Verify AWS Service Limits with an AI Agent

What you’ll get

You will set up a Python agent that can verify current AWS service limits. This agent compares information from official AWS documentation, a knowledge base, and live AWS APIs to provide the most up-to-date service limit. This approach is useful when official documentation might be outdated, and you need to confirm live service quotas before deploying resources.

Tools you need

• Amazon Bedrock (paid): A service that provides access to various AI models for building and scaling generative AI applications.
• Strands Agents (free): The specific Python library or framework used to build the agent, found in the provided GitHub repository.
• Python (free): A popular programming language used to write and run the agent’s code.
• Sanity Context Knowledge Base (freemium): A system for storing and retrieving structured content, used here to hold AWS facts.
• AWS API (paid): The interface that allows programs to interact with AWS services, used here to check live service quotas.

Steps

1. Clone the agent repository: Download the project files to your computer.

   git clone https://github.com/simplynadaf/aws-source-of-truth-agent
   cd aws-source-of-truth-agent

   You should see the project files appear in a new folder named aws-source-of-truth-agent.

2. Install required Python packages: Set up the agent’s environment.

   pip install -r requirements.txt

   This command installs all the necessary libraries for the agent to run. You should see a list of packages being installed.

3. Run the credential-free core check: Test the agent’s ability to reconcile information and check live AWS data without needing your personal AWS credentials or API keys.

   python -m agent.reconcile_offline --service EBS --type limit --region us-east-1

   You should see output indicating the verdict, for example: Verdict: 80000 IOPS (serviceQuotasConsole wins over officialDocs).

4. Run the keyword control test: Compare the agent’s structured approach against a simple keyword search to see how it handles potentially outdated information.

   python -m agent.baseline "maximum IOPS per volume general purpose SSD"

   This command will show how a basic keyword search might return an older, incorrect value, highlighting the agent’s advantage.

5. Set up for the full LLM agent (Optional): To use the full agent with a language model, you need to configure your environment variables. Copy the example environment file:

   cp .env.example .env

   Then, edit the .env file to add your AWS region (e.g., us-east-1) and a Sanity Context Viewer token. You will also need to configure access to Amazon Bedrock, which typically involves setting up AWS credentials and potentially a Bedrock region. Refer to the project’s README for detailed instructions on obtaining a Sanity token and configuring Bedrock.

Original source

This workflow is based on a submission by sarvar_04 to the Sanity Challenge. The author built an AI agent to tackle the problem of outdated information in AWS documentation by cross-referencing multiple sources, including live APIs, to determine the current service limits.

Notes & variations

• Free-tier alternatives: While the core agent logic and installation are free, using Amazon Bedrock and live AWS APIs for more advanced features will incur costs. For a completely free experience, you would need to explore open-source language models and local data sources, which might not provide real-time AWS data.
• Common mistake: Relying solely on keyword searches for AWS service limits can lead to using stale data. Always verify critical information against multiple, authoritative sources, especially live APIs when available.
• Tip for better results: For the most accurate and up-to-date information, ensure your Sanity Context Knowledge Base is regularly updated and that the agent is configured to query the most relevant AWS services and regions for your needs.