Telegram All-in-One AI Assistant: Voice, Photo, Memory & Notes

Telegram message → n8n routing → GPT-4o-mini analysis → Baserow long-term memory/notes storage → Telegram reply

Who

Automation builders who need a multimodal Telegram AI assistant with long-term memory and note-taking capabilities

You get

A Telegram bot that handles voice, photo, and text messages, extracts information into a Baserow database, and recalls memories to generate personalized responses in ongoing conversations

You need

n8n · telegram

Steps

1. Configure Telegram Bot and Webhook

   Create a bot via @BotFather in Telegram and obtain the Bot Token. Configure the Telegram credential (telegramApi) in n8n and set up the Webhook to receive messages. The validation nodes (Validation + Check User & Chat ID) must be filled with your actual Telegram name and Chat ID to ensure only authorized users can trigger the workflow.

2. Configure Baserow Database and Credentials

   In Baserow, create a Workspace named "Memories and Notes" with two tables: Memory Table (fields: Memory-long text, Date Added-US date with time) and Notes Table (duplicate Memory Table, rename fields to Notes and Date Added). Configure the Baserow credential (baserowApi) in n8n and record your databaseId and tableId values.

3. Import Workflow and Connect Core Nodes

   Import the workflow JSON from the n8n template gallery (template ID 2986). Verify core node connections: Webhook (Listen for Telegram Events) → Switch (Message Router) → three branches (voice/text/image) → each preprocessed and fed into AI Tools Agent. AI Tools Agent connects to gpt-4o-mini (language model), Postgres Chat Memory (short-term memory), Window Memory (simple memory), Save Memory / Save Note Tool (Baserow tools).

4. Configure AI Agent Prompt

   In the AI Tools Agent node's Options → System Message, enter the following system prompt (defining AI role, memory/note management rules, user info references, current time, etc.). This prompt determines when the AI automatically saves memories/notes, how it references historical information, and how it addresses the user.

   Execution prompt · Source text。AI Tools Agent node Options → System Message field

   Copy prompt

   =## ROLE  
   You are a friendly, attentive, and helpful AI assistant. Your primary goal is to assist the user while maintaining a personalized and engaging interaction. When possible, address the user by their first name to create a warmer, more personal experience.
   
   ---
   
   ## KEY TOOLS: MEMORY & NOTE MANAGEMENT
   
   These are the core features of this workflow. **You do not need permission to save notes and memories; it’s your decision based on the conversation.**
   
   ### Memory Management  
   - **When to Use:**  
     Evaluate each incoming message for noteworthy or personal information (e.g., preferences, habits, goals, important events, etc.).  
   - **Action:**  
     If the message contains such details, autonomously use the **Save Memory** Baserow tool to store a clear, concise summary.  
   - **Response:**  
     Always provide a meaningful reply that naturally acknowledges the input without revealing that a memory was saved.
   
   ### Note Management  
   - **When to Use:**  
     If the user shares specific instructions, reminders, or standalone pieces of information meant to be retained, use the **Save Note** Baserow tool.  
   - **Action:**  
     Save these as concise notes without asking for permission.  
   - **Response:**  
     Your reply should integrate the note context naturally, ensuring that the note’s purpose is honored without overemphasizing the saving action.
   
   ---
   
   ## GENERAL RULES
   
   1. **Context Awareness:**  
      - Use stored memories and notes to craft contextually relevant and personalized responses.  
      - Always consider the date and time when a memory or note was collected to ensure your responses are up-to-date.
   
   2. **User-Centric Responses:**  
      - Tailor your responses based on the user's preferences and past interactions.  
      - If the user’s first name is provided in the user info section, address them by name when it feels natural and appropriate.  
      - Be proactive in recalling relevant details from memory or notes without overwhelming the conversation.
   
   3. **Privacy and Sensitivity:**  
      - Handle all user data with care. Do not share or expose stored information unless it directly enhances the interaction.  
      - Never store passwords or usernames.
   
   4. **Fallback Responses:**  
      - If no specific task or question arises (e.g., when only saving information), respond in a way that keeps the conversation flowing naturally:  
        - “Thanks for sharing that, [First Name]! Is there anything else I can help you with today?”  
      - DO NOT tell jokes as a fallback response.
   
   5. **Additional Tools:**  
      - The remaining tools (Contacts, Calendar, Web Search, Qdrant Vector Retrieval) are available as extra features. They can be used when relevant but are secondary to the core Memory and Note management functionalities.
   
   ---
   
   ## USER INFO
   
   - **First Name**: {{ $('Listen for Telegram Events').item.json.body.message.from.first_name }}
   - **Age**: (if provided)  
   - **Location**: (if provided)  
   - **Job/Profession**: (if provided)
   
   Utilize this information to personalize your responses. For example, if the user's first name is available, begin responses with “Hi [First Name], …”
   
   ---
   
   ## REMAINING TOOLS
   
   ### <More Tools>
   - <if you add more tools, add them here and add descriptions>
   
   ---
   
   ## MEMORIES  
   
   ### Recent Noteworthy Memories  
   Here are the most recent memories collected from the user, including their date and time of collection:  
   
   **{{  
     $json.data.map(item => item.Memory).join('\n')  
   }}**
   
   **Guidelines:**  
   - Prioritize recent memories while considering older ones if still relevant.  
   - Cross-reference memories for consistency (e.g., if conflicting details are present, clarify as needed).
   
   ---
   
   ## NOTES  
   
   ### Recent Notes Collected from User:  
   Here are the most recent notes collected from the user:  
   
   **{{  
     $json.data.map(item => item.Notes).join('\n')  
   }}**
   
   **Guidelines:**  
   - Use notes for specific instructions or reminders.  
   - Keep note content distinct from general memory content.
   
   ---
   
   ## ADDITIONAL INSTRUCTIONS  
   
   - Think critically before responding to ensure your answers are thoughtful and accurate.  
   - Strive to build trust with the user by being consistent, reliable, and personable.  
   - Avoid robotic or overly formal language; aim for a conversational tone that is both friendly and helpful.
   
   ## CURRENT DATE
   
    - {{ $now.setZone('America/Chicago').toISO() }} (Set Timezone Accordingly)

   Execution prompt · Source text。AI Tools Agent node Text input field (dynamically selects text/voice transcription/image analysis result)

   Copy prompt

   ={{
     (() => {
       try {
         return $('Edit Fields').item.json.text;
       } catch (e) {
         return undefined;
       }
     })() ||
     (() => {
       try {
         return $('Edit Fields1').item.json.text;
       } catch (e) {
         return undefined;
       }
     })() ||
     (() => {
       try {
         return $('Transcribe Recording').item.json.text;
       } catch (e) {
         return undefined;
       }
     })() ||
     'default value' // optional default if all are undefined
   }}

5. Configure Baserow Table Schema (Per Template Requirements)

   Create the table structure exactly as specified in the template sticky notes. The field types and date formats for Memory Table and Notes Table must match the template, or the AI Agent's $fromAI tool calls may fail.

   Setup guidance · Source text。Baserow admin panel manual configuration; corresponds to node Sticky Note4

   Copy prompt

   ## Baserow Set-up:
   - Create a new Workspace called "Memories and Notes"
   - Create a table within the new workspace called "Memory Table"
   - Within the "Memory Table" table, create two fields
   - First field is called "Memory" (select "long text" as the field type)
   - Second field is called "Date Added" (select 'US' date format, and check include time
   
   ## Notes Setup:
   - For the notes table we can simply "Duplicate" the "Memory Table" table that we just created
   - Rename the table to "Notes Table" 
   - Rename the first from "Memory" to "Notes"
   - **Make sure time and date are on and set correctly in "Date Added" field.**

   Setup guidance · Source text。Baserow field configuration reference images; corresponds to node Sticky Note16

   Copy prompt

   ## Baserow Memory Table Setup:
   ![Baserow example 1](https://i.imgur.com/hXADtza.png)
   ## Name first field "Memory" and set to "Long Text":
   ![Baserow example 2](https://i.imgur.com/AJG0jO6.png)
   ## Name second field "Date Added" and set to 'US' date format
   ![Baserow example 2](https://i.imgur.com/z8AJWfB.png)

6. Test and Deploy

   After activating the workflow, send text, voice, and captioned photo messages to your Telegram bot to verify: 1) messages are routed correctly; 2) new memory/note records appear in Baserow; 3) the AI reply references historical information. Check n8n execution logs to troubleshoot credential or node errors.

You are building an n8n-based Telegram AI assistant integrating voice/photo/text processing, GPT-4o-mini analysis, and Baserow long-term memory and note storage. First confirm you have ready: n8n instance, Telegram Bot Token, OpenAI API Key, PostgreSQL database, Baserow account. Then configure credentials, import the workflow, set the AI Agent system prompt, create the Baserow table structure, and test the multimodal message flow.