Chat-Based RAG Agent with Vector Search and Data Logging

This workflow exposes an AI agent through a chat interface and combines conversational input with retrieval-augmented generation (RAG). Incoming chat messages are processed by the agent using a chat model, with lightweight memory to preserve short-term conversational context.

For knowledge retrieval, the agent queries a vector store backed by Pinecone. User inputs are embedded and matched against stored documents, allowing the agent to incorporate relevant contextual information into its responses. The dataset labeled Goldsmith represents a fictional company and is used solely for testing and development purposes.

In addition to generating responses, the workflow appends selected interaction data to a Google Sheets document. This makes it possible to track queries, responses, or metadata for later analysis without affecting the core agent logic.

The overall structure keeps chat handling, retrieval, memory, and external tools clearly separated, making the workflow easy to reason about and extend.