Documents arrive as PDFs, scans, notes and messy text. The job is to turn them into fields your code can use: a form type, a date, a yes or no for each question in a template. A decision model does this by answering typed questions about the document, one field at a time, instead of writing JSON that may or may not parse. Every answer has a type and a confidence value. The examples in this hub run on Jev and other decision models, and Laya appears in some. Examples include kyotofin/tax-doc-classifier, which picks an IRS form for each PDF page, Jev Reviewer by Ahmad Sofi-Mahmudi, which answers extraction forms for systematic reviews, and gtaras7/typesafe-jev, a local CV screening workbench. Before you copy one, check what happens to the document text, how long a document can be, how dates and numbers are handled, and whether a person reviews the low-confidence fields. Most entries are small, so you can read the whole flow before you adopt it.
8 entries
Showing 8 of 8
A TypeScript classifier that sends each PDF page's text to Jev as a choice over 261 IRS forms and 7 page kinds, with a second call only for five corporate forms and their schedules. Its eval reports 0 wrong pages on 314 filled TaxCalcBench forms at about $0.001 per page, 34x cheaper and 6x faster than the Sonnet pipeline it replaced.
Nakshatra Saxena (@nedwize)Openkyotofin/tax-doc-classifier on github.com419 starsCost per page $0.001vs prior setup 34x cheaper
Data extraction for systematic reviewers, built on Jev. It answers extraction forms and risk-of-bias templates like RoB 2, ROBINS-I, QUADAS-2 and TIDieR against a trial report and its supplements, with every answer a verbatim quote and page number to check against the source. Files stay in the browser.
Ahmad Sofi-Mahmudi (@ASofiMahmudi)OpenJev Reviewer on github.com33 starsInput price $0.042/M tok
Local CV screening workbench that asks Jev a small fixed set of typed questions about each CV in a folder, scores the answers with arithmetic you can read, and shows a sortable shortlist. Judgments are stored apart from the arithmetic, so editing the role policy re-scores every candidate in about 20 ms at no cost.
Compares GLiNER 2.5, GLiFormer, GLiClass, Laya, Von and Jev on zero-shot information extraction and classification in one web UI.
Paired comparison of Jev and OpenAI structured outputs on 1,000 synthetic outpatient notes, both given the same note, policy, candidate values and ten typed questions. OpenAI got 98.7% of records fully correct against Jev's 92.4%, while Jev cost about 133 times less with 5.2 times lower mean latency.
A video demo of a per-page PDF router built with Jev that decides which pages need OCR and extracts the rest locally, aiming to cut both cost and processing time.
Extracts dates from text using pre-parsing plus Choice questions, working around the documented weakness of jev-1.13 at date arithmetic.
Recovers structured fields from unstructured plain text by asking typed questions rather than prompting an LLM for JSON.
Nothing matches those filters
Clear a filter to widen the search, or send us the thing you were looking for.
The loop is: split the document into pages or sections, ask a fixed set of typed questions for each, and let code assemble the record. gtaras7/typesafe-jev asks a small fixed set of questions about each CV and scores the answers with arithmetic you can read. kyotofin/tax-doc-classifier sends each PDF page's text as one choice over 261 IRS forms and 7 page kinds.
Choice fits picking a type from a list. Noul fits yes or no template questions. Score is rare here. The TypeSafe AI cookbooks on structure recovery from plain text and on date extraction show the pattern, and the date entry notes it pre-parses the text to work around the model's weakness at date arithmetic. Run your own sample before you trust a pattern.
Free text is the limit. A decision model picks among options you supply, so it does not copy a long passage out on its own. vclic/smoking-extraction-benchmark compares Jev with OpenAI structured outputs on synthetic outpatient notes, with both given the same candidate values and questions. Plan to supply the candidates. Candidates can come from a form template, a list of known values or an earlier extraction step.
It can classify pages and answer typed questions about a document's text, as kyotofin/tax-doc-classifier and Jev Reviewer do. It picks from options you supply, so you give it the candidate values or fields. OCR and layout reading happen before it.
It returns typed answers with a confidence value, which makes review easy. An LLM can copy free text out of a document, which a decision model does not. vclic/smoking-extraction-benchmark compares the two on one task.
Yes. The page-by-page OCR router by Misbah Syed uses Jev to decide which PDF pages need OCR and extracts the rest locally.
