Telegram photo manager for Google Drive with AI edits

Telegram photo manager for Google Drive with AI edits

An n8n workflow for a Telegram bot that keeps your photos in a Google Drive folder. Send it a photo and it saves the
picture and gives it a sensible name. Send it a message and an OpenAI agent renames a file, finds one, or hands an edit or
a merge of two pictures to an image sub-workflow of your own.

What is inside

• workflow.json: the workflow as n8n exports it, 18 nodes. Five are sticky notes that label its parts: the input, the
  agent, its model, the image tools and the file tools. It holds no credential.
• README.md: these notes.

What we changed from the original copy

• Respond had an empty chat ID, so no reply could ever reach Telegram and every run ended in an error there. It now
  answers the chat the message came from: {{ $('Telegram Trigger').item.json.message.chat.id }}.
• The author's two Drive folders are gone. Upload file and Search Raw Files ask for YOUR_PICTURES_FOLDER_ID, and
  Search AI Images for YOUR_OUTPUT_FOLDER_ID.
• Edit Image and Combine Images pointed at sub-workflows in the author's own n8n. Both now arrive empty, and you pick
  your own.

Before you start

• An n8n instance, on n8n Cloud or your own server, reachable from the internet so Telegram can call it. The agent node
  is version 3, so use a recent n8n.
• A Telegram bot and its token, from BotFather.
• A Google account, and two Drive folders: one for the photos you send, one for the images the sub-workflows make.
• An OpenAI API key that can use gpt-5-mini. Read Good to know about its shutdown date.
• Two image sub-workflows of your own, for edits and for merges (below).

Set it up

1. In n8n, create a workflow, open its menu and choose Import from File. Pick workflow.json.
2. Add a Telegram credential with your bot's token. Select it in Telegram Trigger, Download File and Respond.
3. Add a Google Drive OAuth2 credential. Select it in Upload file, Change Name, Search Raw Files and
   Search AI Images.
4. Copy each folder's ID, the last part of its address after /folders/. Put the photos folder's in Upload file and
   Search Raw Files in place of YOUR_PICTURES_FOLDER_ID, and the output folder's in Search AI Images in place of
   YOUR_OUTPUT_FOLDER_ID.
5. Add an OpenAI credential and select it in OpenAI Chat Model.
6. Build the two sub-workflows, then pick them in Edit Image and Combine Images.
7. In Telegram Trigger, add your Telegram user ID under Restrict to User IDs, so strangers cannot use the bot.
8. Save and press Publish. n8n registers the bot's webhook with Telegram when the workflow goes live.

What each sub-workflow receives

The agent fills every field. The image IDs are Google Drive file IDs, so each sub-workflow needs Drive access to download
the pictures. Save what it makes to the output folder, where Search AI Images looks.

Edit Image       imageTitle, imagePrompt, imageID
Combine Images   prompt, image1, image2, imageTitle

Their names in the original copy, Edit Image Nanobanana Tool and Combine Images Nanobanana, point to Google's Gemini
image model, known as Nano Banana, but any image model that edits a picture from a prompt will do.

How a run goes

1. Telegram Trigger receives each new message, and Switch sends a photo one way and text the other.
2. For a photo, Download File fetches it from Telegram, and Upload file saves it to the photos folder under the day's
   date, such as 2026-11-02. Set Text then tells the agent that a picture came in, with its Drive file ID.
3. AI Agent runs gpt-5-mini through OpenAI Chat Model. Its system message tells it to name a new photo at once with
   Change Name and say what it chose, without asking you.
4. For text, the agent works out what you want. It finds files with Search Raw Files or Search AI Images, renames them
   with Change Name, and sends edits to Edit Image and merges to Combine Images.
5. Respond sends the agent's answer to your chat, formatted as HTML.

Change the look

• How it names and behaves: the system message of AI Agent. Its example names show the style it aims for.
• The model: OpenAI Chat Model, now gpt-5-mini.
• The first name a photo gets: the name of Upload file, today's date until the agent renames it.
• What a search returns: the fields of Search Raw Files and Search AI Images, now the ID, the name and the link.

Good to know

• Model shutdown: OpenAI's deprecations page, read on 2 November 2026, lists gpt-5-mini for shutdown on 11 December
  2026, with gpt-5.6-terra as its replacement.
• Photo size: Download File takes the second of the sizes Telegram sends with each photo, a medium copy, so that is what
  lands in Drive. Use {{ $('Telegram Trigger').item.json.message.photo.last().file_id }} in its file ID for the largest.
  A picture sent as a file matches neither rule of Switch, so the bot ignores it.
• No memory: the system message speaks of earlier messages and recent uploads, but the agent has no memory. Each message
  stands alone, so name the file you mean, or let it search.
• Long folders: each search lists every file in its folder, and the whole list goes to OpenAI. A big folder makes every
  search slower and dearer.
• HTML replies: Respond sends the answer as HTML, and Telegram refuses a message with a stray < or & in it. Clear
  the parse mode in Respond if that happens.
• Who can use it: without a user restriction on Telegram Trigger, anyone who finds the bot can write to your Drive and
  spend your OpenAI credit.
• Your data: your photos go from Telegram to your Drive, your messages and the file names go to OpenAI, and the
  sub-workflows get the file IDs the agent passes on.
• Credentials: the JSON holds none, and nothing from its author's Drive.