PerigonInstall

Summary

Generate a concise AI-powered summary synthesizing news articles filtered by a search query and optional parameters such as categories, countries, sources, and date range.

Inputs

• Query (required) — Primary search query string supporting Boolean operators to filter relevant articles.
• Prompt — Instructions guiding how the summary should be composed, with a default that requests a casual, vivid summary in one paragraph no longer than 125 words.
• Category — Comma-separated list of categories to filter articles (e.g., Tech, Business).
• Country — Comma-separated list of country codes to filter articles by origin (e.g., us, gb, ca).
• From Date — ISO date string to filter articles published after this date.
• Max Article Count — Maximum number of articles to include in the summary generation (1-100, default 10).
• Max Tokens — Maximum number of tokens to generate in the summary output, defaulting to 2048.
• Method — Method for selecting articles to summarize: 'ARTICLES' includes all matching articles; 'CLUSTERS' includes one article per cluster.
• Model — The underlying large language model used to generate the summary (default 'gpt-4.1').
• Returned Article Count — Maximum number of articles to return in the API response (1-100, default 10).
• Source — Comma-separated list of source domains to filter articles (e.g., cnn.com, bbc.com).
• Summarize Fields — Fields of each article to include in the summary context; options are Title, Content, and Summary (default all three).

Output shape

a single JSON object containing the AI-generated summary text along with metadata including articles considered in the summary

The response includes the generated summary as well as the articles used to generate it, filtered according to the input parameters. Pagination is controlled by 'Returned Article Count'. Errors are returned via standard API error envelopes.

Examples

Example 1: Summarize recent renewable energy news

Provide a query 'renewable energy', limit to US sources, set maxArticleCount to 20, keep default prompt and model to generate concise insights.