> ## Documentation Index > Fetch the complete documentation index at: https://docs.lindy.ai/llms.txt > Use this file to discover all available pages before exploring further. # Knowledge Base
## Quick Overview Knowledge bases let your agents access information from multiple sources like files, websites, and cloud storage. When your agent needs information, it searches through all your content using both semantic understanding and keyword matching to find relevant answers. This is perfect for customer support bots that need to reference documentation, or any workflow where your agent needs to pull information from a large collection of content. For small datasets (single Google Doc or Sheet), use direct actions like Get Document or Get Spreadsheet for better performance. ## How Search Works ### Semantic Search Matches meaning rather than exact words. Understands context and finds related concepts. * Query: "dark colored jeans" * Finds: "black and navy denim pants" ### Keyword Search Matches exact words in your query. Useful for precise matching of specific terms. * Query: "invoice #12345" * Finds: Exact invoice number ## Adding a Knowledge Base Click the plus button in your workflow and select "Search knowledge base". IMAGE NEEDED: Screenshot showing the Add Step menu with 'Search knowledge base' option highlighted ### Selecting Content Sources IMAGE NEEDED: Screenshot of the content source selection modal showing Files, Text, Website, Google Drive, OneDrive, Dropbox, Notion, and Freshdesk options Choose from multiple content types: **Files** * Supports: PDF, XLSX, CSV, DOCX, TXT, HTML * Maximum: 20MB per file * Audio files supported **Text** * Free-form text input * Ideal for FAQs and reference content * No formatting requirements **Websites** * Single page or full site crawling * Selective page removal * Automatic link following **Cloud Storage** * Google Drive, OneDrive, Dropbox * Notion databases * Freshdesk tickets ## Configuring Search IMAGE NEEDED: Screenshot of Search Knowledge Base configuration showing Query field, Max Results, and Search Fuzziness settings ### Query Field Instructions for what to search in the knowledge base. **Mode options:** * **Auto** — Your agent determines the optimal query * **Prompt AI** - Custom search instructions * **Set Manually** - Use exact search query (no AI) Example: `Search for the answer to the user's question in the email` ### Max Results Number of results returned per query. * Default: 4 results * Maximum: 10 results * Increase for frequently mentioned topics ### Search Fuzziness Balance between semantic and keyword search (0-100 scale). * **0** - Pure keyword matching * **100** - Pure semantic search (default) * Lower values emphasize exact matches Keep fuzziness at 100 to avoid the 1,500 file limit. Lower values restrict search to first 1,500 files only. ## Managing Your Knowledge Base Your knowledge base stays current through automatic and manual updates: * **Automatic updates**: Content refreshes every 24 hours automatically * **Manual updates**: For immediate updates, use Resync Knowledge Base actions to update the entire knowledge base or specific entries when needed IMAGE NEEDED: Screenshot showing Resync Knowledge Base actions in the action menu ## Best Practices * **Organize content** - Group similar information and remove duplicates * **Test queries** - Try different searches and adjust fuzziness for accuracy * **Update regularly** - Keep content current with manual resyncs when needed * **Start simple** - Begin with fewer sources and expand as needed ## Common Use Cases * **Content**: FAQs, product docs, support articles * **Settings**: High semantic search (100) * **Results**: 4-6 max results * **Content**: Templates, documentation, policies * **Settings**: Balanced search (70-80) * **Results**: 4 max results * **Content**: Transcripts, project docs, resources * **Settings**: High semantic search (100) * **Results**: 6-8 max results * **Content**: Mixed files, cloud storage, websites * **Settings**: Flexible based on content * **Results**: 10 max results ## Troubleshooting * Lower search fuzziness for more exact matching * Organize content into clearer categories * Use more specific search queries * Check sync status of your sources * Try exact keyword search (fuzziness near 0) * Verify content is in supported formats * Reduce number of content sources * Use direct actions for single documents * Lower max results if not needed Pro users have 20M character limit. Monitor usage to avoid hitting storage limits. ## Next Steps Enhance your knowledge base workflows with these concepts: Combine knowledge search with other actions Route workflows based on search results Store important findings for future use Let agents intelligently use knowledge bases