This can be your Alternative to Cal AI App

Overview:

A comprehensive n8n workflow demonstrating advanced AI agent orchestration, stateful conversation management, and multi-modal input processing for nutrition tracking applications.

Technical Architecture:

This workflow showcases production-ready patterns for building complex conversational AI systems in n8n:

1. Multi-Agent System Design

Router Agent: Analyzes user intent and routes to specialized sub-agents Registration Agent: Handles new user onboarding with guided data collection Meal Logging Agent: Processes text, image, and voice inputs for nutrition analysis Update Agent: Manages multi-turn conversations for meal corrections Report Agent: Generates daily nutrition summaries Profile Agent: Handles user preference updates

2. Stateful Conversation Management The workflow implements a state machine using n8n Data Tables:

Tracks conversation context across messages Enables multi-step workflows (e.g., "get meal ID → ask for new description → update") Prevents context loss in long conversations Auto-resets state after task completion

3. Multi-Modal Input Processing

Text: Direct AI analysis using Google Gemini Images: Gemini Vision API with specialized nutrition analysis prompts Voice: Audio-to-text transcription followed by analysis Unified output format regardless of input type

4. Advanced Prompt Engineering

Specialized system prompts for each agent role Structured output formatting for reliable data extraction Regional cuisine knowledge (African/West African foods) Error handling and clarification requests built into prompts

5. Data Persistence Layer Uses n8n Data Tables for:

User profiles (calories, protein targets, preferences) Meal logs (descriptions, macros, timestamps) Conversation state (current step, context variables)

6. Telegram Integration

MarkdownV2 formatting with auto-chunking for long messages Handles all Telegram message types (text, photo, voice) "Typing" indicators for better UX

Learning Outcomes:

By studying this workflow, you'll understand:

How to build agent hierarchies with specialized roles State management patterns for complex conversations Multi-modal AI input processing techniques Production data persistence strategies in n8n Scalable webhook-based bot architecture

Technical Requirements:

n8n version 1.0+ (uses Data Tables feature) Google Gemini API access Telegram Bot Token Basic understanding of: AI agents, webhook triggers, data tables

Difficulty Level: Intermediate to Advanced

Use Cases Beyond Nutrition: The patterns here apply to any domain requiring:

Multi-step user onboarding Data collection through conversation Record updates via natural language Report generation Profile management

Components:

50+ nodes demonstrating n8n AI features 6 specialized AI agents 3 data tables with structured schemas Custom JavaScript for MIME type handling and markdown formatting Integration with Google Gemini (text + vision + audio models)

What Makes This Workflow Unique:

Most n8n AI examples show simple Q&A bots. This demonstrates enterprise-grade conversation management with real data persistence, making it suitable for actual production deployment.

SETUP GUIDE

Need Help?

Reach me: https://www.linkedin.com/in/gerald-akhidenor-1ab1a45/

Work with me: https://dominixai.com/

My website: https://jobmonkey.dev

My email: [email protected]

Cal AI Nutrition Bot - Complete Setup Guide

Who This Guide Is For: This workflow is designed for beginner to intermediate no-code builders who want to launch a nutrition tracking bot. You should be comfortable:

Creating accounts on platforms (n8n, Google Cloud, Telegram) Following step-by-step instructions Copy-pasting API keys Basic JSON importing

No programming experience required. If you can use Zapier or Make.com, you can set this up.

Prerequisites (30 minutes to gather)

1. n8n Account

Option A (Easiest): Sign up for n8n Cloud at n8n.io (14-day free trial, then $20/month) Option B (Free): Self-host n8n on DigitalOcean/Railway/your server

2. Google Gemini API Key

Go to ai.google.dev Click "Get API Key" Create new project (free tier: 15 requests/minute) Copy your API key

3. Telegram Bot

Open Telegram and search for @BotFather Send /newbot command Follow prompts to create your bot Copy the bot token (looks like: 1234567890:ABCdefGHIjklMNOpqrsTUVwxyz)

Step 1: Import the Workflow (5 minutes)

Log into your n8n instance Click "New Workflow" in top left Click the three-dot menu → Import from File Upload the My_workflow.json file you downloaded The workflow will appear with all nodes connected

Step 2: Create Data Tables (10 minutes) The workflow uses three data tables to store information. Create them exactly as shown: Table 1: Cal AI Profiles

Navigate to n8n Data Tables section Click Create Data Table Name: Cal AI Profiles Add columns:

User_ID (Text) Name (Text) Email (Text) Calories_target (Number) Protein_target (Number) Country (Text)

Table 2: Cal AI Meals

Create new table: Cal AI Meals Add columns:

Meal_ID (Text) User_ID (Text) Date (Text) Meal_description (Text) Calories (Number) Proteins (Number) Carbs (Text) Fats (Text)

Table 3: Cal AI StateManagement

Create new table: Cal AI StateManagement Add columns:

User_ID (Text) State (Text) Last_Interaction (DateTime) Context_MealID (Text)

Step 3: Configure Credentials (15 minutes) A. Add Google Gemini Credential

In any Google Gemini node, click "Create New Credential" Paste your Gemini API key Click Save This credential will auto-apply to all Gemini nodes

B. Add Telegram Credential

Open any Telegram node Click "Create New Credential" Paste your bot token Click Save

C. Link Data Tables

Open the "Is User Registered?" node Click the Data Table dropdown Select "Cal AI Profiles" Repeat for all nodes with data table connections (n8n will show which ones need updating)

Step 4: Activate Telegram Webhook (5 minutes)

Open the "Telegram Trigger" node Click "Execute Node" to register the webhook You'll see "Webhook registered successfully" Click "Listen for Test Event" to verify Send a message to your bot on Telegram You should see the message appear in n8n

Step 5: Test Each Function (20 minutes) Test 1: User Registration

Send any message to your bot Bot should ask for your name Follow the prompts to complete registration Check the "Cal AI Profiles" data table to confirm your data was saved

Test 2: Meal Logging (Text)

Send: "I had 2 eggs and toast" Bot should respond with nutrition breakdown

Test 3: Meal Logging (Image)

Take a photo of food Send it to the bot Bot should analyze and log the meal

Test 4: Meal Logging (Voice)

Record a voice message: "I had chicken and rice" Send it to your bot Bot should transcribe and analyze

Test 5: Update Meal

Send: "update meal" Follow prompts to provide meal ID and new description

Test 6: Get Report

Send: "show my report" Bot should ask for a date and generate summary

Step 6: Customize (Optional) Change Bot Personality

Edit the systemMessage in each Agent node Adjust tone, add emojis, change coaching style

Add Your Branding

Update bot responses to mention your app name Customize welcome messages

Modify Cuisine Focus

Update the nutrition analysis prompts to focus on your target cuisine Current default: African/West African/Nigerian foods

Adjust Nutrition Calculations

Edit the Gemini Vision prompt in "Analyze image" node Modify portion size assumptions or macro ratios

Troubleshooting Bot doesn't respond:

Check webhook is activated in Telegram Trigger node Verify Telegram credential is correct Check execution logs for errors

"Data Table not found" errors:

Ensure all three tables are created with exact names Relink tables in each node

Gemini API errors:

Verify API key is active Check you haven't exceeded free tier limits (15/min) Try regenerating API key

Meals not saving:

Check "Append Meal Data" node is connected Verify User_ID matches between tables

Going Live Once testing is complete:

Click "Active" toggle in top right Share your bot link: t.me/YourBotUsername Monitor the Executions tab for any errors

Target Users & Support This workflow is perfect for:

Fitness coaches wanting client nutrition tracking without monthly SaaS fees Health tech entrepreneurs building MVP before hiring developers No-code builders learning advanced n8n AI patterns Bootcamp graduates needing portfolio projects with real-world complexity Small app studios prototyping before committing to custom development

You'll succeed with this if you:

Can follow detailed instructions Are comfortable with web-based tools Want to learn by doing (the workflow itself is educational) Need a production-ready solution quickly

You might struggle if:

You've never used n8n before (recommend taking their free course first) You're not comfortable with API concepts (keys, webhooks) You need 100% hand-holding (this assumes basic technical literacy)

Support:

Reach me: https://www.linkedin.com/in/gerald-akhidenor-1ab1a45/ Review Google Gemini documentation for AI customization Telegram Bot API docs for advanced features

Next Steps After Setup

Invite Beta Users - Start with friends/clients to gather feedback Monitor Usage - Watch which features get used most Customize Prompts - Refine AI responses based on real conversations Add Features - Use this as foundation for recipe suggestions, meal planning, etc. Scale Infrastructure - Move to dedicated hosting as user base grows

Estimated total setup time: 1-2 hours for complete beginners, 30 minutes for experienced n8n users.