Summarize API

Produces an extractive summary by scoring sentences on keyword salience and returning the top N in original order. Extractive by design — it never fabricates text, so output is always grounded in the input.

This capability is in beta. The contract is stable, but behavior may still be refined.

What it does

Extractive summary — pick the most salient sentences.

When to use it

Reach for Summarize when you need to answer: “Can you shorten this text to the key sentences?” It sits in the Understand stage of the catalog — compare records and explain results.

Endpoint & cost

POST /api/v1/summarize

Cost: Per credit. Zaps draw down your shared monthly allowance; requests that fail validation are not charged.

Authentication

Send your key as a bearer token: Authorization: Bearer zap_live_…. See Authentication for key management and the alternate header.

Example request

curl https://zapinner.com/api/v1/summarize \
  -H "Authorization: Bearer zap_live_••••••" \
  -H "Content-Type: application/json" \
  -d '{"text":"Data quality matters. Bad data costs money every day. Our tool finds issues fast. It integrates with pipelines easily.","options":{"sentences":2}}'

Every response is structured JSON and includes a request_id. Full field-level request and response schemas for this endpoint live in the OpenAPI spec and the API reference.

Errors & limits

Errors use one consistent shape ( error.code, error.message, request_id). See Errors and Rate limits.

SDK

Summarize has typed methods in both the JavaScript and Python SDKs. Both are implemented and package-ready but not yet published to npm / PyPI — until they ship, vendor the package source or install from a git ref, or call the endpoint over HTTPS with the example above.

Try it

Select Summarize in the playground to see the request, or run it live with your API key.

Match — Decide whether two records refer to the same real-world entity.
Compare — Surface the differences between two records or datasets.
Explain — Get a structured, factor-by-factor explanation of a result.
Extract — Pull caller-declared structured fields out of unstructured text.
Web Extract — Fetch a public web page and extract structured fields from it.
Fuzzy Match — Per-field string-similarity comparison of two records.
Schema Detect — Infer a typed schema object ready to feed into map or validate.
Classify — Single-label classification into caller-declared categories.
Tag — Multi-label tagging: apply every tag whose keywords match.
Profile — Get a per-field profile of any record set in one call.
Statistics — Descriptive statistics for every numeric field.
Missing — Find and quantify missing values per field.
Distributions — Histogram, skew, and outliers for numeric fields.
Quality Score — A composite 0-100 data-quality score with components.
Correlations — Pairwise Pearson correlation across numeric fields.
Trends — Ordinary-least-squares trend, growth, and volatility.
Change Points — Detect level shifts in a numeric series.
Patterns — Discover character-class format patterns per field.
Sentiment — Lexicon-based sentiment with negation handling.
Language — Detect the dominant language of a text.
Entities — Extract emails, phones, money, dates, and proper nouns.
Keywords — Rank the most salient terms in a text.
Text Similarity — Jaccard and cosine similarity between two texts.
Topics — Surface recurring terms across a document set.
HTML Extract — Extract title, headings, links, and text from HTML.
Forecast — Project a numeric series forward with a linear model.