Vectara QA Chain

​

Node Details

• Name: vectaraQAChain
• Type: VectaraQAChain
• Category: [[Chains]]
• Version: 2.0

​

Parameters

1. Vectara Store (Required)

   • Type: VectorStore
   • Description: The Vectara vector store used for document retrieval.

2. Summarizer Prompt Name (Required)

   • Type: options
   • Options:
     • “vectara-summary-ext-v1.2.0” (gpt-3.5-turbo)
     • “vectara-experimental-summary-ext-2023-10-23-small” (gpt-3.5-turbo)
     • “vectara-summary-ext-v1.3.0” (gpt-4.0)
     • “vectara-experimental-summary-ext-2023-10-23-med” (gpt-4.0)
   • Description: The summarizer prompt to use for generating responses.
   • Default: “vectara-summary-ext-v1.2.0”

3. Response Language (Optional)

   • Type: options
   • Options: Multiple language options available (e.g., English, German, French, etc.)
   • Description: The language in which to return the response.
   • Default: “eng” (English)

4. Max Summarized Results (Optional)

   • Type: number
   • Description: Maximum number of results used to build the summarized response.
   • Default: 7

5. Input Moderation (Optional)

   • Type: Moderation[]
   • Description: Moderation tools to detect and prevent harmful input.
   • List: true

​

Input

• A string containing the user’s question or query.

​

Output

• An object containing:
  • text: The summarized answer to the user’s question.
  • sourceDocuments: An array of documents used to generate the answer.

​

How It Works

1. The chain receives a user question.
2. If input moderation is enabled, it checks the input for potential harmful content.
3. The Vectara store retrieves relevant documents based on the question.
4. The retrieved documents are processed and ranked.
5. The specified summarizer prompt is used to generate a concise answer from the top-ranked documents.
6. The answer is formatted with reordered citations.
7. The final answer and source documents are returned as output.

​

Use Cases

• Building advanced question-answering systems with Vectara’s search capabilities
• Creating AI assistants with access to large, complex document repositories
• Implementing intelligent search functionality for enterprise knowledge bases
• Developing summarization tools for research and information retrieval
• Creating multilingual question-answering systems

​

Special Features

• Advanced Retrieval: Utilizes Vectara’s powerful search and ranking algorithms.
• Flexible Summarization: Offers multiple summarizer options for different use cases.
• Multilingual Support: Can generate responses in various languages.
• Citation Ordering: Automatically reorders citations for coherent presentation.
• MMR Reranking: Supports Maximal Marginal Relevance for diverse results.
• Input Moderation: Can implement safeguards against inappropriate or harmful inputs.

​

Notes

• The quality of answers depends on the relevance of retrieved documents and the chosen summarizer prompt.
• Different summarizer prompts may be available based on the user’s Vectara account type (e.g., Growth vs. Scale).
• The chain supports both single-corpus and multi-corpus searches.
• Custom filtering and reranking options are available through the Vectara store configuration.
• The effectiveness of the chain can vary depending on the quality and organization of the indexed documents.
• Proper error handling should be implemented, especially for potential API failures or summarization issues.

The Vectara QA Chain node provides a sophisticated solution for building AI-powered question-answering systems that leverage Vectara’s advanced search and summarization capabilities. It excels in scenarios requiring accurate information retrieval and concise summarization from large document collections. This node is particularly valuable for enterprises needing to extract insights from vast knowledge bases, researchers seeking efficient ways to summarize findings, or developers building multilingual information retrieval systems.