A report generation method based on multi-agent cooperation, medium and device

Abstract

This invention relates to the field of report generation technology, and in particular to a report generation method, medium, and device based on multi-agent collaboration. By parsing user intent, the overall research is broken down into clearly hierarchical and actionable sub-tasks. Combining the research outline, agent capability matrix, and pre-set knowledge graph, agents are scheduled to collaboratively process these sub-tasks, achieving precise matching between agents and sub-tasks. This multi-agent division of labor improves processing efficiency while simultaneously completing key information collection and multi-dimensional specialized analysis, enhancing the depth of data analysis. By integrating the results of specialized analysis into the pre-set knowledge graph, knowledge anomalies are accurately identified and feedback reports are generated. Through closed-loop, targeted iterative optimization, a complete and consistent target knowledge fusion result is formed. This result, combined with user intent, generates a target research report, ensuring accurate, authoritative, and relevant content. This achieves a value loop from research decomposition to results implementation, guaranteeing the report's professionalism, credibility, and logical coherence.

Images