Label: enhancement, agentic, research
Description: The current deep_research mode performs multi-round searching but doesn't explicitly "fact-check" or "verify" contradictory information. We want to move towards a more agentic pipeline where Round 2 and Round 3 are steered by specific contradictions or gaps identified in Round 1.
Requirements:
Update the "Gap Analysis" logic to look for conflicting data points (e.g., different release dates or pricing).
Implement a "Self-Correction" step: The agent should generate a query specifically to resolve a conflict found in previous rounds.
Enhance the report structure to include a "Conflicting Viewpoints" or "Verified Metrics" section.