Deep Research
AI-generatedSummary
Start a deep research task with configurable depth and options.
Inputs
- researchQuery (required) — The research task description
- researchMode — Research depth and duration
- searchType — Type of sources to search
- outputFormats — Output format(s) for the research report
- strategy — Natural language instructions for the research strategy
- includedSources — Comma-separated list of sources to include
- urls — Comma-separated URLs to analyze (max 10)
- codeExecution — Whether to allow code execution during research
- waitForCompletion — Whether to wait for the research to complete or return immediately with task ID
Output shape
a JSON object representing the deep research task outcome
When waitForCompletion is true the response includes deepresearch_id, status, output, output_type, sources, usage, pdf_url, completed_at; when false only deepresearch_id and status are returned; on error an object with an
errorfield is returned.
Examples
Example 1: Run a heavy research on AI trends and wait for full report
researchQuery: 'Explain the impact of generative AI on enterprise software', researchMode: 'heavy', deepResearchOptions: {searchType: 'web', outputFormats: ['pdf'], strategy: 'Prioritize peer‑reviewed journals', includedSources: 'arxiv, nature.com', urls: '', codeExecution: true, waitForCompletion: true}
Example 2: Kick off a fast research task without blocking workflow
researchQuery: 'List top 5 cloud providers', researchMode: 'fast', deepResearchOptions: {searchType: 'web', outputFormats: ['markdown'], waitForCompletion: false}
Example 3: Research limited to specific journals with a custom strategy
researchQuery: 'Compare renewable energy policies', researchMode: 'standard', deepResearchOptions: {searchType: 'proprietary', outputFormats: ['markdown'], strategy: 'Focus on governmental reports', includedSources: 'eia.gov, iea.org', urls: '', codeExecution: false, waitForCompletion: true}