Agentic RAG / Agent Memory / Agent Frameworks
LlamaIndex
Data and retrieval framework for LLM and agent apps.
Best when the agent's value depends on ingestion, indexing, retrieval, and structured access to private knowledge.
Use LlamaIndex when retrieval quality is the core problem, not just an implementation detail behind an agent framework.
Best for
- RAG
- document workflows
- data connectors
Not ideal for
- pure tool orchestration without retrieval
- teams that only need a vector database client
Core concepts
documentsindexesretrieversquery engines
Minimal implementation shape
Ingest documents, build a retriever, inspect retrieved chunks, then let the agent decide whether another retrieval pass is needed.