Build Accurate AI Apps Without Complex RAG: Introducing Amazon Nova Web Grounding
Amazon Nova Web Grounding is an out‑of‑the‑box RAG tool for Amazon Bedrock that automatically retrieves and cites up‑to‑date information, reduces hallucinations, and lets developers create accurate, real‑time AI applications using simple Python code.
Amazon Nova Web Grounding is a new built‑in tool for the Amazon Nova model on Amazon Bedrock that provides an out‑of‑the‑box Retrieval‑Augmented Generation (RAG) solution. The model can automatically decide when to retrieve up‑to‑date information, incorporate cited sources into the context, and thus reduce hallucinations.
Application Scenarios
The feature is useful for knowledge‑base chat assistants, content‑generation tools that require fact‑checking, research assistants that need to aggregate multiple recent sources, and customer‑support applications that demand high accuracy and traceable citations.
Getting Started
Enable the Web Grounding feature in the Amazon Nova model, then use the standard Amazon Bedrock Converse API via the AWS SDK for Python (Boto3). The example below creates a Boto3 session, builds a BedrockRuntime client, and prepares a conversation payload where the user role is “user”.
try:
session = boto3.Session(region_name='us-east-1')
client = session.client('bedrock-runtime')Prepare the request payload with a question that requires fresh information, e.g., “What are the current AWS regions and their locations?”
# Prepare the conversation in the format expected by Bedrock
question = "What are the current AWS regions and their locations?"
conversation = [
{
"role": "user", # Indicates this message is from the user
"content": [{"text": question}]
}
]When Web Grounding is disabled, a simple call to client.converse returns the model’s answer without external retrieval.
model_id = "us.amazon.nova-premier-v1:0"
response = client.converse(
modelId=model_id,
messages=conversation
)
print(response['output']['message']['content'][0]['text'])After enabling Web Grounding, the same call includes a toolConfig that declares nova_grounding as an available system tool, prompting the model to perform a live search.
response = client.converse(
modelId=model_id,
messages=conversation,
toolConfig={
"tools": [
{
"systemTool": {
"name": "nova_grounding" # Enables the model to search real‑time information
}
}
]
}
)The response contains a trace showing which step triggered the external query, a HIDDEN field that masks the raw external data, and a citationsContent object that lists the sources used.
Important Notes
Web Grounding is currently available in all AWS regions but incurs additional charges as described on the Amazon Bedrock pricing page. It currently works only with the Amazon Nova Premier model; support for other Nova models will be added soon.
Developers who have not used Amazon Nova before are encouraged to attend the self‑paced online workshop to practice using the model and its features.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
Amazon Cloud Developers
Official technical community of Amazon Cloud. Shares practical AI/ML, big data, database, modern app development, IoT content, offers comprehensive learning resources, hosts regular developer events, and continuously empowers developers.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
