Agentic RAG

[dnielsen]

Today AllyCat performs basic RAG. It retrieves docs via vector search and passes them into an LLM prompt for answering questions. What else could AllyCat do to make it Agentic RAG? For example, is there a need for it to enable the system to act intelligently and autonomously on tasks? In other words, is there anything it should to that would require it to plan, reason, use tools, need memory, or perform multi-step workflows?