Multi-Agent Research Synthesis Workflow

Content

{
  "steps": [
    {
      "tool": "llm",
      "order": 1,
      "action": "Decompose research question into 4–6 sub-questions via orchestrator agent",
      "output": "sub_question_list"
    },
    {
      "tool": "agent_framework",
      "order": 2,
      "action": "Fan out: spawn one researcher agent per sub-question in parallel",
      "output": "raw_findings_per_subq"
    },
    {
      "tool": "tavily_api",
      "order": 3,
      "action": "Each researcher searches web + arxiv, extracts relevant passages, cites sources",
      "output": "cited_passages"
    },
    {
      "tool": "llm",
      "order": 4,
      "action": "Critic agent identifies contradictions and confidence gaps across findings",
      "output": "contradiction_report"
    },
    {
      "tool": "llm",
      "order": 5,
      "action": "Synthesizer agent merges findings, resolves contradictions, writes structured report",
      "output": "synthesis_report"
    },
    {
      "tool": "tavily_api",
      "order": 6,
      "action": "Fact-check agent verifies 5 most critical claims against primary sources",
      "output": "verified_report"
    }
  ],
  "tools_required": [
    "llm",
    "tavily_api",
    "agent_framework"
  ],
  "expected_output": "Structured research report with citations, confidence levels, and identified open questions",
  "trigger_condition": "Research question submitted requiring synthesis across 10+ sources"
}