Practical overview
When it fits
Use it when you want to deploy and adapt an AWS-based document assistant that routes questions between semantic search and SQL, including multi-step queries.
Who it is for
- Teams building a conversational question-answering system over business documents in AWS.
- Developers adapting an AWS CDK, Bedrock and SageMaker reference solution.
Documented uses
- Retrieve relevant document information to answer factual questions.
- Ask analytical questions, such as which contracts expire within a stated period, by querying extracted entities through SQL.
- Combine retrieval and analytical tools to address multi-step questions.
- Process document data in batches and explore notebooks for updating documents and extracting entities.
Requirements & limits
The documented setup deploys into an AWS account and requires configured Amazon Bedrock access to Anthropic Claude and Amazon Titan in a supported Bedrock region. The hosted app uses Cognito authentication; users create an account to log in.
- The assistant distinguishes factual retrieval from analytical SQL questions using a ReAct agent pattern.
- The solution includes deployment infrastructure and a frontend, not only standalone notebooks.
- The example data-loading pipeline uses a CSV for SQL tables and preprocessed JSON documents for the semantic-search index.
- The repository is licensed under MIT-0.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up the assistant in an AWS account
Deploy the documented backend and frontend, load the example data, and open the chat assistant.
Before you start
- An AWS account.
- Amazon Bedrock model access for Anthropic Claude and Amazon Titan in a supported Bedrock region.
- AWS CDK command-line setup, using Cloud9, CloudShell or a configured local environment.
Steps
Clone the repository, enter the backend folder, and run npm install.
If CDK has not been bootstrapped for the account and region, run npx cdk bootstrap; then run npx cdk deploy in the backend folder and note the SageMaker IAM Policy ARN in the output.
In the frontend folder, run npm install and npx cdk deploy. When the Amplify CI/CD setup is ready, trigger a build in the Amplify console.
Run the SageMaker Pipeline notebook at data-pipelines/04-sagemaker-pipeline-for-documents-processing.ipynb to load the example CSV into SQL tables and use the preprocessed JSON documents to create the semantic-search index.
Open the hosted app, create an account and log in to ask questions.
Expected outcome
A deployed chat assistant populated with the documented example data for factual and analytical questions.
Guide limitations
- The setup requires the stated Bedrock model access in a supported region.
- The documented pipeline loads example CSV and JSON data; updating documents and extracting new entities involves notebooks 1 to 5 in the experiments folder.
Remove the deployed AWS stacks
Remove the backend and frontend stacks created by the setup.
Before you start
- The repository and AWS CDK command-line environment used for deployment.
Steps
In the backend folder, run npx cdk destroy.
In the frontend folder, run npx cdk destroy.
Expected outcome
The documented cleanup commands are issued for both solution stacks.
Guide limitations
- The instructions cover the backend and frontend stacks only.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 24 October 2026
- Original website title
- aws-samples/aws-agentic-document-assistant · Official homepage title · retained in its source language