Learning Lindy

Knowledge Base

A comprehensive guide to leveraging Lindy's intelligent data retrieval system.

A Knowledge Base is a repository of information that enables Lindy to reference and retrieve data from multiple sources. Adding a Knowledge Base to a workflow is helpful when Lindy needs to respond to user queries based on proprietary data sources, reference specific information for certain actions, or perform other activities where access to up-to-date data sources is important for task completion.

Using a hybrid semantic and keyword search approach, Lindy can intelligently query your content to find relevant information.

Note: For small datasets (single Google Doc / Sheet), use direct actions like Get Document or Get Spreadsheet instead of a Knowledge Base for optimal performance.

How Lindy Finds Info

Semantic Search

Matches meaning rather than exact words
Understands context and related concepts
Example: Query "dark colored jeans" matches content about "black and navy jeans"

Keyword Search

Matches exact words in query
Useful for precise matching
Configurable via search fuzziness

Types Of Content Sources

You can add one or a combination of content sources to your knowledge base. You can modify these at any time.

Files

Supports: PDF, XLSX, CSV, DOCX, TXT, HTML
Maximum size: 20MB per file
Audio files supported

Text

Free-form text input
Ideal for FAQs and direct content
No formatting requirements

Websites

Single-page or full-site indexing
Crawl option for linked pages
Selective page removal supported

Cloud Storage

Google Drive
OneDrive
Dropbox
Notion
Intercom
Freshdesk
Zendesk

How To Add A Knowledge Base

At any point in your workflow, you can click “Add step” or the blue plus button, and click “Search knowledge base” to add a knowledge base step.

Knowledge Base Configuration

Model

Choose the LLM used for knowledge base queries

Query

The query field is where you can add instructions (a prompt) on what to search for in the knowledge base.

Mode options:
- Set Manually: use one exact search query for every query (no AI)
- Auto: Lindy determines optimal query
- Prompt AI: Custom search instructions

Example prompt: Search for the answer to the user's question in the email

Max Results

The max results field allows you to customize how many results are returned per query.

Default: 4 results
Maximum: 10 results
If the answer to your question is mentioned many times throughout your Knowledge Base, you may want to increase this up to 10 (default value is 4).

Search Fuzziness

The search fuzziness field allows you to configure what percentage of the query uses a pure keyword search versus semantic search.

The Search Fuzziness is on a 0-100 scale.
0: Pure keyword matching
100: Pure semantic search (default)
Lower values increase keyword emphasis

Knowledge Base Limitations

Pro Users have access to a maximum of 20M characters for storage.

If you adjust search fuzziness below 100, searches are limited to the first 1,500 files. Keeping the default fuzziness of 100 removes this limitation.

Updating Your Knowledge Base

The Knowledge Base automatically refreshes every 24 hours to ensure your content stays up-to-date. However, if you need to push immediate updates—for example, when you've added new product documentation or updated pricing—you can trigger a refresh using two available actions:

Resync Knowledge Base
Resync Knowledge Base Entry

The Resync Knowledge Base action updates your entire knowledge base, while the Resync Knowledge Base Entry action lets you refresh specific entries. Add these actions to your Lindy whenever you need real-time updates between automatic syncs.

Example Knowledge Base Lindies

Use Case	Content Types	Search Settings
Customer Support Chatbot	FAQ documents Product documentation Support articles	Higher semantic search
Email Ticket Responder	Response templates Product documentation Pricing and policies	Balanced search approach
Chat With Your Meetings	Meeting transcripts Project documentation Resource links	High semantic search
Query Your Files	Multiple file types Cloud storage Website content	Flexible search ratios High result limit

Knowledge Base Best Practices

Group similar content and remove duplicates
Test queries and adjust fuzziness for precision
Monitor results and update content regularly

Troubleshooting

For irrelevant results:some text
- Adjust search fuzziness
- Refine content organization
For missing information:some text
- Check sync status
- Use an exact keyword search (dumb the search down and use an exact FAQ style of search query and result)

Guides

Advanced Email Automation: 102