Deep Research

​

Overview

• Autonomous research - Conducts multi-step investigations without human intervention
• Source synthesis - Combines information from multiple sources into coherent insights
• Structured output - Returns organized, formatted research reports
• Citation tracking - Maintains source attribution throughout the research process
• Depth control - Adjust research thoroughness based on your needs

• Market research - Analyze industries, competitors, and trends
• Academic research - Literature reviews and topic exploration
• Due diligence - Company research and background checks
• Content creation - Research for articles, reports, and presentations
• Decision support - Gather comprehensive information for strategic decisions
• Competitive analysis - Deep dives into competitors and market position

​

How Deep Research Works

​

Research Process

1. Query Understanding - Analyzes your research instructions to identify key topics and questions
2. Research Planning - Creates a research strategy and identifies information needs
3. Multi-Step Search - Performs multiple targeted searches to gather comprehensive information
4. Source Evaluation - Assesses source credibility and relevance
5. Information Synthesis - Combines insights from multiple sources into coherent findings
6. Report Generation - Produces structured output with citations
7. Quality Check - Verifies completeness and accuracy

​

Research Depths

• 10-15 sources consulted
• High-level overview
• Key facts and basic insights
• Best for: Quick briefings, initial exploration

• 20-30 sources consulted
• Balanced depth and breadth
• Comprehensive key findings
• Best for: Most research tasks, reports, decision support

• 40+ sources consulted
• In-depth analysis
• Detailed insights and connections
• Best for: Academic research, strategic planning, comprehensive reports

​

Examples

from incredible_python import Incredible

client = Incredible(api_key="YOUR_API_KEY")

response = client.deep_research(
    instructions="Research the history of Python programming language"
)

print(response.output)

if response.citations:
    print(f"\nCitations: {len(response.citations)}")
    for citation in response.citations:
        print(f"  - {citation.title}: {citation.url}")

​

Request Parameters

​

instructions (required)

• ✅ “Research the current state of quantum computing, focusing on commercial applications and major players. Include market size estimates and recent breakthroughs.”
• ✅ “Analyze the impact of remote work on tech company productivity between 2020-2024. Compare different company approaches and outcomes.”
• ✅ “Investigate the environmental impact of electric vehicles, considering battery production, charging infrastructure, and lifecycle emissions. Compare with traditional vehicles.”

• ❌ “Tell me about AI” (too broad, no specific angle)
• ❌ “What is blockchain?” (too simple, better suited for Answer API)
• ❌ “Research everything about climate change” (too vast, needs focus)

• Be specific - Define scope, time period, geographic region
• State your goal - “for a presentation,” “to make a decision,” “for an article”
• Specify format - “executive summary,” “pros and cons list,” “timeline”
• Include constraints - “focusing on US market,” “from 2020-2024,” “excluding speculative sources”

​

depth (optional)

• Fast turnaround for basic questions
• 10-15 sources
• High-level overview and key facts
• Best for: Briefings, quick fact-checking, initial exploration

• Balanced approach for most use cases
• 20-30 sources
• Comprehensive coverage of main topics
• Best for: Business reports, decision support, content research

• Deep investigation with extensive source analysis
• 40+ sources
• Detailed insights, connections, and nuances
• Best for: Academic research, strategic planning, comprehensive analysis

• Quick = Lower cost, faster results, less comprehensive
• Thorough = Higher cost, slower results, more comprehensive
• Standard = Balanced for most needs

​

max_sources (optional)

• 5-10 - Quick fact checks
• 20-30 - Standard research (recommended)
• 50-100 - Exhaustive analysis

​

response_format (optional)

{
  "type": "object",
  "properties": {
    "executive_summary": {"type": "string"},
    "key_findings": {"type": "array", "items": {"type": "string"}},
    "recommendations": {"type": "array", "items": {"type": "string"}},
    "risks": {"type": "array", "items": {"type": "string"}}
  }
}

{
  "type": "object",
  "properties": {
    "companies": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "name": {"type": "string"},
          "strengths": {"type": "array"},
          "weaknesses": {"type": "array"},
          "market_share": {"type": "string"}
        }
      }
    }
  }
}

{
  "type": "object",
  "properties": {
    "events": {
      "type": "array",
      "items": {
        "type": "object",
        "properties": {
          "date": {"type": "string"},
          "event": {"type": "string"},
          "impact": {"type": "string"}
        }
      }
    }
  }
}

• Feeding research directly into applications
• Creating consistent report formats
• Building dashboards and visualizations
• Automating workflows based on research findings

​

Best Practices

• Start with a clear question - Well-defined research goals produce better results
• Scope appropriately - Narrow topics get better depth, broad topics need “standard” or “thorough”
• Specify time periods - “in 2024,” “since 2020,” “historical perspective”
• Define success criteria - What information would make this research complete?

• Use quick for time-sensitive decisions or initial exploration
• Use standard as the default for most business and content needs
• Use thorough for high-stakes decisions or academic research
• Consider cost and time constraints when choosing depth

• Design schemas that match your application’s needs
• Include optional fields for flexibility
• Use enums for constrained values
• Test schemas with sample research first

• Review citations to verify source quality
• Cross-check key facts if stakes are high
• Consider running multiple research sessions with different angles
• Combine Deep Research with human expertise for critical decisions

• Cache research results for repeated queries
• Use appropriate depth levels to balance cost and quality
• Consider batching related research queries
• Monitor API usage to optimize costs

• Automated reporting - Schedule regular research on tracked topics
• Decision support - Research before major business decisions
• Content pipeline - Research for article/content creation
• Competitive intelligence - Regular deep dives on competitors
• Market monitoring - Track industry trends and developments

​

Response

{
  "success": true,
  "report": "Comprehensive research report...",
  "key_findings": [
    "Finding 1",
    "Finding 2"
  ],
  "sources": [
    {
      "title": "Source Title",
      "url": "https://example.com",
      "relevance": "high"
    }
  ]
}