SDK API

Subcategory

evaluate( )

Typed evaluation with experimental_evaluate and Jev

17 patternsSDK API
Jev Composite ScorecardScore a pull request description with three atomic Jev score questions, then combine the answers with weights in code.Sep 2026
Jev Output VerifierWrite an answer, then show it only if Jev is at least 90% sure it actually answers the question. If not, rewrite once.Sep 2026
Jev Loop BreakerLook up facts one step at a time. Stop when Jev is at least 90% sure the notes answer the task, or after 3 lookups.Sep 2026
Jev Sub-agent DelegationPick one specialist with a Jev choice, then run only that agent when confidence and selected probability both clear their floors.Sep 2026
Jev Best-of-N JudgeThree prepared replies sit on the table. A Jev choice lifts the winner only when its probability clears 0.70.Sep 2026
Jev Infinite FAQA FAQ whose last row is a live ask field. Jev opens a canned answer when it matches, or a short grounded answer becomes another row.Sep 2026
Jev Tool PermissionsPrune an agent's tool list before the model sees it. Allow and deny rules run first; Jev asks one relevance boolean per remaining tool in a single call.Sep 2026
Jev Email TagsTag a closed inbox with one Jev boolean per tag. A stamp lands only when its probability clears a floor.Sep 2026
Jev Table QueryFilter or sort a closed table with one Jev question per row. A column name never leaves the app.Sep 2026
Jev MemoryLong-term agent memory with three Jev gates. Keep durable facts, recall what fits a token budget, and evict what went stale. Nothing is rewritten.Sep 2026
Jev + Web ContextExa finds live evidence, Jev answers the same question with and without it, and leave-one-out weights show what each source changed.Sep 2026
Jev EvalsRun a rubric eval suite with jev-evals: one Jev call per case answers every rubric, and a baseline diff catches fractional regressions.Sep 2026
Jev Semantic ClusteringDiscover themes in comments with no labels up front. Embeddings pick likely pairs, Jev answers one boolean per pair, and code does the grouping.Oct 2026
Jev Entity ResolutionResolve messy card statement descriptors to the merchants behind them. Code blocks and vetoes, Jev answers one boolean per pair, and unsure matches wait for review.Oct 2026
Jev Action GateOne command at a time. It passes the gate only when a Jev boolean clears 0.90.
Jev Confidence RouterFour trays are on the table. A message flies to a team only when confidence and the selected probability both clear their floors.
Jev Search FilterFilter a closed catalog with one Jev boolean per item. Matches rise out of the pile when their probability clears a floor.

Explore more

Related patterns

AI SDK Gemini Flash TextGenerate text and analyze market research using Google's Gemini 2.5 Flash. Includes interactive charts and data visualization.Oct 2025
Generate TextGenerate text from prompts using AI. Simple interface with character counting and response rendering.
OpenAI Structured OutputGenerate structured output from AI prompts using OpenAI. Create typed data with generateText and Output.object().
Claude Structured OutputGenerate structured output from AI prompts using Anthropic Claude. Create typed data with generateText and Output.object().
Gemini Structured OutputGenerate structured output from AI prompts using Google Gemini. Create typed data with generateText and Output.object().
Stream TextStream text responses from AI prompts in real-time. Includes character counting and live response rendering with useCompletion.

More in SDK API

generateText( )

Text and structured data generation with Output.object()

5 patternsView →

streamText( )

Streaming text and structured data with Output.object()

2 patternsView →

generateImage( )

Image generation examples

4 patternsView →

generateSpeech( )

Speech synthesis examples

2 patternsView →

transcribe( )

Audio transcription examples

2 patternsView →

tool( )

AI SDK tool API examples and patterns

13 patternsView →

You’ve reached the end of evaluate( ).