Skip to content
Nappai Documentation

FAISS

This component, FAISS, helps Nappai find similar documents within your data using a powerful search technology. Imagine you have a huge library of documents; FAISS helps Nappai quickly locate the most relevant ones based on what you’re looking for.

This component uses FAISS, a very fast and efficient technology developed by Facebook AI Research, to search through your data. FAISS is specialized in finding similar items based on how they’re represented as “vectors” (think of them as mathematical descriptions of your data). This makes searching much faster than traditional methods.

Inputs

  • Index Name: A name to identify your search index. Think of it as a label for your specific search collection. The default is “langflow_index”.
  • Persist Directory: The location where Nappai will save your search index. Nappai will handle this automatically; you usually don’t need to change this.
  • Search Query: This is what you type to search. It’s the question or description of what you’re looking for.
  • Ingest Data: The data you want Nappai to search through. This could be text documents, images, or other data that Nappai can understand.
  • Allow Dangerous Deserialization: This option is usually left as the default (True). It allows Nappai to load saved search indexes more easily. Only change this if you have specific security concerns and understand the implications.
  • Embedding: This setting defines how Nappai represents your data as “vectors” for FAISS to search. Nappai will usually handle this automatically.
  • Number of Results: How many of the most similar documents you want Nappai to return. The default is 4.

Outputs

The component returns a list of documents that are most similar to your search query. These documents will appear in your Nappai dashboard, ready for you to review.

Usage Example

Let’s say you have a collection of customer support tickets. You can use FAISS to quickly find similar tickets when a new one comes in. You would:

  1. Upload your support tickets as “Ingest Data”.
  2. Type your search query (e.g., “problem with login”).
  3. Nappai will use FAISS to find the most similar tickets, showing you the results.

Templates

[List of templates where the component can be seen and its configuration - This section will be populated based on the actual templates used.]

  • VectorStoreInfo: Provides information about the vector store used by FAISS.
  • Self Query Retriever: Helps refine your search queries for better results.
  • [Other Components]: Many other components in Nappai can work with the results from FAISS, such as those for summarizing, categorizing, or sending emails. (A full list is provided in the technical documentation).

Tips and Best Practices

  • For best results, ensure your “Ingest Data” is relevant and well-structured.
  • Experiment with different search queries to refine your results.
  • Consider adjusting the “Number of Results” to control how many documents are returned.

Security Considerations

The “Allow Dangerous Deserialization” setting should be used with caution. Leaving it as the default (True) is generally safe within a controlled Nappai environment. Changing it to False might prevent Nappai from loading previously saved search indexes. Consult Nappai’s security documentation for more details.