DEV Community: Swetha Mudunuri

Amazon Nova Sonic

Swetha Mudunuri — Tue, 20 May 2025 11:06:11 +0000

AWS Nova Sonic: When a Voice Bot Became My Coach

Swetha Mudunuri — Mon, 19 May 2025 05:52:28 +0000

I still remember the last time I tried calling customer support. After being greeted by an overly cheerful but utterly unhelpful automated voice, I spent minutes pressing keys, repeating myself, and getting nowhere. I felt ignored, frustrated, and drained and eventually, I just hung up.

When I attended the AWS Summit last month, I wasn’t expecting anything different until I walked into the AWS Nova Sonic workshop. Initially, I thought: "Not another voice bot..." But something about the way they presented it like a human-like companion rather than a rigid IVR made me curious.

Unlike typical IVRs, this system felt responsive, natural. It wasn’t just routing the user to another department. It was understanding, responding, and completing actions like a smart assistant.

I started thinking: What if this wasn’t just a tool for ticket booking? What if it could become something more like a coach or a friend who understands you and guides you in real-time?

🧩 The Problem It’s Solving

Here's what AWS Nova Sonic is addressing:

❌ Traditional IVRs are rigid, impersonal, and often frustrating.
⌛ Users waste time navigating voice menus or repeating themselves.
❌ Lack of context retention the system forgets your previous queries.
🌍 Limited support for informal speech or multilingual input.
🤖 No sense of empathy — just a robotic process.

✨ Key Features That Impressed Me

Here’s what makes AWS Nova Sonic stand out:

🗣️ Barge-in support you can interrupt mid-sentence just like real conversations.
🌐 Multilingual understanding I spoke in Hindi and it responded in English, accurately!
🧠 Context awareness it maintains chat history for more meaningful responses.
📚 RAG integration with AWS Bedrock Knowledge Bases.
🛡️ Built-in guardrails for security and safe responses.
✅ Responsible AI (RAI) framework to ensure ethical usage.

When I got back from the summit, I was curious to try it out myself. The QR code from the workshop didn’t work, but I found an AWS workshop online and gave it a shot.

As expected, I ran into some Python issues (what's a workshop without a little debugging, right?), but eventually got Nova Sonic up and running.

It was genuinely amazing.

I tested it in Hindi — just to push its limits and was amazed to see it interpret and transcribe in English correctly.

I said, “Mujhe kal ki Bangalore se Delhi ki flight book karni hai subah ke 9 baje ke aaspaas.”

And it responded with a relevant English confirmation flow for booking a flight. That’s when it really clicked — it was listening.

Then, I asked it to be my Speaking Coach.

And it responded with suggestions on breathing exercises, visualization techniques, and most importantly reassurance. It told me it’s okay to feel nervous before a presentation, and that confidence comes with practice.

For a moment, I felt, I wasn’t talking to a bot. I was talking to a companion. I learnt some French words as well 🥰.

02:27

Swetha Mudunuri

May 20 '25

Amazon Nova Sonic

#aws #software

Comments

1 min read

This wasn’t just about booking tickets or giving predefined answers. AWS Nova Sonic gave me a glimpse of what human-centered AI could feel like. Helpful. Understanding. Present.

If you've ever hung up on a frustrating IVR, try giving Sonic a shot not just for what it’s built to do, but for what it could become.

🧾 Conclusion

What started as curiosity turned into a moment of genuine surprise. AWS Nova Sonic isn’t just another voice automation tool — it’s a step toward more human, responsive, and emotionally aware AI interactions.

While I’ve only tried the basic version so far, and haven’t yet tested the RAG integration with Bedrock Knowledge Bases, I can already see the potential. It's exciting to imagine the possibilities as I explore deeper from contextual knowledge fetching to complex workflows.

I'll be sharing more as I experiment further. Until then, take care and keep exploring!

Resources

https://catalog.us-east-1.prod.workshops.aws/workshops/5238419f-1337-4e0f-8cd7-02239486c40d/en-US/200-labs/06-guardrail

Getting Started with Amazon Q CLI and AWS MCP Servers on Windows

Swetha Mudunuri — Fri, 09 May 2025 11:59:12 +0000

If you're curious about how to bring generative AI to your terminal and enhance it with contextual awareness of AWS services, this blog walks you through setting up Amazon Q CLI on a Windows environment and connecting it to one or more MCP (Model Context Protocol) servers.

Whether you're experimenting with infrastructure as code, checking AWS best practices, or planning cost analysis, MCP gives Amazon Q the power to understand your environment better.

What’s Amazon Q CLI and MCP?
Amazon Q CLI is a command-line tool that lets you interact with a generative AI assistant directly in your terminal. Think of it as having a smart AWS-AI assistant that can help with commands, code generation, documentation lookup, and much more.

MCP (Model Context Protocol) is an open protocol that connects AI tools like Q CLI to external services — such as AWS documentation, CDK best practices, cost calculators, or even image generators. MCP servers provide that real-time context the model might not have, especially for newer APIs or service updates.

🏗️ Amazon Q CLI Architecture

You interact with Q CLI
– You enter natural language or AWS-related commands into your terminal.
Q CLI handles your session
– It handles AWS authentication and keeps track of your session.
It checks what you’re asking for
– Q CLI checks whether you're requesting an AWS action, a local file operation, or AI help.
It connects to AWS services or your local files
– It either runs AWS commands or interacts with your local file system, based on the request.
If your request needs AI understanding
– For natural language or complex tasks, Q CLI calls Amazon’s large language model (LLM) to assist.
Q CLI looks for any available MCP tools
– It checks for any local MCP servers that are available and relevant to your request
Relevant parts of the request are sent to MCP servers
– Q CLI sends parts of the request to the appropriate MCP tools for specialized processing.
MCP tools send back results or extra context
– These tools provide extra context or project-specific output to support your query.
Final response is shown in your terminal
– The final result whether its from AWS, local systems, AI, or MCP tools—is provided in your terminal.

Each MCP server you configure exposes a specific domain of knowledge. You can connect multiple MCP servers depending on your use case.

How to setup Amazon Q CLI on Windows (via WSL)

Since Amazon Q CLI is Linux-based, you'll need to use the Windows Subsystem for Linux (WSL) to run it on a Windows machine. Follow these steps:

✅ Step 1: Install WSL on Windows ( I am using Visual Studio Terminal for installation)
Open PowerShell and run:

wsl --install

This installs Ubuntu by default.
During installation, you’ll be prompted to set a username and password for the Linux environment.

✅ Step 2: Download the Amazon Q CLI Installer
From your WSL terminal, download the appropriate zip file for your Linux distribution.

For Linux x86-64, run:

curl --proto '=https' --tlsv1.2 -sSf "https://desktop-release.q.us-east-1.amazonaws.com/latest/q-x86_64-linux.zip" -o "q.zip"

✅ Step 3: Install Dependencies
Before proceeding, make sure required packages are installed:

sudo apt get install unzip

✅ Step 4: Extract the Q CLI Package
Unzip the downloaded file:

unzip q.zip

✅ Step 5: Install Amazon Q CLI
Run the install script:

q/install.sh

During installation, you'll be asked to provide:
Shell Config Modification: Select Yes to allow the script to update your shell configuration automatically.
Login Method: Choose "Use for Free with Builder ID"-Authenticate using your Builder ID.

✅ Step 6: Relaunch and Verify Setup

Close the terminal after installation is complete.
Relaunch the terminal and log back into your Ubuntu (WSL) environment.
Run the following command to verify that Amazon Q CLI is installed correctly:

wsl -d Ubuntu

q --version

For official documentation, refer to: Amazon-Q-CLI Installation Guide

MCP Server Configuration with Amazon Q CLI
Now that Amazon Q CLI is installed, let’s proceed with setting up the MCP Server.

Prerequisites for MCP Server Installation

🔹 Step 1: Install uv
Install uv using curl:

curl -LsSf https://astral.sh/uv/install.sh | sh

ℹ️ Other installation methods (e.g., using pip) are also available — refer GitHub Repository for details.

🔹 Step 2: Verify uv Installation

uv --version

🔹 Step 3: Install Python 3.10 Using uv

uv python install 3.10

🔹 Step 4: Ensure AWS CLI is Installed

Check if AWS CLI is installed:

aws --version

If not, install it using:

curl "https://awscli.amazonaws.com/awscli-exe-linux-x86_64.zip" -o "awscliv2.zip"
unzip awscliv2.zip
sudo ./aws/install

Then configure credentials:

aws configure

Make sure your IAM role/user has permissions to use AWS MCP services (like access to the AWS Pricing API).

🔹 Step 5: Install Required Packages for MCP

Some MCP servers (like AWS Diagram MCP Server) need tools like GraphViz. Install it:

sudo apt-get install graphviz

🗂️MCP Configuration in Amazon Q

Step 6: Create MCP Configuration Folder and File
Create the necessary folder:

Now we need to create a folder amazon q under ~/.aws

mkdir -p ~/.aws/amazonq
cd ~/.aws/amazonq

Create and edit the mcp.json file:

{
  "mcpServers": {
    "awslabs.nova-canvas-mcp-server": {
      "command": "uvx",
      "args": ["awslabs.nova-canvas-mcp-server@latest"],
      "env": {
        "AWS_PROFILE": "**your-aws-profile**",
        "AWS_REGION": "us-east-1",
        "FASTMCP_LOG_LEVEL": "ERROR"
      },
      "disabled": false,
      "autoApprove": []
    }
  }
}

Don't forgot to change your ur profile details.

🔁 You can configure multiple MCP servers by listing them under "mcpServers" like so:

{
  "mcpServers": {
    "mcpserver1": { ... },
    "mcpserver2": { ... }
  }
}

🚀 Launch Amazon Q

Amazon-q-terminal

You’ll now be logged into the Amazon Q CLI with your configured MCP servers.

I've set up four MCP servers — nova-canvas, cost-analysis, aws-diagram, and document. The cover image was generated using the nova-canvas server, while the architecture diagram below was created using the aws-diagram MCP server

Security Considerations

All communication between the CLI and AWS services is secured via HTTPS
Authentication is handled through IAM
Local file system and command execution is limited to the user's permissions
MCP servers operate within the user's environment with appropriate access controls

References
GitHub link for AWS MCP servers
AWS-Cli-Installation
AWS-Cli-User-Guide

Thank you for reading!!!!!!