Member-only story
MCP and A2A — MCP, An Enhancement to LLM Development
From Prompt Engineering to RAG
Large language models (LLMs) can be used out-of-the-box, meaning humans can directly interact with them through prompt engineering to solve problems. As LLMs grow more powerful — especially with the emergence of advanced reasoning models — their responses become increasingly reliable.
However, LLMs are not omniscient. While their pretraining data is periodically updated, it still has a cutoff date. When we want them to analyze personal or proprietary data — such as internal sales records — they inherently lack knowledge of our specific datasets. This is where RAG (Retrieval-Augmented Generation) comes into play.
How RAG Works
RAG is an LLM application development paradigm that enhances accuracy and relevance by combining large language models with external data sources. The process involves:
Retrieval
- A retrieval system (e.g., vector database) searches external data to find information relevant to the user’s query, based on semantic similarity.
Augmentation
- The retrieved data is fed to the LLM as additional context.
