over your own journal. It turns an external caller's offline run into portable, reviewable artifacts without taking trading authority. External Agent or CLI caller → Run Envelope → frozen Benchmark/Evidence Pack → deterministic replay → explicit human promotion gate.

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

Tagscommunity jevbest catalog-import

Works forJev for Classification Jev for Evaluation Jev for Tool calling

Used inJev for Software & DevTools

Languages: Python

Highlights

External caller to run envelope to frozen benchmark and evidence pack to deterministic replay to a human promotion gate.
Optional Jev layer over structured judgments only, with TypeSafe direct and OpenRouter backends.
Frozen finance-jev-v1 suite with 240 cases across four tracks and deterministic replay tooling.
Read-only safety posture with JSON-schema envelopes, sealed pack manifests, and redacted review surfaces.

Watch out

Core runs offline on toy data with no broker connection and no order placement; --sandbox only selects the tmp/sandbox output folder and run permissions are a declarative record. 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-seo

547

Live SEO audit that crawls a site from one homepage URL, checks 52 rules tied to Google Search Central, and asks Jev typed questions about every page to rank fixes.