Upload Trigger

The Upload trigger allows you to trigger flows when documents are uploaded. Perfect for processing invoices, extracting data from forms, analyzing images, converting documents, and any workflow that starts with a file upload. When you add an Upload trigger, you can directly upload files through the trigger interface to test and run your flow with real documents.

---

​

How It Works

1. Add an Upload trigger to your flow
2. Click on “Upload” in the left sidebar
3. Choose whether to trigger the Draft (test version) or Published (live version) of your flow
4. Upload your file using the form (drag and drop or click to upload)
5. Your flow triggers automatically with the uploaded file
6. The file becomes immediately available to your flow agents through the trigger data
7. Files are automatically cleaned up after 2 weeks by default

---

​

Configuration

​

Direct Upload Interface

When you add the Upload trigger to your flow:

1. Click on “Upload” in the left sidebar
2. Select your trigger mode:
   • Draft - Triggers the test version of your flow (for testing before publishing)
   • Published - Triggers the live version of your flow
3. Upload your file:
   • Drag and drop a file into the upload area
   • Click to upload and browse for a file
4. The flow triggers immediately with your uploaded file

Use cases:

• Draft mode: Test your flow with sample documents before going live
• Published mode: Process production documents through your live workflow

​

File Storage

Uploaded files are stored in your account’s microstrate-flow-object-trigger bucket for 2 weeks by default. After the storage period, files are automatically cleaned up to save space.

​

Accepted File Types

Documents:

• PDF (.pdf)
• Microsoft Word (.docx, .pptx)
• Text files (.txt, .md, .rtf)
• CSV (.csv)
• Excel (.xlsx)
• LibreOffice

Images:

• JPEG (.jpg, .jpeg)
• PNG (.png)
• GIF (.gif)

Other:

• JSON (.json)
• XML (.xml)
• ZIP archives (.zip, .gzip, .tar)

​

File Size Limits

Default limits by plan:

• Free: 10 MB per file
• Starter: 25 MB per file
• Pro: 50 MB per file
• Team: 100 MB per file
• Enterprise: Custom limits

Files exceeding the limit will be rejected with an error message.

​

Multiple Files

Single file mode (default): One file per upload request. COMING SOON: Multiple files mode: Allow multiple files in single upload. Useful for batch processing. Each file will be processed by the flow.

---

​

Testing Your Flow

​

Using Draft Mode

Always test your flow in Draft mode first:

1. Open your Upload trigger
2. Select “Draft” as the trigger mode
3. Upload a test file (invoice, form, image, etc.)
4. Monitor the execution in your flow’s draft environment
5. Verify all steps work correctly with real file data
6. Check that agents extract the correct information
7. Ensure integrations and API calls function as expected

Benefits:

• Test with real files without affecting production
• Iterate quickly on your flow design
• Verify file processing logic works correctly
• Safely test with sample documents before going live

​

Using Published Mode

Once you’ve tested in draft mode and published your flow:

1. Select “Published” as the trigger mode
2. Upload production files
3. Flow processes files through your live workflow
4. Monitor executions in your published flow logs

---

​

Working with Uploaded Files

​

File Information Available to Agents

When a file is uploaded, your agents automatically have access to:

• File name - The original name of the uploaded file
• File type - The MIME type (e.g., application/pdf, image/jpeg)
• File size - Size in bytes
• File content - The actual content of the file for processing
• Upload timestamp - When the file was uploaded

​

Using Agents to Process Files

Your AI agents can directly process uploaded files. Simply instruct your agent to:

• Extract data from invoices, receipts, or forms
• Analyze content from contracts or documents
• Recognize text from images using OCR
• Parse data from CSV or Excel files
• Summarize long documents
• Categorize documents by type or content
• Validate that required information is present

Example agent instructions: For invoice processing: “Extract the invoice number, date, total amount, vendor name, and all line items from the uploaded PDF invoice.” For resume screening: “Read the uploaded resume and extract the candidate’s name, email, phone, years of experience, education, and key skills. Then evaluate if they meet our minimum requirements of 3+ years experience in software development.” For form data extraction: “Extract all filled-in fields from the uploaded form image, including name, address, date, and signature status.”

---

​

Use Cases

​

Invoice Processing

Scenario: Automatically process uploaded invoices Flow:

1. Trigger: Upload (accept .pdf)
2. Agent: Extract invoice data
   • Instruct agent to extract: invoice number, date, amount, line items, vendor info
3. Condition: Check if amount is over approval threshold
4. Agent integration request: Create entry in accounting system
5. Agent integration request: Send notification to accounting team

Testing in Draft mode:

1. Upload a sample invoice
2. Verify agent extracts all required fields correctly
3. Check condition logic routes properly
4. Confirm notifications are sent
5. Publish flow and switch to Published mode for production invoices

Benefits: Eliminate manual data entry, faster processing, reduce errors, automatic approval workflow.

---

​

Resume Screening

Scenario: Screen uploaded resumes for job applications Flow:

1. Trigger: Upload (accept .pdf, .docx)
2. Agent: Extract candidate information and evaluate qualifications
   • Instruct agent to extract: name, contact, experience, education, skills
   • Evaluate against job requirements
3. Condition: Does candidate meet minimum qualifications?
   • Yes → Add to applicant tracking system
   • No → Send polite rejection email
4. Agent integration request: Notify hiring manager of qualified candidates

Testing in Draft mode:

1. Upload sample resumes with varying qualifications
2. Verify agent extracts information correctly
3. Test that qualification logic works properly
4. Check both acceptance and rejection paths
5. Validate ATS integration works

Benefits: Automatic screening, consistent evaluation, faster response to candidates.

---

​

Form Data Extraction

Scenario: Extract data from scanned forms Flow:

1. Trigger: Upload (accept .pdf, .jpg, .png)
2. Agent: Perform OCR and extract structured data
   • Instruct agent to identify and extract all form fields
3. Agent: Validate that required fields are filled
4. Condition: Is data complete and valid?
   • Yes → Save to database
   • No → Flag for manual verification
5. Agent integration request: Send confirmation or request additional info

Testing in Draft mode:

1. Upload sample scanned forms with varying quality
2. Test OCR accuracy with different image types
3. Verify data extraction maps to correct fields
4. Check validation catches incomplete forms
5. Test both complete and incomplete form paths

Benefits: Digitize paper forms, reduce manual typing, improve accuracy.

---

​

Document Analysis

Scenario: Analyze contract documents Flow:

1. Trigger: Upload (accept .pdf, .docx)
2. Agent: Read and analyze contract
   • Instruct agent to extract: key terms, dates, obligations
   • Identify non-standard clauses
   • Flag potential risks
3. Agent: Generate summary report
4. Agent integration request: Save analysis to document management system
5. Agent integration request: Email report to legal team

Testing in Draft mode:

1. Upload sample contracts
2. Verify agent identifies key contract terms
3. Check risk flagging works for non-standard clauses
4. Test report generation format and content
5. Validate document management system integration

Benefits: Faster contract review, consistent analysis, risk identification.

---

​

Image Processing

Scenario: Process and optimize uploaded images Flow:

1. Trigger: Upload (accept .jpg, .png)
2. Function: Validate image meets requirements (dimensions, format)
3. Function: Resize image for web use
4. Function: Generate thumbnails
5. Agent integration request: Upload to CDN
6. Agent integration request: Save URLs to database

Testing in Draft mode:

1. Upload test images of different sizes
2. Verify resizing maintains quality
3. Check thumbnail generation works
4. Test CDN upload succeeds
5. Validate database receives correct URLs

Benefits: Automatic optimization, consistent image formats, faster page loads.

---

​

Receipt Processing

Scenario: Extract data from expense receipts Flow:

1. Trigger: Upload (accept .pdf, .jpg, .png)
2. Agent: Extract receipt data and categorize
   • Instruct agent to extract: merchant, date, amount, items, tax
   • Categorize expense type (meals, travel, supplies, etc.)
3. Agent integration request: Create expense record in system
4. Condition: Does amount require manager approval?
   • Yes → Create approval request
   • No → Auto-approve
5. Agent integration request: Notify employee of status

Testing in Draft mode:

1. Upload sample receipts of various formats
2. Test extraction accuracy with different receipt types
3. Verify expense categorization works correctly
4. Check approval threshold logic
5. Test notification delivery

Benefits: Simplified expense reporting, automatic categorization, faster reimbursement.

---

​

Document Conversion

Scenario: Convert documents to different formats Flow:

1. Trigger: Upload (accept .docx, .xlsx, .pptx)
2. Function: Convert to PDF
3. Function: Generate preview images
4. Agent integration request: Upload converted files to storage
5. Agent integration request: Send download links to user

Testing in Draft mode:

1. Upload various document types
2. Verify conversion maintains formatting
3. Check preview image quality
4. Test file storage upload
5. Validate download links work

Benefits: Universal format conversion, automatic processing, secure storage.

---

​

Data Import from Spreadsheets

Scenario: Import data from uploaded spreadsheets Flow:

1. Trigger: Upload (accept .csv, .xlsx)
2. Agent: Parse and validate spreadsheet data
   • Instruct agent to extract all rows and validate format
   • Check for required columns
   • Identify any errors or inconsistencies
3. Condition: Is data valid?
   • Yes → Proceed with import
   • No → Generate error report
4. Agent integration request: Bulk insert to database
5. Agent integration request: Send confirmation or error report

Testing in Draft mode:

1. Upload sample spreadsheets with good and bad data
2. Test parsing with different formats
3. Verify validation catches invalid data
4. Check error reporting is clear
5. Test database insertion with small batches

Benefits: Bulk data import, validation, error handling.

---

​

Best Practices

​

Testing Before Production

Always test in Draft mode:

1. Upload representative sample files
2. Verify each step processes correctly
3. Check error handling works
4. Test with edge cases (corrupted files, wrong formats, etc.)
5. Validate integrations connect properly
6. Only switch to Published mode after thorough testing

​

Instructing Agents Effectively

Be specific in agent instructions:

• Clear extraction requirements: “Extract the invoice number, date, and total amount”
• Validation criteria: “Check that all required fields are filled”
• Conditional logic: “If the amount is over $1,000, flag for approval”
• Error handling: “If the document is unreadable, indicate what information is missing”

Examples of good agent instructions: For invoices: “Extract the following from the invoice: invoice number, invoice date, due date, vendor name, billing address, all line items with descriptions and amounts, subtotal, tax, and total. Format the output as structured data.” For contracts: “Read this contract and identify: contract start date, end date, parties involved, key obligations for each party, payment terms, termination clauses, and any non-standard or unusual provisions. Flag any clauses that deviate from our standard contract template.” For resumes: “Extract candidate information: full name, email, phone, current job title, years of experience, education (degree, school, year), and list of technical skills. Then evaluate: Does candidate have 5+ years of software engineering experience? Does candidate have a bachelor’s degree or equivalent? List any skills matching our requirements: Python, React, AWS.”

​

Using Conditions for Routing

Route files based on content: Use Condition steps after your agent to route based on what was extracted:

• Check if required information is present
• Verify amounts are within acceptable ranges
• Determine if manual review is needed
• Route to different processing paths based on file type or content

Example condition logic: After invoice extraction:

• If amount > $5,000 → flag for approval
• If amount ≤ $5,000 → Auto-approve and process

After resume screening:

• If meets minimum qualifications → Add to ATS and notify hiring manager
• If doesn’t meet qualifications → Send rejection email

​

File Validation

Use agents to validate files: Instruct your agent to check:

• Is the file readable and not corrupted?
• Does it contain the expected type of content?
• Are required fields or information present?
• Is the data in the expected format?

Add a Condition step after the agent to route invalid files to an error handler.

​

Error Handling

Build robust error handling:

• Use Condition steps to check for errors
• Create separate paths for invalid files
• Send clear error notifications
• Errors are automatically logged for review

Example error handling flow:

1. Agent processes file
2. Condition: Did agent successfully extract data?
   • Yes → Continue processing
   • No → Send error notification with file details

​

File Storage Management

Understanding file lifecycle:

• Files are stored for 2 weeks by default
• Automatic cleanup after storage period
• Files are accessible to agents during flow execution
• Extract and save important data within your flow

Best practices:

• Process files promptly after upload
• Don’t rely on files being available after 2 weeks
• Have agents extract and save all needed data during the flow
• Monitor storage usage across your account

---

​

Working with Different File Types

​

PDF Documents

What agents can do:

• Extract all text content
• Identify and extract specific data fields
• Recognize tables and structured data
• Read form fields
• Extract embedded images

Common use cases:

• Invoice processing
• Contract analysis
• Form extraction
• Document conversion

Agent instruction tips:

• Be specific about what data to extract
• Ask agent to identify document structure
• Request structured output format

---

​

Images

What agents can do:

• Perform OCR to extract text
• Identify objects and scenes
• Read handwritten text
• Extract data from forms or documents
• Analyze image content

Common use cases:

• Receipt processing
• Form data extraction
• Business card scanning
• Quality inspection

Agent instruction tips:

• Specify what text or data to extract
• Ask agent to describe image quality
• Request validation of extracted data

---

​

Spreadsheets (Excel and CSV)

What agents can do:

• Read and parse data from rows and columns
• Identify column headers
• Extract specific data ranges
• Validate data formats
• Summarize data

Common use cases:

• Data import
• Inventory updates
• Financial data processing
• Customer list uploads

Agent instruction tips:

• Specify which columns are important
• Define expected data format
• Ask agent to validate data quality
• Request identification of errors

---

​

Word Documents

What agents can do:

• Extract all text content
• Identify document structure
• Parse tables and lists
• Extract specific sections
• Summarize content

Common use cases:

• Resume parsing
• Report analysis
• Template processing
• Content extraction

Agent instruction tips:

• Ask agent to identify document sections
• Specify what information to extract
• Request structured output

---

​

JSON and XML

What agents can do:

• Parse structured data
• Extract specific fields
• Validate data structure
• Transform data format

Common use cases:

• Data integration
• Configuration import
• API data processing

Agent instruction tips:

• Specify which fields to extract
• Define expected data structure
• Request validation of required fields

---

​

Troubleshooting

​

Upload Not Triggering Flow

Check these items:

1. Flow status: Ensure flow is saved
2. Trigger mode: Verify you selected Draft or Published
3. File size: Check file is within plan limits
4. File type: Ensure file extension is allowed
5. Execution logs: Review flow history for errors

​

Agent Not Processing File Correctly

Common solutions:

1. Test in Draft mode: Upload sample files and check agent output
2. Clarify instructions: Make agent instructions more specific
3. Check file format: Ensure file type is supported
4. File quality: Try with higher quality images or clearer PDFs
5. Review execution: Check agent step output in flow logs

​

File Not Readable

Possible causes:

• File is corrupted or damaged
• File format not supported
• File is password protected
• Image quality too low for OCR
• File encoding issues

Solutions:

• Request file be re-uploaded
• Try different file format
• Remove password protection
• Scan at higher resolution
• Save with standard encoding

​

Processing Too Slow

Optimization tips:

1. Test file size: Use appropriately sized files
2. Simplify agent tasks: Break complex tasks into steps
3. Optimize flow: Remove unnecessary steps
4. Check integrations: Ensure external APIs respond quickly

​

Data Extraction Inaccurate

Improvement strategies:

1. Refine agent instructions: Be more specific about what to extract
2. Add validation: Use agents to double-check extracted data
3. Test with samples: Try various file formats in Draft mode
4. Use structured prompts: Ask agent to format output consistently
5. Add error checking: Use Conditions to verify data quality

---

​

Monitoring and Debugging

​

Using Execution Logs

Monitor your flow executions:

1. Go to your flow’s execution history
2. Find the execution triggered by your file upload
3. Review each step’s output
4. Check for errors or unexpected behavior
5. View file information in trigger data
6. Review agent responses and extracted data

​

Testing Different Scenarios

In Draft mode, test:

• Valid files that should succeed
• Invalid file types that should be rejected
• Corrupted files to test error handling
• Files at or near size limits
• Files with edge cases (empty, malformed, etc.)
• Files with missing or incomplete data

​

Common Issues and Solutions

Issue: Agent not extracting data correctly Solution:

1. Upload sample in Draft mode
2. Review agent output in execution logs
3. Adjust agent instructions to be more specific
4. Test with different file samples
5. Add validation steps to check data quality

Issue: Flow timing out on large files Solution:

1. Test with smaller files first
2. Break processing into smaller steps
3. Consider file size limits for your use case
4. Optimize agent prompts to be more efficient

Issue: Integrations not receiving correct data Solution:

1. Verify data extraction in Draft mode
2. Check agent output format matches integration requirements
3. Add a Map step to transform data if needed
4. Test integration separately
5. Review execution logs for errors

---

​

Advanced Flow Building

​

Conditional Processing by File Type

Use Condition steps to route based on file type: Flow example:

1. Upload trigger receives file
2. Condition: Check file extension
   • If PDF → Send to PDF processing agent
   • If Image → Send to OCR agent
   • If Spreadsheet → Send to data import agent
   • Otherwise → Send error notification

​

Sequential Processing Steps

Break complex tasks into multiple agent steps: Example for contract analysis:

1. Agent 1: Extract basic contract information (parties, dates, terms)
2. Agent 2: Analyze risk clauses using extracted information
3. Agent 3: Generate summary report combining both analyses
4. Agent integration request: Save results and notify legal team

​

Validation Workflows

Add validation before main processing: Example flow:

1. Upload trigger
2. Agent: Quick validation check - is file readable? Does it contain expected content?
3. Condition: Is file valid?
   • Yes → Proceed to main processing
   • No → Send error notification and stop
4. Main processing continues…

​

Batch Processing Multiple Files

If multiple files mode is enabled:

1. Upload trigger receives multiple files
2. Loop through files (using flow logic)
3. Process each file with agents
4. Aggregate results
5. Send combined report

​

Human-in-the-Loop Approval

Add approval steps for sensitive operations: Example flow:

1. Upload invoice
2. Agent extracts data
3. Condition: Is amount over $10,000?
   • Yes → Send approval request to manager, wait for response
   • No → Auto-process
4. Continue based on approval decision

---

​

Next Steps