File Upload
Kapa provides an integration that enables uploading single files directly in the platform. This integration offers a quick and simple way to add content to your knowledge base without setting up complex integrations or permissions.
Prerequisites
- Files in one of the supported formats
- File size under 32MB per file
Data ingested
When you upload files to Kapa, the following data is ingested:
- Complete text content of the file
- For OpenAPI files, the complete API structure and documentation
- For Word documents, formatted text and structure
- For PDFs, the text, document structure and tabular data (not images)
Supported file formats
The following file types are supported:
| Format | File Extension | Content Type |
|---|---|---|
| Markdown | .md | Documentation, guides, README files |
| Text | .txt | Plain text content |
| OpenAPI | .json, .yaml, .yml | API specifications |
| StackOverflow | .csv | Exported Q&A content |
| Word | .docx | Formatted documentation |
.pdf | Technical documentation, misc. |
Setup
- Go to the Sources tab in the Kapa platform
- Click Add new source
- Enter a name for your source
- Select File Upload as the source type
- Click Choose File and select the file from your computer
- Click Upload to begin the ingestion process
Best practices
- Use descriptive filenames: Clear filenames help identify the content when reviewing sources
- Structure your content: Well-organized documents with headings and sections are easier for Kapa to understand
- Keep files updated: If your content changes, you'll need to manually re-upload updated files
- Combine with other sources: File Upload works best as a supplement to more comprehensive data sources
- PDFs: Documents with an embedded text layer (not scans) are best
- Format markdown properly: Markdown files require proper structure with headings and standard syntax. See How should I format markdown files for AI ingestion?
Troubleshooting
- File not accepted: Verify your file has one of the supported extensions
- Upload errors: Check that your file isn't corrupted and is under the size limit
- Formatting issues: Complex formatting in Word documents may not be perfectly preserved
- PDFs: As PDFs are unstructured documents, data extraction can have imperfect results