🎬 Codelab: Construyendo un Sistema Multi-Agente Multimodal para Análisis de Evidencia
Video Details
Channel: DevHack URL: https://www.youtube.com/watch?v=4l0bYXTtDIs Relevance: ⭐⭐⭐⭐
Summary
Hands-on codelab building a multi-agent system that analyses evidence from multiple sources (text, images, structured data). The architecture: collector agents gather evidence from different sources, analyst agents process each modality, and a synthesiser agent produces a unified report. Implemented with LangGraph.
PUMA Relevance
The multi-agent evidence analysis architecture closely mirrors PUMA’s Smart PMO vision (Stage 5): collector agents gather issue data from Jira + GitHub, specialist agents analyse priority/effort/risk, and the Manager Agent synthesises into a sprint recommendation. The multimodal processing (text + structured data) parallels PUMA’s issue text + metadata approach.