Retrieval QA
This component, Retrieval QA, is part of Nappai, your automation and AI assistant. It helps you get answers to your questions quickly by searching through your data. You ask a question, and it finds the relevant information and gives you a concise answer.
Relationship with Language Models and Retrievers
This component uses a powerful language model (like a very smart computer program) to understand your questions and a retriever to find the best information from your connected data sources (like databases or files). It combines these technologies to give you the most accurate and relevant answers.
Inputs
- Input: This is where you type your question. It’s the most important part – make sure to ask clear and specific questions.
- Chain Type: This is an advanced setting that controls how the component processes your question. You can usually leave this as the default (“Stuff”). Experimenting with other options might improve results for complex queries.
- Language Model: This setting selects the specific language model Nappai will use to understand and answer your question. This is usually pre-configured and you won’t need to change it.
- Retriever: This setting determines which data sources Nappai will search to answer your question. This is usually pre-configured and you won’t need to change it.
- Memory: This option allows the component to remember previous questions and answers in a conversation. This is helpful for asking follow-up questions.
- Return Source Documents: If you check this box, the component will also show you the specific documents or data it used to create the answer. This is useful for verifying the accuracy of the response.
Outputs
The component provides a text-based answer to your question. If you selected “Return Source Documents,” it will also include references to the data used to generate the answer. You can use this answer directly in your workflow or copy and paste it elsewhere.
Usage Example
Let’s say you want to know the total sales for last month. You would type “What were the total sales for last month?” into the Input field. The component will then search your connected sales data and return the total sales figure.
Templates
[This section will be populated with a list of templates where this component is used once the templates are defined.]
Related Components
[This section will be populated with links to other related components and descriptions once the other components are defined. Examples might include components that provide the data the Retrieval QA component searches, or components that process the output of the Retrieval QA component.]
Tips and Best Practices
- Ask clear and specific questions. The more precise your question, the better the answer.
- If you don’t get the answer you expect, try rephrasing your question.
- Use the “Return Source Documents” option to understand how the component arrived at its answer.
- Experiment with different Chain Types for complex queries.
Security Considerations
The security of this component depends on the security of your connected data sources and the language model used. Ensure that your data sources are properly secured and that you are using a reputable language model provider. Nappai employs security measures to protect your data, but it’s crucial to maintain best practices for data security in your overall system.