User repository on GitHub
A desktop bookkeeping clerk: Jev decides every step, Fable 5.1 rewrites its playbook every ten invoices
- Forks
- 1
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Tools & CLIs
Jev Clerk
A Mac desktop bookkeeping clerk where Jev makes every in-app decision — an action kind from a closed list of 12 and which line to click — while a System 2 model rewrites its playbook every ten invoices.
- Stars
- 8
- Language
- Python
- Reading time
- 3 min
- Published by
- Community
- Author
- stas4000
- Last verified
- Oct 2026
Used inJev for Finance & Fintech
Highlights
- Opens a supplier invoice in Preview, types it into Frappe Books, saves, submits and files the PDF; the read-only accounting database is the referee.
- Jev answers two multiple-choice questions per step at about 0.35 s and $0.0002 per decision; the System 2 model never touches the screen.
- In the recorded run, 30 real invoices over 90 iterations used 1,141 Jev calls for $0.24 and 978 s of System 2 thinking.
- Correct bookings on USD invoices go from 0/9 under playbook v0 to 9/9 by v6 and v8; block 4 regressed to 0/9 after a rewrite.
- Every playbook rewrite is validated by code so the action list stays closed, and two block-8 misses were a parser bug, not Jev.
Watch out
Copy reviewed against the repo README on 2026-10-15; the README is truncated after the run's fairness caveats, so setup instructions and licence details beyond the NOASSERTION metadata are not covered. Results come from one 90-iteration recorded run on one Mac, with the engine changing mid-run.
Questions about this entry
Short answers drawn from the linked project or page.
It works a real Mac desktop as a bookkeeping clerk: it opens a supplier invoice PDF in Preview, reads it, types it into the open-source Frappe Books app, saves, submits and files the PDF, with Jev making every decision along the way.
Each step Jev receives the screen as OCR lines and answers two multiple-choice questions: which kind of action from a closed list of 12, and which line to click. The README measures about 0.35 s and $0.0002 per decision.
Claude Fable 5.1 never touches the screen. After every 10 invoices it reads the full decision log and rewrites the playbook — Jev's question wording, the lessons it sees, typing formats, waits and reflexes — and code validates every rewrite so the action list stays closed.
The referee is the accounting database, opened read-only: an invoice counts only when a submitted Purchase Invoice with the right supplier, date and amount exists. Suppliers and expense items were seeded before the run, since the clerk never creates master data itself.
Across 30 real supplier invoices worked through three times, correct bookings on USD invoices went from 0/9 under the playbook that knew nothing about the app to 9/9 by playbooks v6 and v8, with foreign-currency flagging appearing from v1 onward.
The README totals 1,141 Jev calls for $0.24 across the 90 iterations, plus 978 seconds of System 2 thinking during the nine rewrite passes. Per-invoice Jev cost ranged from $0.0006 to $0.0065 depending on the playbook version.
Yes, and the README lists them: the engine changed mid-run (waits, direct endpoint, loop breaker, labelled OCR lines, faster screen grabs), two blocks were restarted after fixes, and the two misses in block 8 came from an amount-parser bug reading the year in a due-date string, not from Jev.
More Jev resources like this
A Mac-only voice assistant: say Hey Jev or hold right Option and Jev decides every command in one call for $0.00004 per request, while Whisper transcribes locally, Fish Audio S2.1 Pro speaks replies, and an LLM wakes only for questions.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
henryklunaris/hey-jevPython
henryklunaris/hey-jevPythonA local stock decision service that sends market data, indicators, fundamentals, and news collected by AIStock to Jev cloud or a local model and returns a structured Buy, Sell, or Hold signal with probabilities, data-quality notes, and a stored evidence record.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
ethanalgox/jev-tradingTypeScript
ethanalgox/jev-tradingTypeScriptAn experimental self-hosted trading bot for Kraken, Alpaca paper and OKX U.S. spot with Jev-assisted strategies beside deterministic programs. Code enforces sizing and execution limits; the README stresses Jev does not control those safeguards.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
kyaulabs/kairosPython
kyaulabs/kairosPythonA desktop app for dictation, audio-file transcription and meeting notes on Windows, macOS and Linux, using local Whisper models or the OpenAI API, with an opt-in MCP server and local history API so agents can search saved transcripts.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
knuckles92/openwhisperPython
knuckles92/openwhisperPythonA schema-guided document-to-graph pipeline: local parsing feeds entity mentions and bounded candidate pairs to Jev, which chooses among closed-set relations, and every edge lands as proposed, review, or rejected with page-level evidence.
Jev Tools & CLIs#community#catalog-import#madewithjev
Community
chenmingtang830/jevgraphPython
chenmingtang830/jevgraphPythonTraffic signal control with Jev applied to SUMO simulation: each control step packs the road network into one state with a choice question per intersection, and the model's chosen phase, distribution and confidence drive the signal with local-ranking fallbacks.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
usail-hkust/jevlightPython
usail-hkust/jevlightPythonFrom 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.