Automated Extraction of Personal Data via Self-Hosted LLM

This workflow enables automatic extraction of personal information from user messages using a self-hosted Language Model (LLM) called Mistral NeMo. When a chat message is received, it triggers a sequence of nodes that process, analyze, and extract structured data from the message. The process starts with a webhook trigger that activates the workflow upon message receipt. The message is then sent to the Ollama Chat Model, which runs the Mistral NeMo model configured with specific parameters for optimal performance. The response from the LLM is parsed through an ‘Auto-fixing Output Parser’ that ensures the output adheres to a predefined JSON schema. This schema captures details such as the user’s name, surname, contact info, method of communication, and communication timestamp. If the initial response is invalid, the auto-fixer prompts the model again to correct the output. Once validated, the structured data is extracted and available for further use, such as updating a database or CRM system. This workflow is particularly useful for automating customer data collection, lead management, or enhancing chatbot interactions with accurate user details.