Docs
CommunityJev Cookbooks
Infinite Cardinality Choice
Official TypeSafe cookbook that finds a half-remembered film among 5,000 by running one Noul question per film, re-reading the top 80 with longer plots, and putting the top 20 into a single Choice question.
- Format
- Docs
- Reading time
- 2 min
- Published by
- Community
- Added
- Nov 2026
Highlights
- Three-step search over a 5,000-film catalog: one Noul per film, a re-read of the top 80, then one Choice over the top 20.
- The right film comes first 78% of the time, against 33% when embedding search builds the shortlist.
- Choice questions take at most 255 options, and only about forty when each option is a long plot summary.
- Each of the 48 times the Choice question gave its top film 0.8 or more, that film was right.
- Numbers measured on jev-1.14.0 in November 2026; the run replays a downloaded cache by default.
Questions about this entry
Short answers drawn from the linked project or page.
Install openai, ipython, matplotlib, and cooksafe>=0.3.0,<0.4.0 with pip, then set TYPESAFE_API_KEY and, for the embedding search the results are measured against, OPENAI_API_KEY.
No. With DOWNLOAD_CACHE set to True the code downloads the published answers into infinite_cardinality_choice_cache.json, so the run makes no requests. To run live, set DOWNLOAD_CACHE to False and delete that file; a cache already on disk is never overwritten.
The API enforces a limit of 255 options, and 256 options is rejected outright. When each option is a long plot summary, only about forty fit in one Choice question, which is why the search narrows the catalog in earlier steps.
The right film comes first for 78% of the 100 requests. When embedding search picks the twenty films for the final step instead, the right film comes first for 33% of them.
A films.jsonl catalog of 5,000 films with title, year, and two versions of the Wikipedia plot section fetched 2026-11-07, plus 100 requests from Microsoft's tip-of-the-tongue dataset collected from the forum iRememberThisMovie.com.
The numbers came from jev-1.14.0, run in November 2026. The cookbook sets TYPESAFE_MODEL to that version, and jev-latest runs the newest model live.
Its top probability tells you whether to show one film or a short list. Each of the 48 times the Choice question gave its top film 0.8 or more, that film was right.
More Jev resources like this
Official Jev cookbook that checks whether an LLM's answer follows from the paragraphs it was given, using three Noul questions in one request and a risk score that sends doubtful answers to review.
Jev CookbooksDocs#python#guardrails#evaluation
Community
A text dataset of typed decision examples in parquet format, sized 1M to 10M rows. Rows carry state, question, target distribution, and provenance fields drawn from sources like MedQA and procedural rule-firing tasks.
Jev Cookbooks#catalog-import#python#classification
Community
Two-stage Jev control of a Franka Panda in MuJoCo: Jev selects a task intent, then X/Y/Z directions and a gripper command executed against real contacts, with 43 of 50 Jev episodes succeeding versus 48 of 50 for a rule baseline.
Jev Cookbooks#community#python#robotics
Community
lykycy123/RoboJEVPython
lykycy123/RoboJEVPythonPokémon Red under PyBoy where ordinary Python owns the route and arithmetic and Jev is asked only at branches, with each battle turn's faint prediction scored by Brier.
Jev Cookbooks#community#python#games
Spanish-language course repository on Jev and System One, with five Google Colab labs written in pure Python stdlib: fast router, hybrid escalation, realtime guardrails, tool dispatch, and a 1,000-request benchmark, plus an English technical book PDF.
Jev Cookbooks#python#github#routing
Community0utKast/Jev_AI
Independent experiment scoring 400 real job postings against 30 fictional résumés with Jev, producing 12,000 evaluations in about 197 seconds for roughly $0.17, with a 100-posting answer key and an analysis of what the model got wrong.
Jev Benchmarks & Evaluations#github#madewithjev#evaluation
Communitykaucap08/jev-job-matching
From 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.