Voice Note Transcription and Retrieval Augmented Generation Agent

Overview

This n8n workflow automates the process of transcribing voice notes and making them searchable and contextually relevant. It begins with a Webhook Trigger to receive voice note data. The Text Splitter breaks down the transcribed text into manageable chunks. Embeddings (using Cohere) convert these text chunks into vector representations, which are then stored in a Weaviate vector database using Weaviate Insert.

Subsequently, a Weaviate Query node can retrieve relevant vector data. This retrieved data is then used by a Vector Tool to provide context to a RAG Agent (Retrieval Augmented Generation), which combines large language model capabilities with retrieved information. Window Memory helps maintain conversational context. The status of the transcription process is logged to a Google Sheet via Append Sheet, and Slack Alert is used for error notifications.