Where do the numbers in the UI actually come from?

Live company profiles and headlines loaded through yfinance, then read by a System One model into ranked, grounded, routed trade ideas. Every number shown is a typed answer from a single request, and the Requests tab shows the exact state and questions behind any of them, so you can check what was actually asked.

Is this meant to be a working trading strategy?

No. The README says nothing here is investment advice, that profiles, prices and headlines are live and shown as retrieved, and that thresholds are starting guesses rather than values fitted to real outcomes. Read it as a demonstration of how typed answers and routing compose, not as an edge.

What does it take to run it locally?

Copy .env.example to .env with your TYPESAFE_API_KEY, then run the app's run.sh, which serves on 127.0.0.1:8765. The FastAPI server is what holds the API key, and it streams stage events to the UI over SSE rather than polling.

How much does the top-level README explain?

Setup and the file layout: yfinance loaders, cache and the state projection in one module, every question and threshold code decides on in another, the five stages plus call recording, composition and routing in a third, the mechanism catalogue, the server and the static UI. The five stages themselves, the autonomous demo tour and the data sources are documented separately in app/README.md.

Does any of my data leave the machine?

Two places, and only two. The FastAPI server holds TYPESAFE_API_KEY and binds to 127.0.0.1, so the key stays local, and the decisions go to TypeSafe. Market data comes from yfinance, and the README states profiles, prices and headlines are live and shown exactly as retrieved, with nothing normalized or pinned to a point in time.

Where would I change a threshold or a question?

Every question and every threshold the code decides on lives in app/desk/questions.py, while the five stages, call recording, composition and routing sit in app/desk/pipeline.py. Market loading and the state projection are in app/desk/market.py and the mechanism catalogue in app/desk/fixtures.py, so retuning the demo does not mean touching the stage logic.

What does the app-level README add?

app/README.md covers the five stages, the autonomous demo tour and the data sources, none of which is in the top-level README. It is also where the project states what the demo is meant to show, which is the piece to read before judging whether the routing is doing anything.

CommunityJev Practices & Patterns

Research Desk

A Jev demonstration that pulls live company profiles and headlines from yfinance and reads them into ranked, grounded, routed trade ideas, where every number in the UI is a typed answer from one request.

See highlights

Open github.com

Stars: 3
Language: Python
Published by: Community
Author: 0xnairb
Last verified: Oct 2026

0xnairb/research_desk

User repository on GitHub

View on GitHub

TypeSafe Jev demonstration for new analyzation — experimenting with Jev for fast analysis of news and tickers

Forks: 0
Last pushed: Updated Sep 2026

Repo stats from the GitHub API, cached Oct 2026.

About this project

A TypeSafe Jev demonstration: live company profiles and headlines from yfinance, read by a System One model into ranked, grounded, routed trade ideas — fast analysis of news and tickers. Every number the UI shows is a typed answer from one request, and the Requests tab shows the exact state and questions behind any of them. See app/README.md for the five stages, the autonomous demo tour, the data sources, and what the demo is meant to show.

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

The Requests tab shows the exact state and questions behind any number the UI displays, so each figure traces back to one call.
The FastAPI server holds the API key and streams stage events over SSE, serving the app locally on 127.0.0.1:8765.
Code is split by responsibility: yfinance loaders and cache, every question and threshold, the five stages, the mechanism catalogue.

Watch out

The README states plainly that nothing here is investment advice. Profiles, prices and headlines are live and shown exactly as retrieved, so nothing is normalized or pinned to a point in time, and the thresholds are starting guesses rather than values fitted to real outcomes — the routing logic is a demo of grounded typed answers, not a validated strategy. The top-level README covers setup and layout only; the five stages, the autonomous demo tour and the data sources are documented in app/README.md, which is not part of this digest. 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.