Multi-modal expense tracking with GPT-4, Gemini OCR, and voice via Telegram

Open on n8n.io

$20/month : Unlimited workflows

2500 executions/month

Try free

THE #1 IN WEB SCRAPING

Scrape any website without limits

Try free

HOSTINGER

Early Deal
DISCOUNT 20%

Self-hosted n8n

Unlimited workflows - from $4.99/mo

Try free

#1 hub for scraping, AI & automation

6000+ actors - $5 credits/mo

Try free

Important notice

This workflow is provided as-is. Please review and test before using in production.

1. Workflow Overview

This n8n template creates an intelligent expense tracking system that processes text, voice, and receipt images through Telegram. The assistant automatically categorizes expenses , handles curre...

Best for

Multimodal AI automation workflows
advanced n8n builders looking for reusable templates

Tools used

@n8n/n8n-nodes-langchain.agent, n8n-nodes-base.telegramtrigger, @n8n/n8n-nodes-langchain.mcpclienttool, @n8n/n8n-nodes-langchain.mcptrigger, @n8n/n8n-nodes-langchain.toolcalculator, n8n-nodes-base.googlesheetstool, n8n-nodes-base.httprequesttool, n8n-nodes-base.set

Source and attribution

This workflow is cataloged by N8N Workflows and links back to its original n8n.io source page by Oussama.

Original n8n.io source

1.1 Workflow description

Title: Multi-modal expense tracking with GPT-4, Gemini OCR, and voice via Telegram
Workflow name: Multi-modal expense tracking with GPT-4, Gemini OCR, and voice via Telegram

This n8n template creates an intelligent expense tracking system 🤖 that processes text, voice, and receipt images through Telegram. The assistant automatically categorizes expenses, handles currency conversions 🌍, and maintains financial records in Google Sheets while providing smart spending insights 💡.

Use Cases:

🗣️ Personal expense tracking via Telegram chat
🧾 Receipt scanning and data extraction
💱 Multi-currency expense management
📂 Automated financial categorization
🎙️ Voice-to-expense logging
📊 Daily/weekly/monthly spending analysis

How it works:

Multi-Input Processing: Telegram trigger captures text messages, voice notes, and receipt images.
Content Analysis: A Switch node routes different input types (text, audio, images) to appropriate processors.
Voice Processing: ElevenLabs converts voice messages to text for expense extraction.
Receipt OCR: Google Gemini analyzes receipt images to extract amounts and descriptions.
Expense Classification: An LLM determines if the input is an expense or a general query.
Expense Parsing: For multiple expenses, the AI splits and normalizes each item.
Currency Conversion: An exchange rate API converts foreign currencies to USD.
Smart Categorization: The AI agent assigns expenses to predefined categories with emojis.
Data Storage: Google Sheets stores all expense records with automatic totals.
Intelligent Responses: The agent provides spending summaries, alerts, and financial insights.

Requirements:

🌐 Telegram Bot API access
🤖 OpenAI, Gemini, or any other AI model
🗣️ ElevenLabs API for voice processing
📝 Google Sheets API access
💹 Exchange rate API access

Good to know:

⚠️ Daily spending alerts trigger when expenses exceed 100 USD.
🏷️ Supports 12 predefined expense categories with emoji indicators.
🔄 Automatic currency detection and conversion to USD.
🎤 Voice messages are processed through speech-to-text.
📸 Receipt images are analyzed using computer vision.

Customizing this workflow:

✏️ Modify expense categories in the system prompt.
📈 Adjust spending alert thresholds.
💵 Change the base currency from USD to your preferred currency.
✅ Add additional expense validation rules.
🔗 Integrate with other financial platforms.

1.2 Logical Blocks

This catalog entry is organized from the workflow JSON. The node-level section below shows the executable blocks available for review before importing the template.

2. Block-by-Block Analysis

Block 1 - Agent

Type / Role: @n8n/n8n-nodes-langchain.agent - agent
Config choices: Version 1.7

Block 2 - Telegram Trigger

Type / Role: n8n-nodes-base.telegramTrigger - telegramTrigger
Config choices: Version 1.2

Block 3 - MCP Client

Type / Role: @n8n/n8n-nodes-langchain.mcpClientTool - mcpClientTool
Config choices: Version 1

Block 4 - MCP Server Trigger

Type / Role: @n8n/n8n-nodes-langchain.mcpTrigger - mcpTrigger
Config choices: Version 1

Block 5 - Calculator

Type / Role: @n8n/n8n-nodes-langchain.toolCalculator - toolCalculator
Config choices: Version 1

Block 6 - Total_Spent_Tool

Type / Role: n8n-nodes-base.googleSheetsTool - googleSheetsTool
Config choices: Version 4.5

Block 7 - Balance_Tool

Type / Role: n8n-nodes-base.googleSheetsTool - googleSheetsTool
Config choices: Version 4.5

Block 8 - Expenses_Tool

Type / Role: n8n-nodes-base.googleSheetsTool - googleSheetsTool
Config choices: Version 4.6

Block 9 - Exchange_Rate

Type / Role: n8n-nodes-base.httpRequestTool - httpRequestTool
Config choices: Version 4.2

Block 10 - Add_Expenses_Tool

Type / Role: n8n-nodes-base.googleSheetsTool - googleSheetsTool
Config choices: Version 4.5

Block 11 - Read_Rows_Tool

Type / Role: n8n-nodes-base.googleSheetsTool - googleSheetsTool
Config choices: Version 4.6

Block 12 - Rename Fields

Type / Role: n8n-nodes-base.set - set
Config choices: Version 3.4

Block 13 - Sticky Note

Type / Role: n8n-nodes-base.stickyNote - stickyNote
Config choices: Version 1

Block 14 - Sticky Note1

Type / Role: n8n-nodes-base.stickyNote - stickyNote
Config choices: Version 1

Block 15 - Get Voice File from Telegram

Type / Role: n8n-nodes-base.telegram - telegram
Config choices: Version 1.2

Block 16 - Convert Voice to Text (ElevenLabs)

Type / Role: n8n-nodes-base.httpRequest - httpRequest
Config choices: Version 4.2

Block 17 - Route Input by Message Type

Type / Role: n8n-nodes-base.switch - switch
Config choices: Version 3.2

Block 18 - Download Receipt Image

Type / Role: n8n-nodes-base.telegram - telegram
Config choices: Version 1.2

Block 19 - Extract Receipt Data

Type / Role: @n8n/n8n-nodes-langchain.googleGemini - googleGemini
Config choices: Version 1

Block 20 - Classify Message Intent

Type / Role: @n8n/n8n-nodes-langchain.chainLlm - chainLlm
Config choices: Version 1.6

Block 21 - Route to Expense or Query Processing

Type / Role: n8n-nodes-base.if - if
Config choices: Version 2.2

Block 22 - Parse and Split Multiple Expenses

Type / Role: @n8n/n8n-nodes-langchain.chainLlm - chainLlm
Config choices: Version 1.7

Block 23 - Intent Classification Model

Type / Role: @n8n/n8n-nodes-langchain.lmChatAzureOpenAi - lmChatAzureOpenAi
Config choices: Version 1

Block 24 - Expense Parsing Model

Type / Role: @n8n/n8n-nodes-langchain.lmChatAzureOpenAi - lmChatAzureOpenAi
Config choices: Version 1

Showing the first 24 of 33 workflow blocks. Download the JSON for the full node graph.

3. Summary Table

Workflow	Multi-modal expense tracking with GPT-4, Gemini OCR, and voice via Telegram
Complexity	advanced
Nodes	33
Categories	Multimodal AI
Author	Oussama
Published	29 Aug 2025

4. Reproducing the Workflow from Scratch

1. Download the workflow JSON

Use the JSON export at /data/workflows/8035/8035.json as the source template for this automation.
2. Import the template into n8n

Open n8n, import the downloaded JSON, and review each node before activating the workflow.
3. Configure credentials and variables

Replace placeholder credentials, API keys, webhook URLs, account IDs, and environment-specific values with your own settings.
4. Test with sample data

Run the workflow manually or in a staging workspace, inspect node output, and confirm downstream systems receive the expected data.
5. Activate and monitor

Enable the workflow only after testing, then monitor executions, errors, and rate limits during the first production runs.

5. General Notes & Resources

Review imported nodes carefully before activation. This catalog entry is intended to help you inspect the workflow structure, understand required services, and find related templates faster.

Node names, credentials, schedules, webhook paths, and external service limits may need adjustment for your workspace.

Download workflow JSON Original n8n.io source Multimodal AI workflows

Frequently asked questions

What does Multi-modal expense tracking with GPT-4, Gemini OCR, and voice via Telegram do?

This n8n template creates an intelligent expense tracking system that processes text, voice, and receipt images through Telegram. The assistant automatically categorizes expenses , handles curre...

What do I need before importing this workflow?

Review the workflow JSON, configure any required credentials in n8n, and test the automation in a safe workspace before using it in production.

Can I customize this workflow?

Yes. Use the block-by-block analysis and the downloadable JSON to inspect each node, then adjust credentials, prompts, schedules, filters, or destinations for your Multimodal AI use case.

Oussama

3 workflows

Nodes

@n8n/n8n-nodes-langchain.agent n8n-nodes-base.telegramtrigger @n8n/n8n-nodes-langchain.mcpclienttool @n8n/n8n-nodes-langchain.mcptrigger @n8n/n8n-nodes-langchain.toolcalculator n8n-nodes-base.googlesheetstool n8n-nodes-base.httprequesttool n8n-nodes-base.set

Complexity

advanced

Published 29 Aug 2025

Likes 0

View on n8n.io Download Workflow

Install path: /data/workflows/8035/8035.json

Share Your Workflow

Have a useful automation to share? Publish it and help the community.

Submit Your Template How to Submit

Related Workflows

Send daily LinkedIn post drafts with Tavily, Groq Llama 3.3 and Gmail

## Quick overview This workflow runs daily at 8 AM to rotate through a topic list, research it with Tavily web search, generate a LinkedIn post and image prompt with Groq (Llama 3.3 70B), create a Pollinations.ai image URL, and email you the post with the image attached via Gmail. ## How it works 1. Runs every day at 8:00 AM on a schedule. 2. Selects today’s topic from a rotating list based on the day of the year. 3. Searches the web with Tavily to fetch up-to-date results for the selected topic. 4. Uses a Groq-hosted Llama model to write a LinkedIn post and a matching image prompt grounded in the Tavily results. 5. Builds a Pollinations.ai image generation URL from the image prompt and downloads the generated image file. 6. Emails the post text with the image attached via Gmail, and sends a text-only fallback email if image generation or download fails. ## Setup 1. Add a Groq API credential to the Groq Chat Model node. 2. Create a Tavily API key and configure HTTP Header Auth with `Authorization: Bearer <YOUR_TAVILY_KEY>` for the Tavily Search request. 3. Connect a Gmail OAuth2 credential and update the recipient email address in both Gmail send steps. 4. Edit the topic list in the topic-selection code step if you want different rotating themes. ## Requirements - Groq API key (free tier available) - Tavily API key (free tier — 1,000 searches/month) - Gmail account (OAuth2) ## Customization - Edit the topic list inside the "Pick Today's Topic" code node to change what subjects the post rotates through. - Change the Groq model in the "Groq Chat Model" node if you want a different LLM (e.g. a smaller/faster or larger Groq model). - Adjust the schedule trigger if you want a different posting time, or swap the Gmail "send to" address to deliver the draft to someone else for review. ## Additional info This workflow emails you a draft for manual review before posting — it does not post to LinkedIn automatically. If image generation fails, a text-only fallback email is sent instead so you never miss the post text.

View

Send low-stock inventory reorder emails from Snowflake with Claude and Outlook

## Quick overview This workflow runs daily, queries Snowflake for items at or below their reorder level, and groups them by supplier. For each supplier, it uses Anthropic Claude to draft a reorder email, optionally appends details to a Google Docs log, and sends the email via Microsoft Outlook. ## How it works 1. A Schedule Trigger fires every day at 07:00 (Europe/Helsinki timezone) and passes through a Workflow Config node that holds your optional Google Docs log ID. 2. A Snowflake query fetches all inventory rows where current_stock <= reorder_level. If no items qualify, the workflow exits cleanly via a Pass Through node with no further action. 3. A Code node groups the low-stock items by supplier_name, producing one batch per vendor. A SplitInBatches loop then processes each supplier one at a time. 4. For each supplier, a Set node extracts the supplier email, item count, and a pre-rendered Markdown item list, then passes everything to an LLM chain powered by Claude (claude-sonnet-4-6). 5. Claude generates a concise, professional reorder request email body in Markdown, constrained by a system prompt to return only the email content with no preamble. 6. The Markdown output is converted to HTML and optionally appended as a one-line entry to a Google Docs audit log (skipped if no doc ID is configured). The workflow then sends the HTML email to the supplier via Microsoft Outlook before advancing to the next vendor. ## Setup 1. Open the Workflow Config node and set log_doc_id to the document ID from your Google Doc URL (the string between /d/ and /edit). Leave it empty to skip audit logging entirely. 2. Add your Snowflake credentials to the Query Snowflake DB node and verify the query matches your database name, schema, and table. The expected columns are item_id, item_name, current_stock, reorder_level, supplier_name, and supplier_email. 3. Add your Anthropic credentials to the Claude Email Generation node. 4. Add your Microsoft Outlook credentials to both the Send Reorder Email and Send Error Alert nodes. Update the toRecipients field in Send Error Alert to your alert address. 5. If logging is enabled, add your Google Docs credentials to the Log to Google Docs node. 6. In n8n Settings > Error Workflow, point to a dedicated error-handler sub-workflow that wires into the Send Error Alert node (it is intentionally unwired in this workflow). ## Requirements - Snowflake account with an inventory table containing current_stock, reorder_level, supplier_name, and supplier_email columns - Anthropic API credentials (claude-sonnet-4-6) - Microsoft 365 account with Outlook access - Google Docs OAuth2 credentials (only if audit logging is enabled) ## Customization - Adjust the Snowflake query to match your schema, add filters by warehouse or category, or join additional tables for richer item metadata. - Modify the Claude system prompt in Build Email Draft to change tone, language, or email structure — for example to generate Finnish-language emails or include purchase order references. - Change the cron expression in When 7am Daily to run at a different time or frequency. - Extend the log entry format in Log to Google Docs to include item names, SKUs, or a hyperlink back to the workflow execution. - Replace the Google Docs logging step with a Slack notification, a row insert into PostgreSQL, or any other audit destination by swapping the node after Check Configured Doc ID.

View

Generate VEO3 videos from form ideas with ChatGPT, fal.ai, Drive and Gmail

## Quick overview This workflow collects a video idea via an n8n Form, uses OpenAI to turn it into an optimized prompt, generates an 8-second VEO3 video with native audio via fal.ai, saves the MP4 to Google Drive, and emails the Drive link via Gmail. ## How it works 1. Receives a submission from an n8n Form with the requester’s email, a plain-English video idea, and a selected style. 2. Uses OpenAI (via an AI Agent) to convert the idea and style into a structured VEO3 title and detailed prompt returned as JSON. 3. Cleans and parses the AI JSON, creates a filename-safe title, and sets the generation parameters to 16:9 and 8 seconds. 4. Submits the prompt to fal.ai’s queue endpoint for the fal-ai/veo3 model and gets a request ID. 5. Polls fal.ai every 30 seconds until the job status returns COMPLETED. 6. Retrieves the final MP4 URL from fal.ai, downloads the video, uploads it to Google Drive, and emails the Google Drive web link to the requester via Gmail. ## Setup 1. Create a fal.ai account, generate an API key, and replace `YOUR_FAL_AI_API_KEY` in the fal.ai HTTP request headers. 2. Add an OpenAI API credential in the OpenAI Chat Model node and select the model you want to use. 3. Connect Google Drive OAuth2 and replace `YOUR_GDRIVE_FOLDER_ID` with the target folder ID where videos should be saved. 4. Connect Gmail OAuth2 for sending emails, then activate the workflow and use the Form URL (path `veo3-video-generator`) to submit requests.

View

Need Custom Automation?

Get help designing a custom n8n workflow that connects your stack and fits your process.

Multi-modal expense tracking with GPT-4, Gemini OCR, and voice via Telegram

Workflow preview

Important notice

1. Workflow Overview

Best for

Tools used

Source and attribution

1.1 Workflow description

Use Cases:

How it works:

Requirements:

Good to know:

Customizing this workflow:

1.2 Logical Blocks

2. Block-by-Block Analysis

Block 1 - Agent

Block 2 - Telegram Trigger

Block 3 - MCP Client

Block 4 - MCP Server Trigger

Block 5 - Calculator

Block 6 - Total_Spent_Tool

Block 7 - Balance_Tool

Block 8 - Expenses_Tool

Block 9 - Exchange_Rate

Block 10 - Add_Expenses_Tool

Block 11 - Read_Rows_Tool

Block 12 - Rename Fields

Block 13 - Sticky Note

Block 14 - Sticky Note1

Block 15 - Get Voice File from Telegram

Block 16 - Convert Voice to Text (ElevenLabs)

Block 17 - Route Input by Message Type

Block 18 - Download Receipt Image

Block 19 - Extract Receipt Data

Block 20 - Classify Message Intent

Block 21 - Route to Expense or Query Processing

Block 22 - Parse and Split Multiple Expenses

Block 23 - Intent Classification Model

Block 24 - Expense Parsing Model

3. Summary Table

4. Reproducing the Workflow from Scratch

1. Download the workflow JSON

2. Import the template into n8n

3. Configure credentials and variables

4. Test with sample data

5. Activate and monitor

5. General Notes & Resources

Frequently asked questions