File Upload
Upload documents directly to Twig.
Overview
Type
Static (manual upload)
Sync
Manual only (no auto-refresh)
Plan
All plans
Max File Size
50MB per file
Max Files
1,000 per org (Free), 10,000 (Pro), unlimited (Enterprise)
Batch Upload
Via ZIP (max 200MB)
Supported Formats
✅
Text-based preferred, scanned requires OCR
Word
.doc, .docx
N/A
Converted to plain text
PowerPoint
.ppt, .pptx
N/A
Slide text extracted
Text
.txt
N/A
UTF-8 encoding required
Markdown
.md
N/A
Rendered to HTML first
HTML
.html, .htm
N/A
Stripped of tags
Excel
.xls, .xlsx
N/A
Each sheet processed separately
Images
.jpg, .png, .gif
✅
OCR extracts text, accuracy varies
ZIP
.zip
✅
Extracts and processes each file
Unsupported: Password-protected files, encrypted PDFs, corrupted files
Upload Files
Steps
Twig → Data → Add Data Source → Files
Fill form:
Name: e.g., "Product User Manuals"
Description: Optional
Tags: Optional
Click Choose Files or drag-drop into upload area
Select files (multi-select supported)
Wait for upload (progress bar shows %)
Click Save
Processing starts automatically
Expected timeline:
PDF (10 pages): ~30-60 seconds
DOCX (50 pages): ~1-2 minutes
ZIP (20 files): ~3-5 minutes
OCR PDF (100 pages): ~10-15 minutes
ZIP Batch Upload
Create ZIP (include only supported formats):
Max ZIP size: 200MB Max files in ZIP: 1,000
Processing: Each file extracted and processed individually. Status shows "X of Y files processed".
How to Verify
Data → [File Source] → status "Active" (green)
Shows "X files → Y chunks indexed"
Playground → Query about file content → Check citations show filename
Common Mistakes
Symptom: "Unsupported file format" error
Cause: File type not in supported list or corrupted
Fix: Convert to PDF/DOCX, re-upload
Symptom: No text extracted from PDF
Cause: Scanned PDF (image-based) without OCR
Fix:
Verify text is selectable in PDF (not image)
If scanned: Use OCR tool (Adobe Acrobat, online OCR) to convert
Or: Convert to DOCX, re-upload
Symptom: Password-protected file fails
Cause: Encrypted files cannot be processed
Fix: Remove password in source app, re-upload
When This Doesn't Apply
Auto-sync needed: Use dynamic connectors (Google Drive, Confluence) for content that changes frequently
Large file sets: Use Google Drive or SharePoint connectors for 1,000+ files
Using ZIP Files for Batch Upload
ZIP archives allow you to upload multiple files at once, saving time and effort.
Creating a ZIP Archive
On Windows:
Select all files you want to upload
Right-click and choose "Send to" → "Compressed (zipped) folder"
Name your ZIP file
On macOS:
Select all files you want to upload
Right-click and choose "Compress Items"
A ZIP file will be created automatically
On Linux:
Best Practices for ZIP Files
Keep ZIP files under the size limit for your plan
Use clear folder structure inside the ZIP
Include only supported file formats
Avoid nested ZIP files (ZIP containing other ZIPs)
Refresh and Updates
Since the Files connector is static, content updates require manual action:
To Update Files:
Navigate to your data source
Click Edit
Upload the new version of the file
Click Save to reprocess
To Add More Files:
Create a new data source for additional files, or
Update existing data source with a new ZIP containing all files
Best Practices
1. File Organization
Use descriptive file names
Group related files together
Create separate data sources for different topics
Use tags consistently
2. File Preparation
Remove sensitive information before uploading
Ensure text in PDFs is selectable (not images)
Clean up formatting in Word documents
Remove password protection from files
3. Optimize for AI Processing
Use clear headings and structure
Break large documents into smaller files if possible
Include table of contents for long documents
Use consistent terminology
4. File Naming
Good examples:
product-user-guide-v2.pdfapi-documentation-2024.docxtroubleshooting-faq.pdf
Avoid:
document1.pdffinal-FINAL-v3-really-final.docxuntitled.txt
Limitations
File Size Limits
Check your plan for maximum file size
Large files take longer to process
Consider splitting very large documents
Scanned PDFs
Require OCR (Optical Character Recognition)
Processing time is longer
Accuracy depends on scan quality
Unsupported Content
Encrypted or password-protected files
Corrupted files
Proprietary formats without text extraction
Audio files (except transcription files)
Troubleshooting
Upload Failed
Problem: File upload doesn't complete
Solutions:
Check file size against plan limits
Verify file is not corrupted
Try a different browser
Check internet connection stability
Remove special characters from filename
Processing Stuck
Problem: Status shows "PROCESSING" for a long time
Solutions:
Large files can take several minutes
Check the process logs for errors
Contact support if stuck for over 30 minutes
No Content Extracted
Problem: File uploaded but AI can't answer questions
Solutions:
Verify file contains readable text
For PDFs, ensure text is selectable
Check if file format is supported
Try re-uploading the file
Poor Answer Quality
Problem: AI gives incorrect or incomplete answers
Solutions:
Ensure document has clear structure
Add more context files on the same topic
Check if OCR accuracy is low for scanned docs
Use better quality source documents
Examples
Example 1: Product Documentation
Example 2: Training Materials
Example 3: Technical Specifications
Next Steps
After uploading files:
Test your AI agent with relevant questions
Create AI agent personas that use this data
Monitor analytics to see how the data is being used
Add more data sources to expand knowledge coverage
Related Connectors
QnA CSV - Structured question-answer pairs
Data CSV - Tabular data import
Website - Crawl documentation websites
Google Drive - Sync files from cloud storage
Last updated