Step 2: Connect Gmail

Configure Gmail access to read labeled threads, fetch messages, create drafts, and remove labels.

1. Open Retrieve Labeled Threads and keep Resource set to thread with Return All enabled.
2. Credential Required: Connect your gmailOAuth2 credentials in Retrieve Labeled Threads.
3. Open Fetch Message Details and set Message ID to {{ $json.id }} with Operation set to get.
4. Credential Required: Connect your gmailOAuth2 credentials in Fetch Message Details.
5. Open Retrieve Thread Messages and set Thread ID to {{ $json.id }} with Operation set to get and Return Only Messages enabled.
6. Credential Required: Connect your gmailOAuth2 credentials in Retrieve Thread Messages.
7. Open Create Draft in Thread and confirm URL is https://www.googleapis.com/gmail/v1/users/me/drafts and Method is POST.
8. Credential Required: Connect your gmailOAuth2 credentials in Create Draft in Thread.
9. Open Clear AI Label and set Thread ID to {{ $('Map Data Fields').item.json["threadId"] }} with Operation set to removeLabels.
10. Credential Required: Connect your gmailOAuth2 credentials in Clear AI Label.

Step 3: Set Up Query AI Assistant

Configure the AI assistant that generates the response content for each thread.

1. Open Query AI Assistant and set Resource to assistant.
2. Set Text to {{ $json.text }} and Prompt to define.
3. Select your assistant in Assistant ID (replace [YOUR_ID] with your actual assistant).
4. Credential Required: Connect your openAiApi credentials in Query AI Assistant.

Step 4: Configure Processing and Draft Creation

Map the AI response and Gmail metadata, format HTML, build the raw email, encode it, and create a draft in the original thread.

1. In Iterate Through Threads, keep the default batch settings. Note that Iterate Through Threads outputs to both Fetch Message Details and Retrieve Thread Messages in parallel.
2. In Select Latest Message, keep Keep set to lastItems, then connect it back to Iterate Through Threads.
3. In Map Data Fields, map the following values:
   response → {{ $json.output }}
   threadId → {{ $('Fetch Message Details').item.json["threadId"] }}
   to → {{ $('Fetch Message Details').item.json["from"]["text"] }}
   subject → {{ $('Fetch Message Details').item.json["subject"] }}
   messageId → {{ $('Retrieve Labeled Threads').item.json["id"] }}
4. In Render HTML Response, set Mode to markdownToHtml and Markdown to {{ $json.response }}, saving to Destination Key response.
5. In Compose Raw Email, set raw to:
   To: {{ $json.to }} Subject: {{ $json.subject }} Content-Type: text/html; charset="utf-8" {{ $json.response }}
6. In Encode Email Base64, keep the JavaScript that sets encoded using Buffer.from($json.raw).toString('base64').
7. In Create Draft in Thread, set JSON Body to:
   {"message":{"raw":"{{ $json.encoded }}", "threadId": "{{ $('Map Data Fields').item.json["threadId"] }}"}}
8. Ensure Create Draft in Thread flows into Clear AI Label to remove the AI label after drafting.

Step 5: Test and Activate Your Workflow

Validate that messages are drafted correctly and labels are cleared before enabling the schedule.

1. Click Execute Workflow to run a manual test using the current labeled threads.
2. Verify that Query AI Assistant returns a response and that Create Draft in Thread creates a draft in Gmail.
3. Check that Clear AI Label removes the label from the processed thread.
4. Once successful, toggle the workflow to Active to run on the schedule.