Turn incoming emails into fast, accurate replies powered by your own documents. The flow spots general company info requests, searches a private knowledge base built from Google Drive files, writes a short answer, reviews the message, and sends it back to the sender. It suits small support teams that want quick answers without manual copy and paste.

Here is how it works. New emails arrive over IMAP and get converted to clean text. The message is summarized by a language model and classified to check if it is an info request. If it matches, the reply engine uses a Qdrant vector database as a tool to pull facts from your documents. A short reply is drafted, reviewed for tone and structure, formatted as HTML, and sent by SMTP with the original subject line. A separate setup path builds the knowledge base by pulling files from a Google Drive folder, splitting text, creating embeddings with OpenAI, and storing them in Qdrant. Collection create and refresh calls run through HTTP requests.

To run it, you need Google Drive access, a live Qdrant endpoint, and API keys for OpenAI and OpenRouter. You also need IMAP and SMTP credentials for your support inbox. Expect faster first replies, more consistent answers, and fewer tickets that wait for manual triage. It works well for company info emails, partnership requests, and press questions.

• IMAP email intake reads new messages and preserves sender and subject
• Markdown conversion cleans HTML email into readable text for the model
• LLM summarization condenses each email to about 100 words for faster processing
• Classifier filters only company info requests and safely ignores others
• Google Drive ingestion pulls files from a folder, loads text, and splits it into chunks
• OpenAI embeddings create vectors and store them in a Qdrant collection
• Qdrant runs in tool mode so the reply agent can retrieve the most relevant facts
• Reply agent drafts a concise answer and a reviewer model formats and polishes it in HTML
• SMTP sending replies to the original sender with a proper Re subject
• HTTP requests create and refresh the Qdrant collection for clean indexing

• Reduce manual email handling from 15 minutes to under 1 minute per message
• Automate up to 80 percent of general company info replies
• Improve answer consistency by using a single, up to date knowledge base
• Handle up to 10 times more inbound emails without hiring
• Connect email, Google Drive, and Qdrant in one seamless flow
• Lower errors by pulling facts directly from your documents

1. Import the template into n8n: Create a new workflow in n8n > Click the three dots menu > Select 'Import from File' > Choose the downloaded JSON file.
2. You'll need accounts with Google Drive, Qdrant, OpenAI, OpenRouter, IMAP Email and SMTP. See the Tools Required section above for links to create accounts with these services.
3. In the n8n credentials manager, create an IMAP Email credential for your support inbox. Double click the Email Trigger node and select your new credential.
4. Create an SMTP credential for sending replies. Open the Send Email node and choose the SMTP credential. Set from name and address to your support mailbox.
5. Create a Google Drive OAuth2 credential. Open the Get folder and Download Files nodes and select this credential. Copy the folder ID that contains your documents.
6. Create an OpenAI API Key credential from the OpenAI API page. Attach it to the Embeddings OpenAI, OpenAI, and OpenAI 4 o mini model nodes.
7. Create an OpenRouter API Key credential from the OpenRouter dashboard. Attach it to the DeepSeek model nodes used for summarization and review.
8. Set up Qdrant access. In credentials, add a Qdrant API Key or HTTP Header Auth for your endpoint. Open the Create collection and Refresh collection HTTP Request nodes and set QDRANTURL and COLLECTION. Do the same in both Qdrant Vector Store nodes.
9. Configure the ingestion path. In Get folder, set the target folder. In Token Splitter, set chunk sizes that fit your documents. In Embeddings OpenAI nodes, pick the embedding model you prefer.
10. Validate the knowledge base build. Run the manual trigger path to Create collection, Refresh collection, load files, split text, and index into Qdrant. Check your Qdrant dashboard to confirm the collection and points exist.
11. Tune the classification. Open the Email Classifier node and adjust categories if you want more types beyond Company info request. Keep fallback set to other to avoid wrong routing.
12. Test end to end. Send a sample info email to your inbox. Confirm it is summarized, classified, answered with content from Qdrant, reviewed into HTML, and returned to the sender with the correct subject and branding.
13. Troubleshoot common issues. If emails do not arrive, check IMAP connection and mailbox permissions. If Qdrant indexing fails, verify the URL, API key, and collection name. If Drive files do not load, confirm the folder ID and sharing. For model errors, check API key limits and model names.