How asinghcsu/agenticrag-survey works
Project type and purpose
• This is an open-source survey and documentation repository titled “Agentic Retrieval-Augmented Generation: A Survey On Agentic RAG.”
• Its purpose is to present a comprehensive taxonomy, patterns, workflows, comparative analysis, applications, and references for Agentic RAG systems, supported by illustrations in the assets directory.
File structure
• Top-level:
• README.md – the single entry point containing all documentation and links.
• assets/test – a placeholder or test assets folder.
• No source code or runtime components; it’s purely documentation and visual materials.
README for additional insights
• Sections correspond to major conceptual components:
• Agentic Patterns (reflection, planning, tool use, multi-agent).
• Agentic Workflow Patterns (prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer).
• Taxonomy of system architectures (single-agent, multi-agent, hierarchical, corrective, adaptive, graph-based, document workflows).
• Comparative analysis, applications, challenges, and future directions.
• Each section is illustrated with an image file under assets/.
• Links to external notebooks, papers, and tutorials act as “external services.”
Elements
• Clear labels: “README.md,” “assets/agentic_planning.png,” “ArXiv Paper 2501.09136,” etc.
• README → assets (include images)
• README → external URLs (links to papers, notebooks)
• assets → renderer (visual display)
• Blue rectangles for file-system components (README.md, assets/).
• Yellow parallelograms for rendered output (GitHub page).
Detail
• Every section of the README as a distinct node (e.g., “Agentic Patterns,” “Taxonomy,” “Applications”).
• For each node, show the associated image file from assets/.