Knowledge Folders
Knowledge folders let you work with up to 1,000 files in a single collection, extending beyond the 20-file limit for direct attachments. This guide covers setup, usage, and key differences from direct file attachments.Knowledge folders use vector search to find relevant content sections, which enables large document collections but may not consider every document part for each response.
Understanding the Context Window Limitation
Direct File Attachments vs Knowledge Folders
Direct attachments: Up to 20 files sent directly to the model’s context window
Knowledge folders: Up to 1,000 files with vector search selecting relevant sectionsThe 20-file limit exists because of model context windows (the maximum information models can process simultaneously).
Supported File Types
Knowledge folders support most document formats including PDF, DOCX, TXT, and Markdown files. Tabular Data Limitation: XLSX and CSV files cannot be added to knowledge folders. For spreadsheet data analysis, use direct file attachments (up to 20 files) or convert your data to a supported text format.Setting Up Knowledge Folders
1
Access Knowledge Folders
Navigate to Integrations in your Langdock workspace.
2
Create and Populate Your Folder
Choose your upload method:
- Manual upload: Select files directly from your computer
- API integration: Push files programmatically using our REST API
3
Attach to Assistant
- Open your target assistant
- Go to the Knowledge section
- Click Add Action
- Search for your knowledge folder name
- Select Add for the
"Search in [folder name]"
action
Sharing and Permissions
Knowledge folders support the same sharing capabilities as assistants:Share with specific users in your workspace
API Integration
Knowledge Folder API
Rate limit: 50 requests per minute
Authentication: API key required (generated by workspace admins)
Documentation: Complete schema available in our API guides