Cookbook
Cookbook: Agent skill suggestion
Picks which agent skill to load for a given task by scoring candidate skills against the current state.
Official cookbooks and patterns
Worked examples published by the model vendor rather than by the community: the documented patterns, the walkthroughs for each primitive, and the code the SDK reference is built around. These are the canonical references our recipes are derived from, and they are the first place to check when something in a community project looks wrong.
25 cookbooks
Showing 25 of 25
All entries
- Cookbook
Cookbook: Autoresearch feature discovery
Generates and tests candidate features over a dataset by scoring each one with Jev.
- Cookbook
Cookbook: Batching parallel questions
Shows how to send many questions about the same state in a single request instead of issuing one call per question.
ChoiceScoreNoul - Cookbook
Cookbook: Classification using confidence
Uses the confidence value returned with Choice and Score answers to decide when to act automatically and when to escalate.
- Cookbook
Cookbook: Classifying RAG passages
Filters retrieved passages for relevance before they reach the generation step of a RAG pipeline.
- Cookbook
Cookbook: Date extraction
Extracts dates from text using pre-parsing plus Choice questions, working around the documented weakness of jev-1.13 at date arithmetic.
ChoiceStructured extraction - Cookbook
Cookbook: Double-checking citations
Verifies that each citation in generated text is actually supported by the cited source.
- Cookbook
Cookbook: Function calling from natural language
Maps a natural-language request onto a function and its arguments using Choice questions, as a cheaper substitute for LLM tool calling.
- Cookbook
Cookbook: Guardrails for LLMs
Screens LLM inputs and outputs with typed checks so unsafe or off-policy content is caught before it is used.
- Cookbook
Cookbook: Hierarchical classification
Classifies into a deep taxonomy by walking the tree one Choice question at a time instead of flattening every leaf into one list.
- Cookbook
Cookbook: Knowledge graph entity alignment
Decides whether two records refer to the same entity when merging knowledge graphs or databases.
- Cookbook
Cookbook: Line-by-line semantic search (semantic find)
Runs a semantic grep over a file by asking a noul question per line, turning meaning-based lookup into a batch of typed decisions.
- Cookbook
Cookbook: Pre-parsed value extraction
Uses code to enumerate candidate values first, then asks Jev to choose among them, avoiding the numeric weaknesses of the model.
- Cookbook
Cookbook: Re-ranking with semantic search
Uses Jev scores to re-rank candidate results returned by a first-stage retriever.
- Cookbook
Cookbook: SDE cascade
Chains cheap Jev decisions in front of expensive steps so only the uncertain cases escalate.
- Cookbook
Cookbook: Self-consistency with Choice questions
Applies the same repeated-sampling self-consistency check to Choice questions so disagreement between samples becomes a signal to escalate.
- Cookbook
Cookbook: Self-consistency with Noul questions
Samples a noul question repeatedly and routes uncertain probabilities to human review while keeping the underlying values visible. Worked example uses insurance claim evaluation.
- Cookbook
Cookbook: Structure recovery from plain text (autoformat)
Recovers structured fields from unstructured plain text by asking typed questions rather than prompting an LLM for JSON.
- Cookbook
Demo: Smart home assistant
Interactive demo that evaluates smart home requests with speculative questions and falls back to an LLM when Jev is uncertain.
- Cookbook
Jev 1.13 jaggedness (known limitations)
Official page listing what jev-1.13 does badly: literal interpretation, unreliable counting, weak numerics and date ordering, multi-hop indirection, distraction from irrelevant state, and no built-in prompt-injection defence.
- Cookbook
Pattern: Composite scoring
Combines several independent score dimensions into a single ranking or decision number.
- Cookbook
Pattern: Confidence-gated routing
Uses the returned confidence value as a routing input: act on high confidence, confirm on medium, escalate to a human or an LLM on low.
- Cookbook
Pattern: Intent routing
Classifies an incoming request into an intent with a Choice question and dispatches it to the matching handler.
- Cookbook
Pattern: Speculative fan-out
Sends many questions in one call, including speculative ones, and lets application code decide afterwards which answers matter.
ChoiceScoreNoul - Cookbook
TypeSafe HTTP API reference
Reference for POST https://api.typesafe.ai/v1/systemone, covering the state/model/questions request body, the answer shapes, and error codes 401, 422, 429 and 529.
ChoiceScoreNoul
Nothing matches those filters
Clear a filter to widen the search, or send us the thing you were looking for.