Skip to content
Nappai Documentation

Self Query Retriever

This component in Nappai helps you find the documents you need quickly and easily. It works by understanding your search request (in plain language) and then using that to search through your stored data. Think of it as a super-powered search engine for your data.

Relationship with Language Models and Vector Stores

This component uses a language model (like a sophisticated AI) to understand what you’re looking for, and a vector store (a special type of database) to efficiently search through your documents. It combines these two technologies to give you highly relevant results.

Inputs

  • Query: This is where you type your search request. Use clear and concise language, just like you would when searching on Google.
  • Vector Store: This is the location where your documents are stored. Nappai will automatically connect to the correct vector store you’ve already set up. You don’t need to worry about this setting unless you’re changing your data storage.
  • Metadata Field Info: This provides extra information about your documents (like author, date, or topic). Nappai will usually automatically fill this in, but you can adjust it if needed. This helps the component refine its search.
  • Document Content Description: This is a brief description of what kind of information is contained in your documents. This helps the component understand the context of your search.
  • LLM: This refers to the language model used for understanding your query. Nappai will select an appropriate model for you. You typically won’t need to change this setting.

Outputs

  • Retrieved Documents: This is a list of documents that match your search. The component will display these documents directly in the dashboard, making it easy to review the results.

Usage Example

Let’s say you have a collection of customer support tickets stored in Nappai. You want to find all tickets related to “billing issues.” You would:

  1. Enter “billing issues” in the Query field.
  2. Nappai will automatically select the correct Vector Store containing your support tickets.
  3. The other input fields will likely be pre-filled by Nappai.
  4. Click “Run”.
  5. The component will return a list of relevant support tickets under the Retrieved Documents output.

Templates

[List of templates where the component is used. This section will be populated based on the actual templates used.]

  • [Component Name 1]: [Brief description of how it relates to the Self Query Retriever, e.g., “This component preprocesses the documents before they are added to the vector store.”]
  • [Component Name 2]: [Brief description]
  • …and so on for all components listed in the provided text. (This section requires expanding with descriptions for each component.) For example:
    • PGVector: PGVector Vector Store with search capabilities: This is where your data is stored for the Self Query Retriever to search.
    • Speech to Text: Transcribe speech to text: You could use this to convert audio files into text before using the Self Query Retriever to search their content.
    • Summarizer: Summarize large bodies of text using AI: You could use this to summarize the retrieved documents after using the Self Query Retriever.

Tips and Best Practices

  • Use clear and specific keywords in your query for better results.
  • If you’re not getting the expected results, try refining your query or checking the metadata information.
  • For very large datasets, consider using more specific metadata to improve search speed and accuracy.

Security Considerations

  • Ensure your vector store and LLM are configured securely according to Nappai’s security guidelines. Access to sensitive data should be appropriately restricted.