Vectara Vector Store

​

Node Details

• Name: Vectara_VectorStores
• Type: Vectara
• Version: 2.0
• Category: Vector Stores

​

Base Classes

• Vectara
• VectorStoreRetriever
• BaseRetriever

​

Credentials

• Type: vectaraApi
• Required Parameters:
  • API Key
  • Customer ID
  • Corpus ID

​

Input Parameters

​

Main Inputs

1. Document (optional, list)
   • Type: Document
   • Description: List of documents to be indexed in Vectara.
2. File (optional)
   • Type: file
   • Description: File to upload to Vectara. Supports various file types as per Vectara documentation.

​

Additional Parameters

1. Metadata Filter (optional)
   • Type: string
   • Description: Filter to apply to Vectara metadata.
2. Sentences Before (optional)
   • Type: number
   • Default: 2
   • Description: Number of sentences to fetch before the matched sentence.
3. Sentences After (optional)
   • Type: number
   • Default: 2
   • Description: Number of sentences to fetch after the matched sentence.
4. Lambda (optional)
   • Type: number
   • Default: 0.0
   • Description: Balance between neural search and keyword-based search (0 to 1).
5. Top K (optional)
   • Type: number
   • Default: 5
   • Description: Number of top results to fetch.
6. MMR K (optional)
   • Type: number
   • Default: 50
   • Description: Number of top results to fetch for MMR (Maximal Marginal Relevance).
7. MMR diversity bias (optional)
   • Type: number
   • Default: 0.0
   • Description: Diversity bias for MMR (0.0 to 1.0).

​

Outputs

1. Vectara Retriever
   • Type: retriever
   • Base Classes: [Vectara, VectorStoreRetriever, BaseRetriever]
2. Vectara Vector Store
   • Type: vectorStore
   • Base Classes: [Vectara, VectaraStore]

​

Functionality

• Upsert: Allows adding new documents or files to the Vectara index.
• Retrieval: Provides similarity search capabilities with various configuration options.
• File Handling: Supports both direct file uploads and file retrieval from storage.
• Filtering: Enables metadata filtering for more precise searches.
• Context Configuration: Allows setting the number of sentences before and after the matched sentence.
• Hybrid Search: Supports balancing between neural and keyword-based search through the lambda parameter.
• MMR: Implements Maximal Marginal Relevance for diverse result sets.

​

Use Cases

• Document indexing and retrieval
• Similarity search in large document collections
• Integration of advanced search capabilities in AI applications
• Enhancing chatbots or question-answering systems with relevant document retrieval

​

Notes

• Requires Vectara API credentials for operation.
• Supports various file types for indexing.
• Offers flexible configuration for search behavior and result formatting.