Retrieval QA Chain

​

Node Details

• Name: retrievalQAChain
• Type: RetrievalQAChain
• Category: [[Chains]]
• Version: 2.0

​

Parameters

1. Language Model (Required)

   • Type: BaseLanguageModel
   • Description: The language model used for generating answers.

2. Vector Store Retriever (Required)

   • Type: BaseRetriever
   • Description: The retriever used to fetch relevant documents from a vector store.

3. Return Source Documents (Optional)

   • Type: boolean
   • Description: Whether to return the source documents used to generate the answer.

4. System Message (Optional)

   • Type: string
   • Description: Custom system message to set the behavior of the AI assistant.
   • Default: A predefined system message template.
   • Rows: 4
   • Additional Params: true

5. Chain Option (Optional)

   • Type: string
   • Options: [“stuff”, “map_reduce”, “refine”]
   • Description: The method used to process retrieved documents.
   • Default: “stuff”
   • Additional Params: true

6. 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

• If Return Source Documents is false:
  • A string containing the answer to the user’s question.
• If Return Source Documents is true:
  • An object containing:
    • text: The 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 vector store retriever fetches relevant documents based on the question.
4. Depending on the chosen Chain Option:
   • “stuff”: Combines all retrieved documents into a single context.
   • “map_reduce”: Processes each document separately and then combines the results.
   • “refine”: Iteratively refines the answer using each document.
5. The language model generates an answer based on the processed documents and the original question.
6. The answer (and optionally source documents) is returned as output.

​

Use Cases

• Building question-answering systems based on specific knowledge bases
• Creating AI assistants with access to large document repositories
• Implementing intelligent search functionality for databases or document collections
• Developing automated customer support systems with access to product documentation
• Creating educational tools that can answer questions based on course materials

​

Special Features

• Flexible Retrieval: Uses vector store for efficient and relevant document retrieval.
• Multiple Processing Options: Supports different methods for handling retrieved documents.
• Source Attribution: Option to return source documents for transparency.
• Customizable System Message: Allows fine-tuning of AI behavior.
• Input Moderation: Can implement safeguards against inappropriate or harmful inputs.
• Scalability: Can handle large document collections efficiently.

​

Notes

• The quality of answers depends on both the underlying language model and the relevance of retrieved documents.
• The choice of Chain Option can significantly impact performance and accuracy for different types of queries.
• Custom system messages can be used to guide the AI’s behavior and response style.
• The chain supports streaming responses for real-time interaction in compatible environments.
• Proper error handling and input validation should be implemented for production use.
• The effectiveness of the chain can vary depending on the quality and organization of the knowledge base.

The Retrieval QA Chain node provides a robust solution for building AI-powered question-answering systems with access to large document repositories. By combining efficient document retrieval with advanced language model processing, it enables the creation of intelligent systems that can provide accurate, context-aware answers to user queries. This node is particularly valuable in scenarios where answers need to be derived from a specific body of knowledge, such as in specialized customer support, educational platforms, or domain-specific research tools.