← Về thư mục
📄 / / projects / personal / Vedic-astro-mar / docs / archive / mas_legacy / research_lab_mas_guide.md

Research Lab: Multi-Agent System (MAS) & Learning Guide

This document defines a structured approach for a Researcher & Learner role using a 5-agent MAS. It is optimized for extracting actionable lessons and steps from external sources (YouTube, forums, documents).

1. Role: Researcher & Learner (MAS)

Core Limitations


2. The 5-Agent Architecture

Agent Task Input Output
Fetcher Scrape & Ingest Source URL/Playlist /research_lab/raw/*.txt
Synthesizer Group & Summarize Raw Text files /research_lab/structured/*.md
Distiller Extract Lessons Structured Theme files /research_lab/distilled/*.md
Archivist Compile Final Distilled Lesson files /research_lab/final/lessons_learned.md
Reviewer Verify & Validate Final Report Verified lessons ready for integration.

3. The Workflow (MAS)

graph TD
    A[Source Content] -->|Fetcher| B[Raw Data]
    B -->|Synthesizer| C[Structured Themes]
    C -->|Distiller| D[Distilled Lessons]
    D -->|Archivist| E[Compilation]
    E -->|Reviewer| F[Verified Output]
    F -->|Ready to Move| G[Main Workspace]

Phase 1: Fetching (Data Ingestion)

Phase 2: Synthesizing (Structuring)

Phase 3: Distilling (Extraction)

Phase 4: Archiving (Final Compilation)

Phase 5: Reviewing (Verification)


4. Research Workspace Structure

Ensure your /research_lab folder has the following children: - raw/: Raw scraped data. - structured/: Processed themes and summaries. - distilled/: Lesson highlights and actionable steps. - final/: Ready-for-integration reports. - archive/: Old research for historical reference.