Overview
When you upload a file to Glossa, it goes through an automated processing pipeline that analyzes the content, identifies requirements, and creates detailed citations linking requirements back to specific sections or timestamps in the source file.
Understanding how file processing works helps you upload the right files, troubleshoot processing issues, and get the best results from Glossa's AI-powered requirements generation.
Processing Pipeline
Step 1: Upload and Validation
When you upload a file:
File is validated for type and size
File is encrypted and stored securely
File status changes to Pending
Step 2: Queue and Processing
Once uploaded:
File enters the processing queue
Status changes to Processing
Glossa's AI analyzes the file content
Requirements are extracted and generated
Citations are created linking to source content
Step 3: Completion
After processing:
Status changes to Ready
Generated requirements appear in Requirements tab
File is available for preview with citation highlights
"Files currently processing" counter decreases
Processing Time
Processing time varies based on several factors:
File Size
Small files (< 5 MB): 2-4 minutes
Medium files (5-50 MB): 4-8 minutes
Large files (50-500 MB): 8-15 minutes
Very large files (500 MB - 1 TB): 15-30+ minutes
File Type
Text documents (PDF, Word): Fastest (2-6 minutes)
Spreadsheets (Excel): Fast (3-7 minutes)
Images with OCR: Moderate (5-10 minutes)
Audio files: Slower (transcription required, 5-15 minutes)
Video files: Slowest (video + audio processing, 10-30 minutes)
Content Complexity
Simple, structured documents: Process faster
Complex formatting: May take longer
System Load
Off-peak times: Faster processing
Peak usage times: May be slower
Multiple large files uploading: Process sequentially
How Different File Types Are Processed
Documents (PDF, Word, PowerPoint)
Processing method:
Text extraction from document structure
Maintains formatting and section organization
Identifies headings, lists, tables
Requirements extraction:
Analyzes paragraphs for requirement-like content
Identifies feature requests, business rules, user needs
Captures process descriptions and workflows
Citations created:
Highlight specific paragraphs or sections
Show relevant portions in yellow highlighting
Link to exact page and section in document
Best results:
Well-structured documents with clear headings
Spreadsheets (Excel, CSV)
Processing method:
Reads cell data, rows, and columns
Understands table structure and headers
Processes formulas and data relationships
Requirements extraction:
Identifies requirement lists in rows
Extracts business rules from data patterns
Captures field definitions and validation rules
Citations created:
Reference specific cells or ranges
Show relevant rows and columns
Maintain table context
Best results:
Data dictionaries and field specifications
Business rules documented in spreadsheets
Audio Files (MP3, WAV, M4A)
Processing method:
Transcribes audio to text using AI
Identifies speaker changes (when possible)
Timestamps all content
Requirements extraction:
Analyzes transcription for requirements
Identifies feature discussions
Captures decisions and action items
Citations created:
Include start and end timestamps
Link to specific moments in audio
Play from cited timestamp when clicked
Best results:
Clear audio quality (minimal background noise)
Single speaker or distinct speakers
Structured discussions (meetings, interviews)
Video Files (MP4, MOV, AVI)
Processing method:
Transcribes audio track
Timestamps all content
Requirements extraction:
Analyzes audio transcription
Citations created:
Include start and end timestamps
Link to specific moments in video
Play from cited timestamp when clicked
Best results:
Clear audio (use microphone, not camera audio)
Structured meetings or demos
Images (JPG, PNG, TIFF)
Processing method:
OCR (Optical Character Recognition) extracts text
Analyzes image structure and layout
Identifies text regions
Requirements extraction:
Analyzes OCR text for requirements
Processes whiteboard photos
Reads handwritten notes (if legible)
Citations created:
Show full image with extracted text highlighted
Reference regions where text was found
Best results:
High-resolution images
Clear, readable text (typed or printed)
Well-lit photos of whiteboards
Scanned documents with good contrast
Requirements Generation
How AI Identifies Requirements
Glossa's AI looks for:
Feature requests: "The system should...", "Users need to..."
Business rules: "When X happens, do Y"
User needs: "As a user, I want..."
Process steps: "First..., then..., finally..."
Constraints: "Must support...", "Cannot exceed..."
Decisions: "We decided to...", "The client wants..."
Requirements Granularity
The level of detail in generated requirements is controlled by the Requirement Detail & Visibility Level in Project Settings:
Broad Strokes:
Fewer, high-level requirements
Each requirement covers broader scope
Good for sales discovery, initial scoping
Balanced:
Moderate number of requirements
Good level of detail for most projects
Default setting, recommended for discovery
Granular:
Many detailed requirements
Each requirement is very specific
Good for technical implementation planning
Important: Detail level for the project is captured at the time of upload. Files processed with "Balanced" cannot be retroactively changed to "Granular"—you must delete and re-upload.
Quality Scoring
Each generated requirement receives an AI quality score:
Assesses clarity, completeness, testability
Identifies vague or ambiguous requirements
Suggests improvements
See the AI Review article for details.
Citation Creation
Document Citations
For text-based files:
Paragraph-level precision: Highlights specific paragraphs
Yellow highlighting: Shows exact content that supports requirement
Multiple paragraphs: Citations can span multiple paragraphs if needed
Preview opens to citation: Clicking citation jumps to highlighted section
Audio/Video Citations
For recordings:
Timestamp ranges: Shows start and end time (e.g., 15:30 - 17:45)
Playback from timestamp: Clicking citation jumps to that moment
Transcript excerpts: Shows what was said at that timestamp
Speaker identification: May show who said it (when detectable)
Multiple Citations
One requirement can have citations from multiple sources:
Mentioned in a meeting AND an email
Discussed in multiple meetings
Referenced in multiple documents
All citations appear in the Reference Data tab
Processing Status and Indicators
Status Values
Pending:
File uploaded successfully
Waiting in queue to begin processing
Usually brief (< 1 minute)
Processing:
AI is actively analyzing the file
Requirements being generated
Citations being created
Duration varies by file (see Processing Time section)
Ready:
Processing complete
Requirements have been generated
File is fully searchable and citable
Error:
Processing failed
Detailed error message explains why
File needs attention (see Troubleshooting section)
Processing Counter
At the bottom of the Files tab:
Files currently processing: [number]
This shows:
How many files are actively being processed
Updates in real-time as files complete
When the indicator is not displayed, all processing is complete
What's NOT Processed
Glossa does NOT generate requirements from:
Administrative content:
Meeting scheduling emails
Calendar invites
Out-of-office messages
Pleasantries and greetings
Irrelevant content:
Invoices and receipts
Blank templates
Marketing materials (unless about requirements)
Legal boilerplate
Low-quality input:
Severely corrupted files
Completely illegible handwriting
Unintelligible audio
Files with no extractable content
If a file contains no requirements, processing will complete successfully but no requirements will be generated.
Troubleshooting
File Stuck in Pending
If a file stays "Pending" for more than 5 minutes:
Possible causes:
High system load
Processing queue backed up
File validation issue
Solutions:
Wait 10-15 minutes and refresh
Check "Files currently processing" counter
If still pending after 30 minutes, contact support
File Stuck in Processing
If a file stays "Processing" longer than expected:
Possible causes:
Very large file (1+ hour video)
Complex file structure
Processing in progress (may just need more time)
Solutions:
Check file size and type—large videos can take 30+ minutes
Wait for processing to complete (don't interrupt)
Refresh page after 20-30 minutes to check status
If processing for multiple hours without progress, contact support
Error Status with Detailed Message
When a file shows Error status, Glossa displays a specific error message explaining what went wrong.
Common error types:
1. "File is corrupted or unreadable"
File is damaged
File format is invalid
File has internal errors
Solutions:
Try opening file on your computer
Re-export from source application
Use a different file format
Try a different copy of the file
2. "File is password-protected"
Document requires password to open
Cannot extract content from encrypted file
Solutions:
Remove password protection
Export unprotected version
Provide decrypted copy
3. "File type not supported"
File extension doesn't match supported types
File is actually a different format than extension suggests
Solutions:
Convert to supported format (PDF, Word, MP4, MP3, etc.)
Check file extension matches actual file type
See Uploading Files for supported types
4. "Processing timeout"
File is too large or complex
Processing took longer than maximum allowed time
Solutions:
Split large file into smaller chunks
Reduce file size (compress video, reduce resolution)
Simplify file structure
Contact support for assistance with very large files
No Requirements Generated
If processing completes but no requirements appear:
This is normal when:
File contains no actionable requirements
Content is too vague or general
File is a template or example
Content is purely administrative
What to check:
Open the file preview in Glossa
Review content—does it actually describe requirements?
Check if content is too high-level
Verify Requirements Granularity Dial setting
If requirements should exist:
Try uploading a more detailed version
Consider manually creating requirements and adding file as citation
Check granularity setting and adjust if needed
Contact support with file details
Low-Quality Results
If generated requirements are poor quality:
Possible causes:
Source content is vague or incomplete
Audio/video quality is poor
Document is poorly structured
Wrong granularity setting
Solutions:
Upload more detailed source material
For meetings, ensure clear audio and structured discussions
For documents, use clear headings and structure
Adjust Requirements Detail & Visibility Level
Review and edit requirements manually
Use AI Review to identify improvements
Citations Not Working
If citations don't link correctly:
Possible causes:
File was deleted after requirement creation
Citation link is broken
Preview not loading
Solutions:
Verify file still exists in Files tab
Try clicking citation again
Refresh the page
If file was deleted, re-upload it
Contact support if citations consistently fail
Best Practices
Optimize File Quality
For documents:
Use clear headings and structure
Write in complete sentences
Use bullet points for lists
Include requirement keywords ("must", "should", "will")
For audio/video:
Use good quality microphone
Minimize background noise
Speak clearly and at moderate pace
Record in quiet environment
For images:
Use high resolution (300+ DPI for scans)
Ensure good lighting and contrast
Keep text horizontal and readable
Avoid shadows and glare
Upload Structured Content
Files with clear structure process better:
Meeting agendas with discussion topics
Documents with numbered requirements
Spreadsheets with requirement tables
Organized notes with headings
Set Granularity Before Upload
Decide what level of detail you need
Set Requirements Granularity Dial
Upload files
Avoid needing to delete and re-upload
Monitor Processing
After uploading:
Check "Files currently processing" counter
Refresh periodically for large files
Review generated requirements promptly
Address any errors immediately
Handle Errors Quickly
When files show Error status:
Read the detailed error message
Fix the underlying issue
Delete the error file
Re-upload corrected version
Don't leave error files in your project—they clutter the Files tab and can be confusing.
Performance Tips
Batch Similar Files
Upload similar files together:
All meeting recordings from one week
All discovery documents for one feature
Related files process more efficiently
Avoid Very Large Batches
Instead of uploading 50 files at once:
Upload 10-15 at a time
Let batch complete before uploading more
Easier to monitor and troubleshoot
Use Appropriate File Formats
Choose the best format for your content:
PDFs for finalized documents
Word for editable documents
MP3 for audio-only (smaller than video)
MP4 for video with screen shares
Images for whiteboards and handwritten notes
Compress When Possible
For large files:
Compress videos before uploading
Reduce image resolution if very high
Remove unnecessary content from documents
Balance quality vs. file size
Security & Privacy
Processing Location
Files are processed using AI services
Processing follows Glossa's Data Processing Agreement (see glossapro.ai/dpa)
Content is encrypted in transit and at rest
Data Retention
Processed files stored permanently in project
AI processing data not retained after requirements generation
Files deleted within 60 days of account termination
Compliance
See glossapro.ai/dpa for Data Processing Agreement
See glossapro.ai/security-policy for security details
GDPR and CCPA compliant