NEW: Knowledge Base Search Step

Introduction

Introduction

In a major update to our AI tools, we are excited to introduce the Knowledge Base Search Step. This innovative feature will significantly enhance the accuracy of answers retrieved from your knowledge base. If you have experience with the Knowledge Base Query API, you'll find this new step familiar, yet it’s designed to be more efficient and faster. Below, we will walk you through how to implement this feature and make the most of it.

Getting Started

To integrate the Knowledge Base Search Step, navigate to the Dev section of your platform. Here’s how to set it up:

1. Drag and Drop the KB Search Step: This step sends a question to the knowledge base and retrieves raw information based on that query.
2. Saving the Data: For example, you can use something like "last utterance" to store the raw information in a variable called "chunks".
3. Executing the Query: Upon executing the flow and entering a question, such as "What is Voice Flow?", you will see output displaying raw information pulled from your knowledge base.

Template Example

To illustrate how the Knowledge Base Search Step functions, we’ve created a new template specifically for chat projects. This template incorporates:

• An AI Step, which optimizes user queries using memory.
• A search within the knowledge base to reap raw information.
• A well-crafted response that accurately answers the user’s question.

For instance, if you ask, “What is Voice Flow?”, the system provides a pertinent answer. If you follow this up with, “Tell me more about number three,” the system can delve deeper into that specific point without requiring additional side flows or complex setups. Additionally, the system supports multilingual queries, allowing questions to be asked in languages such as Spanish, and responds appropriately.

How It Works

Behind the scenes, the flow utilizes several components:

• Memory Optimization: Before invoking the KB Search Step, the user’s query is optimized with previously stored memory.
• Language Detection: The flow can determine the language of the user’s input, enhancing the response's relevance.
• Response Crafting: The raw information from the KB Search is synthesized into a coherent output that aligns with the user's memory and preferred language.

Moreover, if the retrieved data doesn't meet a certain confidence level, a no-match path can be created to manage queries that lack sufficient information. This updated approach makes building a high-accuracy AI agent more streamlined than before, eliminating the labor-intensive processes previously required.

Conclusion

The Knowledge Base Search Step is a game-changer for those looking to leverage AI in their projects. With templates pre-loaded with this flow, users can hit the ground running and develop an effective AI agent with ease and efficiency.

Keyword

knowledge base search, AI answers, information retrieval, query optimization, user memory, multilingual support, no-match path, Voice Flow, efficiency

FAQ

Q1: What is the Knowledge Base Search Step?
A1: It is a new feature that allows you to retrieve raw information more efficiently from your knowledge base, enhancing the accuracy of AI-generated answers.

Q2: How do I implement the Knowledge Base Search Step?
A2: You can find it in the Dev section of your platform, drag it into your flow, and set it up to work with previously stored user data.

Q3: Can this feature handle multiple languages?
A3: Yes, the Knowledge Base Search Step can respond to queries in different languages, providing flexibility and broader accessibility.

Q4: What should I do if the retrieved information does not match the user's question?
A4: You can create a no-match path to handle situations where the confidence level of the returned data is insufficient.

Q5: Is the Knowledge Base Search Step included in existing templates?
A5: Yes, every new template comes pre-loaded with this flow, making it easy to get started quickly.