#341 Demo: Autonomous AI Research Agent

## Overview Orchestration workflow for deep research tasks over several days. ## Workflow Steps 1. User inputs research topic 2. LLM breaks into sub-questions 3. ForEach sub-question: Search API → Scrape → LLM summarize 4. Sleep for rate limits 5. Human approval gate (optional) 6. Aggregate final report ## Tools Required - tools.http.request ✅ - tools.llm.call ✅ - tools.approval.request ✅ - tools.web.scrape ❌ (or parse HTTP response) ## DSL Operators TaskOperator, ForEachOperator, WaitOperator, WaitForEventOperator, ConditionalOperator ## Durable Advantages - Auto-retry on rate limits - Human-in-the-loop approval - Multi-day execution - Resume after restart