​
What is Deep Research?

Deep Research refers to the ability of Agentset to perform thorough, methodical analysis of large document sets to extract insights, answer complex questions, and synthesize information that may be scattered across multiple sources.

This capability allows users to:

  • Analyze complex topics across hundreds or thousands of documents
  • Perform multi-step research queries that require synthesizing information
  • Extract insights that aren’t explicitly stated in any single document
  • Draw connections between related information from different sources

​
How Deep Research Works

Agentset’s Deep Research capability builds on our Agentic RAG foundation with additional features focused on thorough analysis:

  1. Comprehensive Retrieval: The agent intelligently retrieves multiple document sets based on sub-questions
  2. Information Synthesis: Data from multiple sources is combined to form new insights
  3. Gap Analysis: The agent identifies information gaps and proactively seeks to fill them
  4. Chain-of-Thought Research: Each research step builds upon previous findings
  5. Structured Output: Research findings can be organized into structured formats

​
Deep Research in Action

​
Enabling Deep Research

Deep Research capabilities are available with Agentset’s chat functionality. For complex research tasks, you can configure the chat to perform more thorough analysis:

const researchResponse = await agentset.chat({
  namespaceId: namespace.id,
  messages: [
    { role: "user", content: "Analyze the impact of recent regulatory changes on our European market strategy" }
  ],
  options: {
    // Enable deep research mode
    researchMode: "deep",
    // Allow longer processing time for complex research
    maxProcessingTime: 300, // seconds
    // Return structured research findings
    outputFormat: "structured",
  }
});

// Access structured research findings
console.log(researchResponse.findings);

​
Structuring Research Output

For more complex research tasks, you can request structured output formats:

const researchResponse = await agentset.chat({
  namespaceId: namespace.id,
  messages: [
    { role: "user", content: "Analyze our competitor landscape in the healthcare sector" }
  ],
  options: {
    researchMode: "deep",
    outputFormat: "structured",
    outputStructure: {
      // Define the structure of the research output
      competitors: {
        type: "array",
        items: {
          name: "string",
          marketShare: "string",
          keyProducts: "array",
          strengths: "array",
          weaknesses: "array"
        }
      },
      marketTrends: "array",
      strategicRecommendations: "array"
    }
  }
});

// Access structured competitor analysis
const competitors = researchResponse.findings.competitors;

​
Research Report Generation

For comprehensive research tasks, you can generate detailed reports:

const report = await agentset.generateReport({
  namespaceId: namespace.id,
  query: "Comprehensive analysis of our market position and growth opportunities",
  options: {
    // Generate a formal research report
    format: "research_report",
    // Include executive summary
    includeExecutiveSummary: true,
    // Include visualizations in the report
    includeVisualizations: true,
  }
});

console.log(report.url); // URL to the generated report