Uploading Files

The Document Library is where your organization's permanent files live, get processed by AI, and become searchable knowledge for your team. Files uploaded here are available platform-wide - they appear in the File Manager, can be shared with team members, and can be referenced from any chat conversation using #.

This is different from uploading a file directly in a chat conversation. Chat uploads are scoped to that specific conversation and do not appear in the Document Library. If you want a file to be a lasting part of your organization's knowledge base, upload it here.

Here is how to get files in.

You have two options. Drag files directly onto the upload area in the Document Library, or click the upload button to open your system's file picker. Both methods support selecting multiple files at once.

Select as many files as you need in a single operation. Each file uploads independently, so one failure does not block the rest. A progress indicator shows the status of every file in the batch.

Once upload begins, you will see a real-time progress bar for each file. The status moves through uploading, processing, and ready. You can continue working in the platform while files process in the background.

After a file reaches the server, Butterflai's AI pipeline takes over. The system extracts text content, detects tables, and creates vector embeddings that make the file searchable by the Knowledge Agent. This means every document you upload becomes part of your organization's queryable intelligence.

Excel files (.xlsx) receive special treatment. After upload, Butterflai previews the sheet names in your workbook. You choose which sheets to import -- each selected sheet is converted into its own CSV for structured data processing. This gives you control over exactly what data enters the system.

After uploading, you can add a description to any file. This is not just metadata -- the AI uses your description as additional context when answering questions. A clear, specific description (e.g., "Q3 2025 marketing spend by channel and region") significantly improves the accuracy of AI responses that reference the file.