Skip to main content

How Workspace Files Work

Every workspace has a shared file library. When you upload a document, RelayHub processes it through the RAG (Retrieval-Augmented Generation) pipeline — extracting text, splitting it into chunks, and generating vector embeddings. Once indexed, the AI can search and reference the file’s contents when answering questions in any workspace conversation.
Files uploaded to a workspace are only available within that workspace. They do not appear in your personal file library or in other workspaces.

Supported File Types

RelayHub supports a wide range of document formats:

Documents

PDF, Word (.docx), PowerPoint (.pptx), plain text (.txt), Markdown (.md)

Spreadsheets

Excel (.xlsx, .xls), CSV — with full column-aware querying

Images

PNG, JPG, WEBP — analyzed via AI vision for charts, diagrams, and text extraction

Uploading Files

1

Open the Files tab

Navigate to your workspace and click the Files tab. You will see all currently uploaded files and their processing status.
2

Upload your documents

Click the Upload button or drag and drop files into the upload area. You can upload multiple files at once.
3

Wait for processing

Each file goes through a processing pipeline: text extraction, chunking, and embedding generation. You will see a progress indicator for each file. Small documents process in seconds; large PDFs may take a minute or two.
4

Start asking questions

Once a file shows a Ready status, it is fully indexed. Open any conversation in the workspace and the AI will automatically search the file when relevant.
For spreadsheets, the AI can run natural-language queries against the data (for example, “What were total sales in Q3?”). Upload CSV or Excel files to take advantage of this.

File Processing Pipeline

When a file is uploaded, it passes through several stages:
StageDescription
UploadFile is stored securely in object storage (MinIO).
ExtractionText and tables are extracted from the document. PDFs use OCR when needed.
ChunkingThe extracted text is split into overlapping chunks optimized for retrieval.
EmbeddingEach chunk is converted into a vector embedding and stored in the vector database.
ReadyThe file is fully indexed and available for AI search.
If a file is stuck in a processing state for more than a few minutes, try removing it and re-uploading. If the issue persists, contact your administrator.

Reindexing Files

Occasionally you may want to reindex a file — for example, if the processing pipeline has been upgraded or if the original indexing encountered issues. To reindex a file:
  1. Open the Files tab in your workspace.
  2. Find the file you want to reindex and click the options menu (three dots).
  3. Select Reindex. The file will be re-processed through the full extraction and embedding pipeline.
Reindexing does not require re-uploading the original document. The stored file is used as the source.

Removing Files

Owners, Admins, and Members can remove files from a workspace. Open the Files tab, click the options menu on the target file, and select Remove. This deletes the file from storage and removes all associated embeddings from the vector database.
Removing a file is permanent. The AI will no longer be able to reference its contents in future conversations. Past conversations that cited the file will retain their existing messages, but new queries will not find the removed content.

How Files Affect AI Responses

When you send a message in a workspace conversation, the AI performs a semantic search across all indexed files in the workspace. The most relevant chunks are included as context in the AI’s prompt, allowing it to ground its response in your actual documents. This means:
  • More files, better answers. The more relevant documents you upload, the richer the AI’s context.
  • Workspace scoping matters. The AI only searches files in the current workspace, so keep workspaces focused on a specific topic or project for the best results.
  • Citations are traceable. When the AI references a document, it indicates the source file so you can verify the information.

Next: Workspace Memory

Learn how the AI remembers important facts and detects patterns across your workspace conversations.