summarize.do
Intelligent Summaries at the Edge
Process documents, conversations, and data streams with AI-powered summarization. Automatic chunking, semantic compression, and multi-format support with edge caching.
import { Summarize } from 'summarize.do' // Summarize documents at the edgeconst summary = await Summarize.document({ content: longArticle, length: 'brief', format: 'bullets'}) // Real-time conversation summarieson.Chat.Message.sent($ => { if ($.conversation.length % 10 === 0) { const recap = Summarize.conversation({ messages: $.conversation }) $.Chat.send.Summary({ recap, threadId: $.threadId }) }})
One of 48 primitives
summarize.do sits in Agents: reason, research and act. Every .do shares one grammar, Business-as-Code, and one SDK.
Data
Records, vocabulary and search
Insight
Measure, test and trace
Frequently asked questions
Summarize.do runs at the edge with automatic caching (90% cost reduction), intelligent chunking for large documents, and multi-model routing for optimal quality and cost. Handles rate limits, retries, and fallbacks automatically. Pre-optimized prompts for different content types (articles, code, conversations, data) deliver better results than generic API calls. Edge caching means identical content returns summaries in <10ms.
Automatic intelligent chunking with semantic boundary detection. Documents are split at natural breakpoints (paragraphs, sections) while preserving context. Map-reduce summarization aggregates chunk summaries into coherent final output. Handles documents up to 1M tokens with maintained coherence. Hierarchical summarization for extremely large corpora (millions of documents).
All major document formats: PDF, DOCX, HTML, Markdown, plain text, and structured data (JSON, CSV, XML). Code summarization for 50+ programming languages. Audio transcription + summarization for meetings and podcasts. Supports 100+ languages with automatic language detection. Preserves formatting, links, and metadata through summarization process.
Yes. Configure output length (brief/medium/detailed), format (bullets/paragraphs/executive), and focus areas (technical details, business implications, action items). Custom prompts and examples for domain-specific summarization. Fine-tune models on your content for improved quality. A/B test different approaches and track which summaries users engage with most.
Multi-model ensemble with automatic quality scoring and confidence intervals. Low-confidence summaries are automatically regenerated with different approaches. Citation tracking ensures summary claims link to source material. Human feedback loop improves quality over time. Battle-tested across billions of documents with 95%+ user satisfaction scores. Automatic factuality checking flags potential hallucinations.