A serverless document analyzer: upload a .txt file to an S3 bucket, Lambda fires automatically, reads the file, sends it to Claude via Bedrock for a summary, and saves the result as filename_summary.txt back to S3.
Set up input and output buckets
In the AWS console, go to S3 → Create bucket. Create two buckets:
ai-pipeline-input-[your-name] — where you'll drop documents
ai-pipeline-output-[your-name] — where summaries get saved
Keep all defaults. Don't enable public access.
Bucket names must be globally unique. Add your initials or a random number at the end. If it says "bucket already exists," just pick a different name.
Write the Lambda function
In the console, go to Lambda → Create function. Choose "Author from scratch," Python 3.12, and attach the bedrock-lambda-role you created on Day 1.
Paste this code into the function editor:
import boto3
import json
import os
s3 = boto3.client("s3")
bedrock = boto3.client("bedrock-runtime", region_name="us-east-1")
OUTPUT_BUCKET = os.environ["OUTPUT_BUCKET"]
def lambda_handler(event, context):
# Get the uploaded file info from the S3 event
bucket = event["Records"][0]["s3"]["bucket"]["name"]
key = event["Records"][0]["s3"]["object"]["key"]
# Read the file content from S3
obj = s3.get_object(Bucket=bucket, Key=key)
text = obj["Body"].read().decode("utf-8")[:4000] # cap at 4K chars
# Send to Claude via Bedrock
response = bedrock.invoke_model(
modelId="anthropic.claude-opus-4-5-20251101-v1:0",
body=json.dumps({
"anthropic_version": "bedrock-2023-05-31",
"max_tokens": 512,
"messages": [{
"role": "user",
"content": f"Summarize this document in 3 bullet points:\n\n{text}"
}]
}),
contentType="application/json",
accept="application/json"
)
summary = json.loads(response["body"].read())["content"][0]["text"]
# Save summary to output bucket
output_key = key.replace(".txt", "_summary.txt")
s3.put_object(
Bucket=OUTPUT_BUCKET,
Key=output_key,
Body=summary.encode("utf-8")
)
return {"statusCode": 200, "body": f"Saved summary to {output_key}"}In the Lambda configuration, add an environment variable: OUTPUT_BUCKET = your output bucket name.
Increase the timeout to 30 seconds (Bedrock calls can take a few seconds). Default 3s will timeout.
Wire the S3 trigger
In Lambda, click Add trigger → S3. Select your input bucket. Event type: PUT. Suffix: .txt (so only text files trigger it).
Now upload any .txt file to your input bucket and watch it work. Check your output bucket for the _summary.txt file. Check CloudWatch Logs for the Lambda execution log.
This is the core serverless pattern for AI processing: event triggers function, function calls AI, result is stored. Swap S3 for SQS, SNS, API Gateway, or EventBridge and you have every major AWS AI pipeline pattern.
Day 3 Complete
- Created S3 input and output buckets
- Built a Lambda function that reads from S3 and calls Bedrock
- Wired S3 PUT event as a Lambda trigger
- Tested the full pipeline with a real document
Next: Knowledge Bases — Managed RAG
Day 4 builds a document Q&A system using Bedrock Knowledge Bases. Upload PDFs, ask questions, get answers grounded in your documents.
Go to Day 4