We use cookies to enhance your experience and measure how the site performs. Choose "Essential Only" to disable analytics. Read our Privacy Policy.

    Odeus Docs

    Knowledge Folders

    Organize and search through collections of documents in your chats using Knowledge Folders. Build a reusable knowledge base that can be shared across your team.

    Knowledge Folders

    Organize and search through collections of documents in your chats using Knowledge Folders. Build a reusable knowledge base that can be shared across your team.

    Knowledge Folders have moved to the Library. You can now find and manage them at Library → Knowledge Folders. The full documentation has been updated there. This page is kept for reference.

    Knowledge Folders let you create searchable collections of documents that can be accessed directly in chat. Instead of uploading the same files repeatedly, you can build organized knowledge bases and reference them with a simple @mention. They are the per-workspace surface of Odeus's Knowledge Hub retrieval system.

    They use embedding search (vector search) to find the most relevant sections from your documents, so you can keep large collections of files and have the AI pull only the parts that answer your question.

    Odeus's retrieval is admin-configurable. Workspace admins can choose the embedding model (for example OpenAI text-embedding-3-small/3-large, Jina, Cohere, or Voyage) and an optional reranker (Cohere, Jina, or Voyage) to tune quality for your content. Documents are chunked at multiple resolutions, and answers come back with citations that point to the source file and page (with bounding-box highlights for PDFs).

    Key features:

    • Can be shared with individuals, groups, or your entire workspace
    • Works in chat via @mentions and in Agents as actions

    Knowledge Folders are perfect for large documentation sets, company policies, product manuals, or any scenario where you need to search across many documents. For smaller document sets where you need the AI to consider the entire content, use direct file attachments instead.

    Creating a Knowledge Folder

    Click **Library** in the left sidebar and find the Knowledge Folders section.
    
    
    
    Click **New Knowledge Folder** and give your folder a descriptive name. You can also add an optional description to help others understand what documents it contains.
    
    
    
    Your Knowledge Folder is now ready for documents. You can start adding files right away or configure sharing settings first.
    

    Adding Documents

    To add documents, upload files directly:

    1. Open your Knowledge Folder
    2. Click Add Files or drag and drop files into the folder
    3. Wait for the files to be processed (indexing takes a few moments)

    Supported formats include PDF (with layout and OCR, including Arabic), DOCX, XLSX, PPTX, CSV, TXT, Markdown, HTML, and images. Check our supported file types for the complete list.

    Spreadsheets (XLSX, CSV) are best worked with as direct file attachments when you need the model to read the full table. Knowledge Folders are tuned for retrieval across many documents rather than whole-table analysis.

    Using Knowledge Folders in Chat

    Once your Knowledge Folder has documents, you can reference it in any chat using the @mention feature.

    In any chat input field, type `@` to open the actions menu.
    
    
    
    Start typing the name of your Knowledge Folder. You'll see it appear in the search results alongside integrations, Agents, and saved prompts.
    
    
    
    Click on your Knowledge Folder or press Enter to select it. The folder name will appear in blue in your chat, confirming it's active. Now ask your question, and the AI will search through the folder's documents to find relevant information.
    

    When you @mention a Knowledge Folder, the AI uses embedding search to find the most relevant sections from your documents. It doesn't send all documents to the model, just the parts most likely to answer your question.

    Example Usage

    Let's say you have a Knowledge Folder called "Product Documentation" with your company's user guides:

    @Product Documentation How do I reset my password?
    

    The AI will search through all documents in the folder, find the sections about password reset, and provide an answer based on that content.

    Sharing Knowledge Folders

    Knowledge Folders support flexible sharing options, making them ideal for team knowledge bases.

    Sharing Options

    Share with specific people in your workspace. Great for department-specific knowledge or project collaborations.
    
    
    
    Share with defined user groups. Efficient for team-wide access without managing individual permissions.
    
    
    
    Make the folder available to everyone in your workspace. Ideal for company-wide resources like policies or product documentation.
    

    Permission Levels

    PermissionEditorUser
    View folder and contentsYesYes
    Search documents in chatYesYes
    Add filesYesNo
    Remove filesYesNo
    Share with othersYesNo
    • Editor: Can add and remove documents, and share with others.
    • User: Can search the folder in chat and Agents.

    The person who created the folder has full control, including the ability to delete the folder. Workspace admins also have full control over all folders.

    How to Share

    1. Open your Knowledge Folder
    2. Click the Share button
    3. Search for users or groups
    4. Select the permission level (Editor or User)
    5. Optionally send a notification with a custom message

    Workspace admins can configure which roles are allowed to share Knowledge Folders with the entire workspace. Check User Management > Roles for these settings.

    Knowledge Folders vs Direct Attachments

    Choosing between Knowledge Folders and direct file attachments depends on your use case:

    • Use Knowledge Folders When — * You have more than 20 documents * You need to share documents across multiple chats * You're building a team knowledge base * You only need relevant sections, not entire documents * Documents are reference material (manuals, policies, FAQs)

    • Use Direct Attachments When — * You have 20 or fewer documents * You need the AI to consider the entire document content * Documents are specific to one conversation * You need the AI to read a full spreadsheet table (XLSX, CSV) * You need maximum accuracy for that specific document

    Pro tip: For the best results, keep direct attachments for documents where every detail matters (contracts, specific reports), and use Knowledge Folders for larger reference libraries where finding relevant sections is the goal.

    Technical Differences

    FeatureDirect AttachmentsKnowledge Folders
    File limit20 files1,000 files
    ProcessingFull document to contextEmbedding search for relevant sections
    Whole-table spreadsheetsBest fitIndexed for retrieval, not whole-table
    SharingPer chat/agentStandalone sharing
    Best forComplete document analysisLarge knowledge bases

    Using Knowledge Folders with Agents

    Knowledge Folders integrate seamlessly with Agents as actions:

    1. Open your Agent configuration
    2. Navigate to the Knowledge section
    3. Click Add Action
    4. Search for your Knowledge Folder name
    5. Add the "Search in [folder name]" action

    Now your Agent can automatically search through the Knowledge Folder when users ask relevant questions.

    Best Practices

    • Organize by topic: Create separate folders for different subjects (HR Policies, Product Docs, Technical Guides)
    • Use descriptive names: Help users understand what each folder contains at a glance
    • Keep folders updated: Remove outdated documents and add new ones regularly
    • Set appropriate permissions: Use the principle of least privilege when sharing
    • Combine with direct attachments: Use both methods together when needed

    Troubleshooting

    * Verify you have at least User access to the folder
    * Check that you're spelling the folder name correctly
    * Try typing more of the name to narrow down search results
    
    
    
    * Ensure your documents are properly processed (check for processing status)
    * Try rephrasing your question with more specific keywords
    * Verify the information exists in the uploaded documents
    * Consider if direct attachments might work better for this use case
    
    
    
    Check our [supported file types](/en/using-odeus/troubleshooting/faq/supported-file-types) for the complete list. For whole-table spreadsheet work, attach the file directly to the chat instead.
    

    Related Resources