Building a RAG agent
In previous tutorials, we built a pipeline that embeds the chunks of text similar to user's query to a system message, which allows the chatbot to access the external knowledge base. However, in practice, this approach may be too naive, as it:
All of these limitations can be addressed by building a more sophisticated pipeline logic, that might have a routing and query-rewriting mechanisms. However, a viable alternative is to use an Agent which can inherently perform all of these tasks.
The fundamental concept of agents involves using a language model to determine a sequence of actions (including the usage of external tools) and their order. One possible action could be retrieving data from an external knowledge base in response to a user's query. In this tutorial, we will develop a simple Agent that accesses multiple data sources and invokes data retrieval when needed.
As an example of external knowledge bases, we will use three webpages containing release announcement posts about recently released generative models:
