In order to avoid switching applications and contexts, there is a consistent customer feedback that AI Assistant is most useful if you can interface with users in the productive tools already used. Web applications such as Amazon Q Business and Slack have become an essential environment for the latest AI assistant deployment. In this post, various interfaces will investigate how to enhance user interaction, improve accessibility, and respond to various tastes.
By providing seamless experiences throughout the environment, organizations can increase user satisfaction and recruitment rate. Assistant uses search expansion generations (RAG) to integrate the reliable sources in response to the entire interface and enhance reliability and educational value. This multi-interface, the Rag-Powered approach, not only strives to meet the latest user’s flexibility, but also promotes more information-based user-based and ultimately effective assistants. Maximizes sex and reach. By combining RAG and multiple interfaces, assistants provide consistent and accurate information, regardless of user priority environment and productivity tools.
Overview of solutions
The following figure shows the architecture design of the application.
You can find a complete code and the procedure for deploying solutions in the GitHub repository.
Click here to open the AWS console and follow.
Prerequisite
The following prerequisites are required:
Expand the solution
For the setup procedure, see Readme in the GitHub repository.
Solution component
This section describes two important components in solutions. Data source and vector database.
Data source
Use the Spack Document RST (reconstructed text) file uploaded to the Amazon Simple Storage Service (Amazon S3) Bucket. Every time the assistant returns it as a source, it is not the top of the source page, but a link to a specific part of the spack document. For example, a spack image of Docker Hub.
Spack is a multipurpose package manager of supercomputers, Linux, and Maco, which revolutions scientific software installation so that multiple versions, composition, environment, and compilers can coexist on a single machine. Masu. Spack, developed by Todd Gamblin at Lawrence Livermore National Laboratory in 2013, has dealt with restrictions on conventional package managers in the high -performance computing (HPC) environment. Brian Weston, the lead of Cloud Transformation, a mission science program in LLNL, advised on the development of this assistant.
In addition, use a text file uploaded to the S3 bucket that can be accessed from the Amazon CloudFront link. There is also an automated intake work, from Slack conversation data to S3 bucket equipped with the AWS Lambda function. This allows the assistant to answer questions using previous conversations from the user and quote the source. If this source is quoted in Amazon Q, the user may not be able to access the Slack data, so we chose to use the CloudFront link instead of using the Slack link. Some are replacing this methodology using Amazon Kendra’s Slack connector.
This solution can support other data types as long as you can extract text, supplied to vector databases, change the code, and change the code, such as PDF and word documentation. Their raw files can be provided by distribution of cloudfront.
The next screenshot shows the sample cloud front URL.
When expanded, the existing data is automatically uploaded to the S3 bucket and processed by the assistant. This solution also uses Amazon Eventbridge to include daily intake of data from Slack to application.
Vector database
In this solution, Amazon Kendra is used as a vector database and provides a great advantage of simplicity and cost -effective. Amazon Kendra reduces both development costs and maintenance costs as a completely managed AWS service. Amazon Q, which supports two types of retriber (native retriber and Amazon Kendra), is seamlessly integrated into this setup. By using Amazon Kendra, the solution efficiently uses the same retriever for both Amazon Q and Slack interfaces. This approach not only streamlines overall architectures, but also provides more consistent user experience in both environments. As a result, regardless of the interface selected by the user, a highly cost -efficient system that maintains the uniformity of information search and presentation is formed.
Amazon Kendra also supports the use of metadata for each source file. This allows both UIs to provide links to the source, whether on the Spack Document website or the CloudFront link. In addition, Amazon Kendra can support relevant tuning and enhance certain data sources. This solution has increased the results of spack documents.
User interface
This section describes the UI used in this solution.
Amazon Q business
Amazon Q Business offers a safe and enhanced AI assistant using RAG. As an AWS native solution, it is seamlessly integrated with other AWS services and has a unique user -friendly interface. This integration provides a smooth implementation experience in combination with its simple setup and deployment process. Amazon Q Business provides an accurate and conscious response to accurate and contests that are firmly rooted in specific data and documents of the tissue by combining intelligent information search and generated AI functions from enterprise systems. Increase sex and accuracy.
The next screenshot is an example of Amazon Q Business UI.
slack
Slack is a popular collaboration service that is an essential part of many organizations’ communication forums. Its versatility is beyond team messaging and functions as an effective interface of the assistant. By integrating assistants equipped with AI into Slack, companies can use familiar environments to provide user immediate access to information.
The next screenshot shows an example of Slack UI using a message thread.
Monitoring
Amazon Q has an analysis dashboard that provides insights on user engagement in a specific Amazon Q business application environment. It provides valuable data on the patterns, conversation dynamics, user feedback, and query trends, and can analyze and optimize the performance of AI assistants and the interaction of users.
Slack collects user feedback as shown in the screenshot in front of the UI. Users can add “thumb” or “thumb down” to the assistant’s response to track the performance. In addition, we used the Amazon CloudWatch dashboard to imitate the Amazon Q Analytics dashboard and build a custom solution that further consistently consistently consistently consistence of two applications.
The next screenshot shows an example of the Slack Cloudwatch dashboard.
In addition, there is a daily scheduled Slack message that summarizes the SlackBot data of the past day, as shown in the screenshot below.
cleaning
To prevent the ongoing rates from being charged, clean up the resources created as part of this post using the commands mentioned in README.
Conclusion
The implementation of a multi -interface AI assistant using RAG indicates a leap in AI -led organizational communication. By integrating Amazon Q Business and Slack interfaces with a robust backend equipped with Amazon Kendra, this solution provides accurate and conscious information with a seamless and unreasonable access. The architecture strengths are the consistency of the entire environment, the processing process, and the comprehensive monitoring function. This approach not only improves user engagement and productivity, but also positions organizations to quickly adapt to evolving communication needs in AI -centered landscapes, and to a more efficient and intelligent information management system. Mark a very important step.
For more information about the AWS service used in this solution, expand the Amazon Bedrock Slack Gateway and Amazon Kendra Developer Guide in Amazon Q User Guide.
About the author
Nick Viso I’m an AWS professional service machine learning engineer. He uses data science and engineering to solve complex organizational and technical issues. In addition, he builds and develops an AI/ML model in the AWS cloud. His passion extends to his trends and diverse cultural experiences.
Dr. Ian Landsford AWS Professional Service Aerospace Cloud Consultant. He integrates cloud services into the Aerospace application. In addition, Ian focuses on building AI/ML solutions using AWS services.