Definition
An advanced form of RAG where an AI agent dynamically orchestrates the RAG pipeline, making decisions on how to retrieve, process, and generate information based on the query and available tools.
Why it matters (in Poovi’s context)
Allows for more sophisticated and adaptable RAG systems that can handle complex queries, self-correct errors, and optimize the retrieval process.
Key properties or components
- Dynamic Pipeline Selection
- Query Translation
- Self-Correction
- Tool Utilization
- Reasoning Capabilities
Contradictions or debates
While it significantly improves accuracy, it comes at the cost of increased latency compared to simpler RAG methods.