Skip to content
Nappai Documentation

Vectara RAG

This component, Vectara RAG, is like a super-powered search assistant within Nappai. Instead of just giving you a list of links, it understands your questions and provides you with direct answers based on the information it finds in your connected data sources. It does this by intelligently searching and summarizing the most relevant information.

Relationship with Vectara

This component uses the Vectara platform to perform powerful searches and generate summaries. Vectara is a sophisticated search engine that understands the meaning of your questions, allowing for more accurate and relevant results.

Inputs

  • Search Query: This is the question you want answered. Type your question here.
  • Hybrid Search Factor: This setting controls how much Vectara relies on simple word matching versus its more advanced understanding of the meaning of words. Leave this at the default setting unless you have a specific reason to change it. Adjusting this might improve results in certain cases.
  • Metadata Filters: This allows you to refine your search by specifying certain characteristics of the data you want to search within. For example, you might only want to search documents from a specific date or project. Leave this blank for a general search.
  • Reranker Type: This selects the method used to order the search results. Leave it at the default setting (“mmr”) unless you have a specific need for a different method.
  • Number of Results to Rerank: This sets how many of the initial search results are considered for re-ordering. The default is usually sufficient.
  • Diversity Bias: This setting (only applicable to the default “mmr” reranker) influences how diverse the results are. A higher value means more varied results. Leave this at the default unless you need more diverse answers.
  • Max Results to Summarize: This determines how many of the best search results are used to create the final answer. A higher number might provide a more comprehensive answer, but it could also take longer.
  • Response Language: Choose the language you want the answer in. “eng” (English) is the default.
  • Prompt Name: This selects the type of summary generated. Leave this at the default setting unless you have a specific need for a different summary style.

Outputs

  • Answer: This is the final answer generated by Vectara RAG, based on your search query and the settings you provided. It’s a concise summary of the relevant information found.

Usage Example

Let’s say you want to know “What were the sales figures for Q3 2023?”. You would enter this into the “Search Query” input. The component would then search your connected data, find the relevant sales information for Q3 2023, and provide a summarized answer in the “Answer” output.

Templates

[List of templates where the component can be seen and its configuration - This section needs to be populated with actual template information]

[Links to other related components and a brief description of each - This section needs to be populated with actual component information and links. For example, components that provide data Vectara RAG can search.]

Tips and Best Practices

  • Be specific in your search queries: The more precise your question, the better the results.
  • Experiment with settings: If you’re not getting the desired results, try adjusting the settings like “Hybrid Search Factor” or “Max Results to Summarize”.
  • Use metadata filters to refine searches: This can significantly improve the accuracy of your results when searching large datasets.

Security Considerations

[If applicable, include appropriate security considerations - This section needs to be populated based on the specific security implications of using Vectara and handling sensitive data within Nappai.]