User repository on GitHub
hello, jev
- Forks
- 0
- Last pushed
- Updated Sep 2026
Repo stats from the GitHub API, cached Oct 2026.
Jev directory
CommunityJev Demos & Experiments
Jev Decoder
A set of experiments that uses TypeSafe's Jev as a decoder through its Choice primitive, sampling ASCII characters, BPE tokens and grammar-constrained SQL from Jev's probability distributions, and measuring next-character loss against uniform sampling on tiny-shakespeare.
- Stars
- 0
- Language
- Python
- Reading time
- 2 min
- Published by
- Community
- Author
- panalexeu
- Last verified
- Oct 2026
- Languages
- Python
- Capabilities
- Open sourceRequires API key
Highlights
- ascii_char_gen and token_gen sample characters or BPE tokens from Jev Choice; Choice takes at most 255 options, so the tokenizer vocabulary is capped.
- char_eval measures next-character loss over a 256-character window: 7.71 nats versus 4.17 for uniform sampling, with 45 catastrophic misses above 20 nats.
- sql_gen builds SQL from a Lark grammar plus NUMBER and STRING tokens taken from the reference query; 489 of 1,034 Spider dev queries matched the reference.
- Setup uses uv scripts to fetch tiny-shakespeare and train a 255-token BPE tokenizer; the Spider archive is a separate Google Drive download.
Quickstart
bash
uv run ./scripts/dwnld_shk.py
uv run ./scripts/train_tokenizer.pyWatch out
Grounded in the truncated repository README and GitHub metadata reviewed on 2026-10-15; the author presents personal experiments rather than a library, no licence is listed, and the SQL figure is the README's own Spider dev result. The sql_gen instruction text was cut off mid-sentence.
Questions about this entry
Short answers drawn from the linked project or page.
A collection of experiments by panalexeu that leans on Jev's Choice primitive to generate text: ASCII characters, BPE tokens and grammar-constrained SQL sampled from the probability distributions Jev returns.
The README says Jev is not a decoder model but some form of encoder that evaluates state and returns typed answers and probabilities. The experiments use Choice to sample among characters, tokens or SQL grammar rules.
It asked Jev to predict the next character after a 256-character window and measured average loss of 7.71 nats over 256 predictions, versus 4.17 for uniform sampling, with 45 catastrophic misses above 20 nats.
On the Spider dev set, 489 of 1,034 generated queries (47.3%) returned the same results as the reference queries, per the README. The Lark grammar lives in sql.lark, and NUMBER and STRING tokens are taken from the reference query.
Use uv: uv run ./scripts/dwnld_shk.py downloads tiny-shakespeare and uv run ./scripts/train_tokenizer.py trains the 255-token BPE tokenizer. The Spider archive for sql_gen is a separate Google Drive download extracted into ./data.
Neither. The README presents personal experiments, the repo lists no licence, and there is no package or API beyond the scripts. Treat the numbers as one author's exploratory results on tiny-shakespeare and Spider.
Choice accepts a maximum of 255 options, so token_gen trains the BPE tokenizer with a vocabulary of 255. That cap shapes every experiment where the model must choose among characters, tokens or grammar rules.
More Jev resources like this
Text generation with a model that never generates text: jev-gpt asks Jev one typed question per choice — word type, WordNet category, a group of about 200 words split by embeddings, then the word — about 400 calls and 2 cents per prompt.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
florian-hoenicke/jev-gptPython
florian-hoenicke/jev-gptPythonLatitude is an open-source observability platform for AI agents: one line of telemetry captures traces, failing traces group into tracked signals, and fixes are dispatched to coding agents and replayed against real traces.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
latitude-dev/latitude-llmTypeScript
latitude-dev/latitude-llmTypeScriptA reproducible lab measuring how several coding agents coordinate on one repository, plus Médula, a kernel that asks TypeSafe Jev before every write whether a change collides with what the other agents are doing.
Jev Benchmarks & Evaluations#community#catalog-import#cobanov
Community
joaquinruiz/medulaPython
joaquinruiz/medulaPythonAn npm CLI that scans TypeScript, JavaScript and Python codebases for LLM calls that are really decisions, converts them into Jev modules with confidence-gated fallbacks, and shadow-tests them on recorded traffic before you swap.
Jev Tools & CLIs#community#catalog-import#logicrw
Community
0xjba/jev-swapHTML
0xjba/jev-swapHTMLBenchmark harness that scores TypeSafe Jev against any OpenRouter model or local checkpoint on labelled classification data: accuracy, calibration (ECE), confidence distribution, latency and cost across five tasks with 300 items each.
Jev Benchmarks & Evaluations#community#catalog-import#cobanov
Community
4esv/jev-evalPython
4esv/jev-evalPythonA workshop of small, inspectable Python experiments for TypeSafe System One models: ten critical bays, from fraud screening to emergency dispatch, each giving Jev a realistic state while deterministic policy keeps authority over the next action.
Jev Demos & Experiments#community#catalog-import#logicrw
Community
jgalego/jevs-garagePython
jgalego/jevs-garagePythonFrom 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.