Cookbooks - TypeSafe

Each cookbook is a worked example: a real dataset, the TypeSafe questions that decide something about it, and the code that turns those decisions into a working system. Read one when you want to see how the primitives and patterns come together on a concrete problem, or copy one as the starting point for your own. This section assumes you know the TypeSafe primitives and understand how confidence works. If not, read those first. Each cookbook is listed once, under the problem it solves. Many also use techniques from other sections.

LLMs and agents

Use Jev alongside an LLM or agent: pick tools and skills, screen what goes in and out, and check what comes back.

Cookbook	What it does	Level
Function calling	Turns natural-language trading requests into calls to ordinary typed functions by mapping function names and closed-set arguments to confidence-aware TypeSafe questions.	Intermediate
Skill suggestion	Picks at most one skill for an agent turn out of the 182 in Nous Research’s Hermes catalog, using two TypeSafe requests to rank and re-check the top candidates.	Intermediate
Guardrails for LLMs	Screen every message going into and out of an LLM app with one TypeSafe request, thresholding hazard probabilities and severity to pass, review, block, or route.	Intermediate
Double-checking citations	Catch wrong or hallucinated citations by checking against the source document. One Choice question decides whether the quote’s context supports the claim.	Beginner
Classifying RAG passages	Score each retrieved passage with one TypeSafe request, then decide in code which ones reach the answering model.	Intermediate
Simple memory system	Let Jev decide which requests and answers are worth remembering, then wrap the Jev call in a small function that adds memories to later calls.	Beginner

Classification

Assign inputs to categories at any depth.

Cookbook	What it does	Level
Hierarchical classification	Classifies documents through deep patent, retail product, biomedical, and source-code hierarchies using parallel beam search over TypeSafe Choice probabilities.	Intermediate
Classification using confidence	Classify SEC annual reports into 75 industry groups with one Choice each, then read the answer’s own confidence to decide whether to report that group or the broader division above it.	Beginner
Twitch chat filter	Hide insults, spam, and unasked-for gameplay tips in Twitch chat, blur spoilers, and highlight questions. Each message gets one TypeSafe request containing multiple questions. Up to eight messages are processed concurrently, and the rules are defined in code.	Beginner

Extraction

Pull typed values and structure out of messy text.

Cookbook	What it does	Level
SDE cascade	Uses a 2-stage structured-data-extraction cascade (mini → verify → reasoning) to get most of the quality of a big reasoning model at a fraction of the cost.	Intermediate
Date extraction	Extracts absolute and relative dates by asking TypeSafe for the parts named in a document, then resolving and validating them in code with confidence-based review.	Beginner
Pre-parsed value extraction	Uses regexes to find candidate emails, phone numbers, and amounts, then has TypeSafe select the requested span so code can normalize a verbatim value.	Beginner
Structure recovery	Reconstructs Markdown from plain text that lost its formatting in two requests: one stitches hard-wrapped lines back together, one classifies every block (heading, list, code, callout).	Beginner

Search and matching

Find the lines, passages, or records that match a query or each other.

Cookbook	What it does	Level
Re-ranking	Builds 30-passage BM25 shortlists for 40 CLERC legal queries, then uses one TypeSafe question per query-candidate pair to raise top-1 accuracy from 5% to 18% and top-10 accuracy from 38% to 62%.	Beginner
Line-by-line search	Build semantic search for GitHub’s Terms of Service. In one request, score 218 line ids against a plain-language query with a Choice question, and use a Noul question to check whether the document contains an answer.	Beginner
Knowledge graph entity alignment	Decides which of 450 candidate pairs from two beer catalogues describe the same product using one Score question plus three companion Nouls that surface which fields disagree.	Beginner

Techniques

Get more out of any question: repeat it, batch it, or turn its answers into features.

Cookbook	What it does	Level
Self-consistency: choices	Add an uncertain outcome to moderation decisions and compare label agreement with the share of automatic actions.	Beginner
Self-consistency: nouls	Route uncertain probabilities to human review while keeping the underlying noul values visible.	Beginner
Parallel questions	Runs a 13-question regulatory briefing over the GDPR Wikipedia article, showing that batching every question into one TypeSafe call is 12.2x cheaper and 10.0x faster with no change in answers.	Beginner
Autoresearch feature discovery	Runs an autoresearch loop that proposes TypeSafe questions, converts free text into numeric features, and uses model errors to improve a supervised CatBoost regressor.	Advanced

We’re always keen to learn how people are making use of our primitives. If you’ve built something worth a cookbook, drop us a note!

​LLMs and agents

​Classification

​Extraction

​Search and matching

​Techniques

LLMs and agents

Classification

Extraction

Search and matching

Techniques