Uploading Your Data

Flow Myna's AI-powered wizard makes data upload simple—just bring your CSV files and let our intelligent agents automatically understand your data structure and create the necessary transformations. No data preparation or technical expertise required.

Overview of the Upload Process

Flow Myna uses a 5-step wizard that combines AI automation with human review to transform your raw data into process-ready format:

1. Process Selection - Choose your process type (1 minute)
2. File Upload - Upload your CSV files (1-2 minutes)
3. Auto-Mapping - AI analyzes and maps your data (2-3 minutes)
4. Check Mapping - Review and approve the AI's work (2-5 minutes)
5. Transform Data - Complete the import (1-2 minutes)

Total time: 8-15 minutes from upload to analysis-ready data

The key advantage: Our AI agents do the heavy lifting of understanding your data structure, identifying objects and events, and creating the right relationships. You just verify their work and make any needed adjustments.

CSV Requirements

Flow Myna works with standard CSV files. No special formatting is required, but following these guidelines ensures the best results:

Basic Requirements

File Format

• Standard CSV format (comma-separated values)
• UTF-8 encoding recommended
• Headers in first row
• One row per record

Required Columns At minimum, your data should include:

• Timestamps: When events occurred (date and/or datetime)
• Activities/Events: What happened
• Case/Object identifiers: What the event relates to

Recommended Columns For richer analysis, include:

• Multiple object types (customer ID, product ID, etc.)
• Event attributes (amount, status, category, etc.)
• Actor information (who performed the activity)
• Any relevant metadata

Example: Loan Application CSV

Here's what a typical loan application CSV might look like:

loan_id,customer_id,timestamp,activity,amount,risk_level,officer_name,notes
L-001,C-12345,2024-01-15 09:00:00,Application Submitted,50000,Medium,J.Smith,Online application
L-001,C-12345,2024-01-15 10:30:00,Credit Check Complete,50000,Medium,System,Score: 720
L-001,C-12345,2024-01-16 14:00:00,Risk Assessment,50000,Medium,M.Johnson,Requires review
L-001,C-12345,2024-01-17 11:00:00,Manager Approval,50000,Medium,S.Williams,Approved
L-001,C-12345,2024-01-17 14:30:00,Loan Disbursed,50000,Medium,System,Transferred to account
L-002,C-67890,2024-01-15 10:00:00,Application Submitted,25000,Low,J.Smith,Branch application
L-002,C-67890,2024-01-15 11:00:00,Credit Check Complete,25000,Low,System,Score: 780
L-002,C-67890,2024-01-15 15:00:00,Auto Approval,25000,Low,System,Meets criteria
L-002,C-67890,2024-01-16 09:00:00,Loan Disbursed,25000,Low,System,Transferred to account

What the AI will discover:

• Objects: Loans (L-001, L-002), Customers (C-12345, C-67890)
• Events: Application Submitted, Credit Check Complete, Risk Assessment, etc.
• Relationships: Each loan belongs to a customer
• Attributes: Amount, risk level, officer name for context

The 5-Step AI Wizard

Let's walk through each step of the upload process in detail:

Step 1: Process Selection

What happens: Choose the type of business process you're analyzing.

Your actions:

1. Navigate to Data Hub in your workspace

2. Click "Create New Dataset"

3. Select process type:

   • Order Fulfillment: E-commerce, logistics, shipping processes
   • Customer Support: Tickets, cases, support interactions
   • Loan Processing: Applications, approvals, underwriting
   • Invoice Processing: Accounts payable/receivable workflows
   • Custom Process: Any other business process

4. Enter dataset name (e.g., "Q1 2024 Loan Applications")

5. Add description (optional but recommended)

6. Click "Continue"

Why this matters: The process type helps the AI understand your data context. Don't worry too much about picking the "perfect" type—the AI adapts to your actual data structure regardless.

Step 2: File Upload

What happens: Upload your CSV files containing process data.

Your actions:

1. Drag and drop CSV files into the upload area
   • Or click to browse and select files
2. You can upload multiple files if your data is split
3. Files are validated for basic format
4. Wait for upload to complete (usually seconds)
5. Click "Start Analysis" to proceed

File upload tips:

• Start with one file if you're new to Flow Myna
• You can add more files later with "Add to Dataset"
• Large files (100MB+) may take a minute to upload
• Make sure you have a stable internet connection

Step 3: Auto-Mapping (AI Analysis)

What happens: The AI agent pipeline analyzes your data automatically. This is where the magic happens.

What the AI does:

1. Reader Agent (30-60 seconds)

• Reads your CSV file structure
• Identifies columns and data types
• Extracts sample data for analysis
• Detects date formats and encodings

2. Mapper Agent (60-90 seconds)

• Uses LLM to understand your data semantics
• Identifies objects (loans, customers, etc.)
• Recognizes events (activities that happened)
• Creates object-event relationships
• Infers object-to-object relationships

Your actions:

• Just wait and watch the progress
• You'll see status updates as each agent completes
• Total time: 2-3 minutes typically

What if analysis fails? Rare, but if it happens:

• Check that your CSV has headers
• Verify it has timestamp/date columns
• Ensure it has activity/event information
• Contact support if you need help

Step 4: Check Mapping

What happens: Review the AI's work and make adjustments if needed.

This is the most important step—you verify that the AI correctly understood your data structure.

What you'll see:

Object Types The AI identifies different types of objects in your data:

• Loan Applications (from loan_id column)
• Customers (from customer_id column)
• Officers (from officer_name column)

Event Types The AI discovers all activities that occur:

• Application Submitted
• Credit Check Complete
• Risk Assessment
• Manager Approval
• Loan Disbursed

Object-Event Relationships The AI maps which events relate to which objects:

• Each loan application goes through multiple events
• Each event may touch multiple objects

Mapping Visualization You'll see a visual diagram showing:

• Source CSV columns on the left
• Object and event types in the middle
• Relationship arrows showing connections

Your actions:

If mapping looks good:

1. Review the discovered objects and events
2. Verify the relationships make sense
3. Check that key columns are correctly identified
4. Click "Approve and Continue"

If adjustments needed:

1. Click on any object or event to edit
2. Rename if needed (e.g., "Loan" → "Loan Application")
3. Add or remove object types
4. Adjust which columns map to which entities
5. Click "Save Changes" then "Approve and Continue"

Common adjustments:

• Renaming objects for clarity ("obj_123" → "Loan")
• Combining similar events ("submit" and "submitted" → "Application Submitted")
• Removing irrelevant columns from mapping
• Adding object relationships the AI missed

Step 5: Transform Data

What happens: The final processing step that transforms your data into the Flow Myna format.

What the AI does:

1. Script Writer Agent

• Generates data transformation scripts based on approved mapping
• Creates code to extract objects, events, and relationships
• Handles data type conversions and cleaning

2. Script Runner Agent

• Executes the transformation scripts safely
• Validates output data quality
• Handles any transformation errors

3. Data Syncer Agent

• Synchronizes processed data to the database
• Creates objects, events, and relationships
• Builds indexes for fast querying

Your actions:

• Watch the progress
• Wait for completion (1-2 minutes typically)
• See success confirmation

When complete, you'll see:

• Total objects created
• Total events created
• Dataset now available for project creation
• Option to create first project immediately

Example: Complete Loan Application Upload

Let's walk through a full example using loan application data:

Your Starting Data

You have a CSV file: loan_applications_q1_2024.csv with 1,000 rows:

loan_id,customer_id,timestamp,activity,amount,risk,officer
L-001,C-12345,2024-01-15 09:00,Application Submitted,50000,Medium,J.Smith
L-001,C-12345,2024-01-15 10:30,Credit Check,50000,Medium,System
# ... 998 more rows

Step 1: Process Selection

• Select "Loan Processing"
• Name: "Q1 2024 Loan Applications"
• Description: "Jan-Mar 2024 loan application and approval data"

Step 2: Upload

• Drag and drop loan_applications_q1_2024.csv
• File uploads in 3 seconds (12MB file)
• Click "Start Analysis"

Step 3: AI Analysis (1 min 30 sec)

The AI discovers:

• Reader: 1,000 rows, 7 columns, datetime format detected
• Mapper: Found 1 object type (Loans Applications), 8 event types

Step 4: Review Mapping

You review and see:

• ✅ Loans correctly identified (250 unique loans)
• ✅ Customers correctly identified (180 unique customers)
• ✅ 8 event types properly named
• ✅ Loan→Customer relationship detected
• 🔧 Rename "Credit Check" → "Credit Check Complete"
• Approve mapping

Step 5: Transform (1 min 30 sec)

• Creates 250 loan objects
• Creates 180 customer objects
• Creates 1,000 events
• Links all relationships
• Success!

Result

Your dataset is ready. You can now:

• Create a project to analyze the data
• View objects and events in the workspace
• Start asking the AI co-pilot questions

Total time: 7 minutes from starting upload to analysis-ready data

Troubleshooting Common Issues

CSV File Won't Upload

Issue: File upload fails or gives an error.

Solutions:

• Check file size (should be under 500MB for smooth upload)
• Verify it's a valid CSV format
• Try opening in Excel/Google Sheets to verify it's not corrupted
• Check for special characters in filename
• Ensure stable internet connection

AI Mapping Looks Wrong

Issue: AI didn't correctly identify objects or events.

Solutions:

• In Step 4, manually adjust the mapping
• Rename objects/events for clarity
• Check if your CSV has clear column headers
• Verify timestamp columns are properly formatted
• If very wrong, try uploading a cleaner subset of data first

Missing Timestamps

Issue: AI says it can't find timestamp information.

Solutions:

• Ensure your CSV has a date or datetime column
• Check that dates are in a recognizable format (ISO 8601, MM/DD/YYYY, etc.)
• If dates are split across multiple columns, consider combining them first
• Add a header to the timestamp column if it's missing

Too Many Events Found

Issue: AI creates hundreds of event types from subtle variations.

Solutions:

• In Step 4 mapping review, combine similar events
• Standardize activity names in your source data before upload
• Use event groups to organize related events

Transform Takes Too Long

Issue: Step 5 transformation seems stuck.

Solutions:

• Large datasets (100K+ rows) can take 5-10 minutes
• Check your internet connection hasn't dropped
• Wait at least 10 minutes before assuming it's stuck
• If truly stuck, refresh and try re-uploading
• Contact support if problem persists

Next Steps

Once your data is uploaded and transformed, you're ready to start analyzing:

Create Your First Project

• Creating a Project - Set up a focused analysis view of your data

Explore Your Process

• Process Map - Visualize your process flow
• Process Variants - Analyze different execution paths

Ask Questions

• AI Co-Pilot - Use natural language to explore your data