Hybrid retrieval & agentic RAG — sparse/structural, web agents & multimodal grow
Key Questions
What is Agentic RAG?
Agentic RAG supercharges retrieval with MCP, ReAct, Zep KG for memory and tools. It goes beyond basic RAG using AI agents. Videos explain integrations.
What is Zep MCP integration?
Zep MCP with Claude Code via Composio enables graph/memory ops in Agentic RAG. It provides structured retrieval enhancements. Tutorials guide setup.
What are PageIndex and AgentLink?
PageIndex offers vectorless RAG without vectorDB or chunking. AgentLink demos Gemini browsing internet remotely, solving CAPTCHAs zero-cost. They advance hybrid retrieval.
What parsing tools like MinerU and DataSieve?
MinerU2.5-Pro pushes data-centric document parsing limits at scale. DataSieve aids structural parsing for multimodal RAG. They support web agents.
Why do most LLM KBs fail?
LLM Knowledge Bases fail due to poor architecture in external stores read at query time. Types include sparse/structural hybrids. Agentic RAG with GraphRAG addresses this.
Agentic RAG (MCP+ReAct+Zep KG); PageIndex; AgentLink; GraphRAG hybrids; MinerU/DataSieve parsing; ReAct events; LLM KB failures; web rendering.