Advanced AI-Powered Document Parsing & Text Extraction with LlamaParse

This n8n workflow demonstrates an advanced method for parsing documents and extracting structured text using the LlamaParse API, triggered by new emails containing attachments or a manual webhook. It intelligently processes documents, classifies their content, and then either saves the extracted data to Google Sheets or sends it via Telegram, based on the classification.

What it does

This workflow automates the following steps:

1. Triggers on New Emails or Webhook: It can be initiated either when a new email with an attachment is received in Gmail, or manually via a webhook.
2. Downloads Attachments: If triggered by Gmail, it downloads the attached files from the email.
3. Parses Documents with LlamaParse: It sends the document attachments to the LlamaParse API for advanced parsing and text extraction.
4. Classifies Extracted Text: The extracted text is then fed into a Text Classifier (powered by LangChain and OpenAI) to categorize the document's content.
5. Conditional Routing:
   • If the document is classified as "Invoice", it extracts specific fields (e.g., invoice number, amount, date) and saves them to a Google Sheet.
   • If the document is classified as "Receipt", it extracts relevant details and sends a summary message to a Telegram chat.
   • For any other classification, it performs a "No Operation" (effectively ignoring it or allowing for future expansion).
6. Merges Paths: After conditional processing, the workflow merges the different execution paths.
7. Saves to Google Drive (Optional): If the document was processed, it saves the original attachment to Google Drive.

Prerequisites/Requirements

To use this workflow, you will need:

• n8n Instance: A running instance of n8n.
• Gmail Account: Configured with n8n credentials for the Gmail Trigger node.
• Google Drive Account: Configured with n8n credentials for the Google Drive node.
• Google Sheets Account: Configured with n8n credentials for the Google Sheets node.
• LlamaParse API Key: An API key for the LlamaParse service. This will be used in the HTTP Request node.
• OpenAI API Key: An API key for OpenAI, used by the "OpenAI Chat Model" node within the Text Classifier.
• Telegram Bot Token and Chat ID: Configured with n8n credentials for the Telegram node.

Setup/Usage

1. Import the workflow: Download the JSON provided and import it into your n8n instance.
2. Configure Credentials:
   • For the Gmail Trigger, Google Sheets, and Google Drive nodes, set up your Google OAuth2 credentials.
   • For the Telegram node, set up your Telegram Bot API credentials.
   • For the HTTP Request node (which interacts with LlamaParse), you will need to add an API key as a header or query parameter, depending on LlamaParse's authentication method.
   • For the OpenAI Chat Model node, configure your OpenAI API key.
3. Customize LlamaParse HTTP Request:
   • Review the "HTTP Request" node. Ensure the URL, headers (including your LlamaParse API key), and body are correctly configured for the LlamaParse API endpoint you intend to use for document parsing.
4. Configure Text Classifier:
   • Adjust the categories and examples in the "Text Classifier" node to accurately reflect the types of documents you expect and the classifications you want to make (e.g., "Invoice", "Receipt", "Other").
5. Set up Google Sheets:
   • Create a new Google Sheet with appropriate headers (e.g., "Invoice Number", "Amount", "Date") if you intend to save "Invoice" data. Update the "Google Sheets" node to point to your spreadsheet and worksheet.
6. Configure Telegram:
   • Ensure the "Telegram" node is configured with the correct Chat ID to send messages to.
7. Activate the workflow: Once all credentials and configurations are set, activate the workflow.

The workflow is now ready to automatically process documents from incoming emails or manual triggers, extract information, classify it, and route it to the appropriate destination.