The system starts by retrieving emails from a Gmail account using the Composio Gmail API component. These emails are then fed into a loop component that processes each message individually. For each email, the system converts the data format and sends it to an AI agent equipped with Gmail tools and detailed labeling instructions.

The AI agent analyzes each email's content using a comprehensive prompt template that defines 17 different categories including Newsletter, Inquiry, Invoice, Proposal, Action Required, and others. The agent examines the subject line and body content to determine the most appropriate label, then uses Gmail API tools to create and assign the label directly to the email.

The flow begins with trigger options that include webhooks for receiving events from automation tools, direct API calls using the API Request component to Gmail REST endpoints, or third-party connectors. The system normalizes message text with Parser components, then feeds the data through prompt templates and language models for analysis.

For more precise categorization, the flow can use Structured Output to emit typed JSON fields like category, urgency, and entities. The system routes emails by category using the Smart Router for LLM-based decisions or If-Else components for rule-based routing.

The processed results are converted back to the appropriate data format and collected, with the final output displayed through a chat interface showing the completion status of the email organization process. Authentication is handled through global variables that store API keys and OAuth tokens securely.

The flow can be extended using other Langflow nodes such as embedding models to create vector databases for email search and deduplication, additional API Request components to integrate with CRM systems or project management tools, or conditional logic nodes to create more sophisticated routing rules based on sender reputation, email content analysis, or time-based criteria.