⚙️ SQS AI Lambda — Tender Summarization Service

A production-ready, high-throughput AWS Lambda that processes South African tender messages from multiple sources via Amazon SQS (FIFO) and generates intelligent summaries using Anthropic Claude 3 Sonnet via Amazon Bedrock. It continuously polls the source queue, validates and enriches messages with AI-generated summaries, and publishes results to a write queue—while safely routing failures to a DLQ. Built for reliability, observability, and smart prompt management through AWS Parameter Store.

Sources: eTenders, Eskom, Transnet, SARS, SANRAL (government and utility)
AI Model: Anthropic Claude 3 Sonnet via Amazon Bedrock
Prompts: Dynamic, source-specific prompts managed in AWS Parameter Store
Queues: AIQueue (input), TagQueue (success), FailedQueue (errors)
Focus: Quality AI summaries, speed, resilience, clear monitoring

📚 Table of Contents

✨ Features
🧭 Architecture (Simple View)
💡 Why SQS + Lambda + Bedrock + Parameter Store?
🧩 What's Inside (Project Structure)
⚙️ Configuration
🚀 Quick Start
🧠 How It Works
📦 Tech Stack
🗂️ Message Models (At a Glance)
📮 Queues
📈 Monitoring & Metrics
🛡️ Error Handling & Recovery
⚡ Performance & Scaling
🔒 Security
🧪 Testing
📦 Deployment
🧰 Troubleshooting
🗺️ Roadmap

✨ Features

🚀 High throughput: continuous polling + batch processing (up to 10 messages)
🧠 Advanced AI Summarization: Uses Anthropic Claude 3 Sonnet via Amazon Bedrock
🎯 Source-specific Prompts: Fetches tailored instructions from AWS Parameter Store
🔄 Multi-source support: eTenders, Eskom, Transnet, SARS, SANRAL (easy to add more)
⚙️ External Prompt Management: Update AI instructions without code deployments
⚡ Efficient Caching: In-memory prompt caching reduces API calls
🧯 Safe-by-default: DLQ routing, partial-batch success, Bedrock throttling handling
🧭 Observable: structured logs, clear metrics, error context
🧩 JSON-friendly: case-insensitive and camelCase support

🧭 Architecture (Simple View)

Tender Scrapers
    ↓
TenderQueue (TenderQueue.fifo)
    ↓
Deduplication Lambda
    ├─ Check RDS Database (duplicate validation)
    └─ Route based on deduplication results
           ↓
AIQueue (AIQueue.fifo)
    ↓
AI Summary Lambda (AILambda)
    ├─ Message Factory (eTenders, Eskom, Transnet, SARS, SANRAL)
    ├─ Prompt Service ← AWS Parameter Store (dynamic prompts)
    ├─ Bedrock Service → Amazon Bedrock (Claude 3 Sonnet)
    ├─ Message Processor (validate → AI summarize)
    └─ SQS Service (I/O)
           ├─ TagQueue (TagQueue.fifo)        ← success + AI summaries
           └─ FailedQueue (FailedQueue.fifo)  ← errors/DLQ
                ↓
AI Tagging Lambda
    ├─ Apply relevant tags based on content
    └─ Route tagged messages
           ↓
WriteQueue (WriteQueue.fifo)
    ↓
Final Processing → RDS Database

💡 Why SQS + Lambda + Bedrock + Parameter Store?

Resilient by design: retries, visibility timeouts, and DLQs
Pay-per-use: scales with load, no idle cost
Intelligent Processing: State-of-the-art AI summaries via Claude 3 Sonnet
Flexible Prompts: External prompt management allows quick iterations
Simple ops model: logs + metrics in CloudWatch, minimal moving parts
Easy extension: add new sources via the Message Factory pattern

🧩 What's Inside (Project Structure)

Sqs_AI_Lambda/
├── Function.cs                     # Lambda entry point with continuous polling
├── Models/
│   ├── TenderMessageBase.cs        # Base model with AISummary property
│   ├── ETenderMessage.cs           # eTenders model
│   ├── EskomTenderMessage.cs       # Eskom model
│   ├── TransnetTenderMessage.cs    # Transnet model
│   ├── SarsTenderMessage.cs        # SARS model
│   ├── SanralTenderMessage.cs      # SANRAL model
│   ├── SupportingDocument.cs       # Attachments
│   └── QueueMessage.cs             # Internal SQS wrapper
├── Services/
│   ├── MessageProcessor.cs         # Business pipeline with AI integration
│   ├── MessageFactory.cs           # Creates typed messages by MessageGroupId
│   ├── BedrockSummaryService.cs    # Claude 3 Sonnet integration with retries
│   ├── PromptService.cs            # Parameter Store prompt management
│   └── SqsService.cs               # SQS operations
├── Interfaces/
│   ├── IMessageProcessor.cs
│   ├── IMessageFactory.cs
│   ├── IBedrockSummaryService.cs
│   ├── IPromptService.cs
│   └── ISqsService.cs
├── Converters/
│   └── StringOrNumberConverter.cs  # Flexible type conversion
├── aws-lambda-tools-defaults.json  # Deployment config
└── README.md

⚙️ Configuration

AWS Parameter Store Setup

Create these String parameters in AWS Systems Manager Parameter Store:

/TenderSummary/Prompts/System      # General instructions for Claude 3
/TenderSummary/Prompts/eTenders    # eTenders-specific prompts
/TenderSummary/Prompts/Eskom       # Eskom-specific prompts
/TenderSummary/Prompts/Transnet    # Transnet-specific prompts
/TenderSummary/Prompts/SARS        # SARS-specific prompts
/TenderSummary/Prompts/SANRAL      # SANRAL-specific prompts

Environment Variables

SOURCE_QUEUE_URL="https://sqs.us-east-1.amazonaws.com/211635102441/AIQueue.fifo"
TAG_QUEUE_URL="https://sqs.us-east-1.amazonaws.com/211635102441/TagQueue.fifo"
FAILED_QUEUE_URL="https://sqs.us-east-1.amazonaws.com/211635102441/FailedQueue.fifo"

IAM Permissions Required

SQS: ReceiveMessage, DeleteMessage, GetQueueAttributes, SendMessage
Bedrock: bedrock:InvokeModel (Claude 3 Sonnet model ARN)
Parameter Store: ssm:GetParameter (on /TenderSummary/Prompts/*)
CloudWatch Logs: logs:CreateLogGroup, logs:CreateLogStream, logs:PutLogEvents

🧠 How It Works

Ingest (continuous polling, small time budget safety)

while (context.RemainingTime > TimeSpan.FromSeconds(30))
{
    var messages = await PollMessagesFromQueue(10);
    if (!messages.Any()) break;
    await ProcessMessageBatch(messages);
}

Route & Create (by MessageGroupId, now supports 5 sources)

var result = groupIdLower switch
{
    "etenderscrape" or "etenderlambda" => CreateETenderMessage(body, groupId),
    "eskomtenderscrape" or "eskomlambda" => CreateEskomTenderMessage(body, groupId),
    "transnettenderscrape" or "transnetlambda" => CreateTransnetTenderMessage(body, groupId),
    "sarstenderscrape" or "sarslambda" => CreateSarsTenderMessage(body, groupId),
    "sanraltenderscrape" or "sanrallambda" => CreateSanralTenderMessage(body, groupId),
    _ => HandleUnsupportedMessageGroup(groupId)
};

Fetch Prompts (cached from Parameter Store)

var systemPrompt = await _promptService.GetPromptAsync("System");
var sourcePrompt = await _promptService.GetPromptAsync(sourceType);

AI Summarization (Claude 3 Sonnet via Bedrock)

var summary = await _bedrockService.GenerateSummaryAsync(message, systemPrompt, sourcePrompt);
message.AISummary = summary;

Process & Tag

Adds lifecycle tags (Processed, AI-Enhanced, handler name, UTC timestamp)

Output (transactional)

Success → TagQueue + delete from AIQueue
Error → FailedQueue (DLQ) + keep in AIQueue for retry

📦 Tech Stack

Runtime: .NET 8 (LTS)
Compute: AWS Lambda
AI Model: Anthropic Claude 3 Sonnet (via Amazon Bedrock)
Configuration: AWS Systems Manager Parameter Store
Messaging: Amazon SQS (FIFO)
Serialization: System.Text.Json
Logging/DI: Microsoft.Extensions.*

NuGet (excerpt):

<PackageReference Include="Amazon.Lambda.Core" Version="2.5.0" />
<PackageReference Include="Amazon.Lambda.Serialization.SystemTextJson" Version="2.4.4" />
<PackageReference Include="Amazon.Lambda.SQSEvents" Version="2.2.0" />
<PackageReference Include="AWSSDK.SQS" Version="4.0.1.2" />
<PackageReference Include="AWSSDK.BedrockRuntime" Version="3.7.400.3" />
<PackageReference Include="AWSSDK.SimpleSystemsManagement" Version="3.7.400.3" />
<PackageReference Include="Microsoft.Extensions.DependencyInjection" Version="9.0.8" />
<PackageReference Include="Microsoft.Extensions.Logging" Version="9.0.8" />

🗂️ Message Models (At a Glance)

Shared fields: title, description, tenderNumber, reference, contact info, province, supportingDocs, tags, AISummary

eTenders: id, status, datePublished, dateClosing, url
Eskom: source, publishedDate?, closingDate?
Transnet: institution, category, location, contactPerson
SARS: briefingSession, dates
SANRAL: category, region, fullNoticeText

Base model excerpt:

public abstract class TenderMessageBase
{
    public string Title { get; set; } = string.Empty;
    public string Description { get; set; } = string.Empty;
    public string TenderNumber { get; set; } = string.Empty;
    public List<SupportingDocument> SupportingDocs { get; set; } = new();
    public List<string> Tags { get; set; } = new();
    public string? AISummary { get; set; } = null; // Generated by Claude 3 Sonnet
    public abstract string GetSourceType();
}

📮 Queues

Queue	Purpose	URL	Groups (examples)	Notes
AIQueue	Incoming tender messages	AIQueue.fifo	etenderscrape, eskomtenderscrape, transnettenderscrape, sarstenderscrape, sanraltenderscrape	Retention 5d
TagQueue	AI-enhanced messages	TagQueue.fifo	Source-based grouping	Includes AI summaries + tags
FailedQueue	Error recovery	FailedQueue.fifo	Error-type grouping	Enriched error info

Full URLs:

SOURCE_QUEUE_URL="https://sqs.us-east-1.amazonaws.com/211635102441/AIQueue.fifo"
TAG_QUEUE_URL="https://sqs.us-east-1.amazonaws.com/211635102441/TagQueue.fifo"
FAILED_QUEUE_URL="https://sqs.us-east-1.amazonaws.com/211635102441/FailedQueue.fifo"

📈 Monitoring & Metrics

Watch these:

Lambda: Invocations, Duration, Errors, Throttles
SQS: ApproximateNumberOfMessages (queue depth)
Bedrock: Model invocation metrics, throttling events
Parameter Store: GetParameter API calls

Structured log example:

{
  "level": "Information",
  "message": "AI summary generated",
  "properties": { 
    "SourceType": "SANRAL", 
    "TenderNumber": "RFP-2025-001",
    "SummaryLength": 247,
    "BedrockDuration": 1.2
  }
}

Custom metric-style logs:

_logger.LogInformation("Batch done - Processed: {Processed}, AI-Enhanced: {Enhanced}, Failed: {Failed}, Duration: {Duration}ms",
  successCount, aiEnhancedCount, failedCount, processingDuration);

🛡️ Error Handling & Recovery

Bedrock throttling: Automatic retries with exponential backoff
Prompt fetching failures: Graceful degradation, detailed logging
AI processing errors: Messages routed to DLQ with context
Automatic retries via visibility timeout + Lambda retry behavior
Enriched error payloads for fast diagnosis

Redrive example:

aws sqs start-message-move-task \
  --source-arn arn:aws:sqs:us-east-1:211635102441:FailedQueue.fifo \
  --destination-arn arn:aws:sqs:us-east-1:211635102441:AIQueue.fifo

⚡ Performance & Scaling

Prompt caching: Reduces Parameter Store API calls
Bedrock concurrency control: Prevents overwhelming the AI service
ReadyToRun to reduce cold starts
Batch operations for send/delete
Async/await throughout the pipeline
Keep JSON compact (camelCase, no indentation)

Scaling:

Concurrency scales with demand and Bedrock capacity
Typical: dozens of AI-enhanced messages/minute per Lambda
Cost: pay-per-invocation + Bedrock token usage

🔒 Security

Data handling:

TLS in transit, SSE at rest on SQS and Parameter Store
Public tender data only
IAM least-privilege for Bedrock and Parameter Store access
Full audit trail via CloudWatch

Prompt security:

Parameter Store provides secure configuration management
Prompts can be updated without code deployments
Version control for prompt changes

📦 Deployment

This section covers three deployment methods for the AI Lambda Function (Tender Summarization Service). Choose the method that best fits your workflow and infrastructure preferences.

🛠️ Prerequisites

Before deploying, ensure you have:

AWS CLI configured with appropriate credentials 🔑
.NET 8 SDK installed locally
AWS SAM CLI installed (for SAM deployment)
Access to AWS Lambda, SQS, Bedrock, Parameter Store, and CloudWatch services ☁️
Visual Studio 2022 or VS Code with C# extensions (for AWS Toolkit deployment)
Claude 3 Sonnet model access in AWS Bedrock

🎯 Method 1: AWS Toolkit Deployment

Deploy directly through Visual Studio using the AWS Toolkit extension.

Setup Steps:

Install AWS Toolkit for Visual Studio 2022
Configure AWS Profile with your credentials in Visual Studio
Open Solution containing Sqs_AI_Lambda.csproj

Deploy Process:

Right-click the project in Solution Explorer
Select "Publish to AWS Lambda" from the context menu
Configure Lambda Settings:
- Function Name: AILambda
- Runtime: .NET 8
- Handler: Sqs_AI_Lambda::Sqs_AI_Lambda.Function::FunctionHandler
- Memory: 512 MB
- Timeout: 900 seconds (15 minutes)

Set Environment Variables:

FAILED_QUEUE_URL=https://sqs.{region}.amazonaws.com/{account-id}/FailedQueue.fifo
SOURCE_QUEUE_URL=https://sqs.{region}.amazonaws.com/{account-id}/AIQueue.fifo
TAG_QUEUE_URL=https://sqs.{region}.amazonaws.com/{account-id}/TagQueue.fifo

Configure IAM Role with required permissions:
- SQS: ReceiveMessage, DeleteMessage, GetQueueAttributes, SendMessage
- Bedrock: bedrock:InvokeModel for Claude 3 Sonnet
- Parameter Store: ssm:GetParameter on /TenderSummary/Prompts/*
- CloudWatch Logs: Standard logging permissions
Set up SQS Trigger manually after deployment

Parameter Store Setup:

Create these parameters before deployment:

# System prompt for Claude 3 Sonnet
aws ssm put-parameter \
    --name "/TenderSummary/Prompts/System" \
    --value "You are an AI assistant specialized in analyzing South African tender documents..." \
    --type "String"

# Source-specific prompts
aws ssm put-parameter \
    --name "/TenderSummary/Prompts/eTenders" \
    --value "Focus on government procurement requirements..." \
    --type "String"

aws ssm put-parameter \
    --name "/TenderSummary/Prompts/Eskom" \
    --value "Emphasize power generation and electrical infrastructure..." \
    --type "String"

aws ssm put-parameter \
    --name "/TenderSummary/Prompts/Transnet" \
    --value "Highlight logistics and transportation aspects..." \
    --type "String"

aws ssm put-parameter \
    --name "/TenderSummary/Prompts/SARS" \
    --value "Focus on tax administration and revenue services..." \
    --type "String"

aws ssm put-parameter \
    --name "/TenderSummary/Prompts/SANRAL" \
    --value "Emphasize road infrastructure and highway projects..." \
    --type "String"

Post-Deployment:

Test the function using the AWS Toolkit test feature
Monitor logs through CloudWatch integration
Verify Bedrock integration and prompt retrieval
Test SQS message processing

🚀 Method 2: SAM Deployment

Use AWS SAM for infrastructure-as-code deployment with the provided template.

Initial Setup:

# Install AWS SAM CLI
pip install aws-sam-cli

# Install .NET 8 SDK
# Download from https://dotnet.microsoft.com/download/dotnet/8.0

# Verify installations
sam --version
dotnet --version

Build and Deploy:

# Build the .NET 8 application
dotnet build -c Release

# Build the SAM application
sam build

# Deploy with guided configuration (first time)
sam deploy --guided

# Follow the prompts:
# Stack Name: ai-lambda-stack
# AWS Region: us-east-1 (or your preferred region)
# Parameter FailedQueueURL: https://sqs.{region}.amazonaws.com/{account-id}/FailedQueue.fifo
# Parameter SourceQueueURL: https://sqs.{region}.amazonaws.com/{account-id}/AIQueue.fifo
# Parameter TagQueueURL: https://sqs.{region}.amazonaws.com/{account-id}/TagQueue.fifo
# Confirm changes before deploy: Y
# Allow SAM to create IAM roles: Y
# Save parameters to samconfig.toml: Y

Environment Variables Setup:

The template already includes the required environment variables:

# Already configured in AILambda.yaml
Environment:
  Variables:
    FAILED_QUEUE_URL: https://sqs.us-east-1.amazonaws.com/{account-id}/FailedQueue.fifo
    SOURCE_QUEUE_URL: https://sqs.us-east-1.amazonaws.com/{account-id}/AIQueue.fifo
    TAG_QUEUE_URL: https://sqs.us-east-1.amazonaws.com/{account-id}/TagQueue.fifo

Pre-Deployment Parameter Store Setup:

# Create all required prompts before deploying
./scripts/setup-parameters.sh  # Create this script with parameter creation commands

Subsequent Deployments:

# Quick deployment after initial setup
dotnet build -c Release
sam build && sam deploy

Local Testing with SAM:

# Test function locally (requires Docker)
sam local invoke AILambda

# Start local SQS simulation
sam local start-api

SAM Deployment Advantages:

✅ Complete infrastructure management including SQS queues
✅ Environment variables defined in template
✅ IAM permissions automatically configured for Bedrock and Parameter Store
✅ Easy rollback capabilities
✅ CloudFormation integration
✅ SQS trigger automatically configured

🔄 Method 3: Workflow Deployment (CI/CD)

Automated deployment using GitHub Actions workflow for production environments.

Setup Requirements:

GitHub Repository Secrets:

AWS_ACCESS_KEY_ID: Your AWS access key
AWS_SECRET_ACCESS_KEY: Your AWS secret key
AWS_REGION: us-east-1 (or your target region)

Pre-existing Lambda Function: The workflow updates an existing function, so deploy initially using Method 1 or 2.
Parameter Store Setup: Ensure all prompts are configured in Parameter Store.

Deployment Process:

Create Release Branch:

# Create and switch to release branch
git checkout -b release

# Make your changes to the .NET code
# Commit changes
git add .
git commit -m "feat: update AI prompt processing logic"

# Push to trigger deployment
git push origin release

Automatic Deployment: The workflow will:
- Checkout the code
- Set up .NET 8 SDK
- Install AWS Lambda Tools
- Build and package the Lambda function using Sqs_AI_Lambda.csproj
- Configure AWS credentials
- Update the existing Lambda function code
- Maintain existing configuration (environment variables, triggers, etc.)

Manual Trigger:

You can also trigger deployment manually:

Go to Actions tab in your GitHub repository
Select "Deploy .NET Lambda to AWS" workflow
Click "Run workflow"
Choose the release branch
Click "Run workflow" button

Workflow Deployment Advantages:

✅ Automated CI/CD pipeline
✅ Consistent deployment process
✅ Audit trail of deployments
✅ Easy rollback to previous commits
✅ No local environment dependencies
✅ Automatic .NET build and packaging

🔧 Post-Deployment Configuration

Regardless of deployment method, verify the following:

Environment Variables Verification:

Ensure these environment variables are properly set:

# Verify environment variables via AWS CLI
aws lambda get-function-configuration \
    --function-name AILambda \
    --query 'Environment.Variables'

Expected output:

{
    "FAILED_QUEUE_URL": "https://sqs.us-east-1.amazonaws.com/{account-id}/FailedQueue.fifo",
    "SOURCE_QUEUE_URL": "https://sqs.us-east-1.amazonaws.com/{account-id}/AIQueue.fifo",
    "TAG_QUEUE_URL": "https://sqs.us-east-1.amazonaws.com/{account-id}/TagQueue.fifo"
}

Parameter Store Verification:

Verify all required prompts exist:

# Check system prompt
aws ssm get-parameter --name "/TenderSummary/Prompts/System"

# Check source-specific prompts
aws ssm get-parameter --name "/TenderSummary/Prompts/eTenders"
aws ssm get-parameter --name "/TenderSummary/Prompts/Eskom"
aws ssm get-parameter --name "/TenderSummary/Prompts/Transnet"
aws ssm get-parameter --name "/TenderSummary/Prompts/SARS"
aws ssm get-parameter --name "/TenderSummary/Prompts/SANRAL"

Bedrock Model Access:

Verify Claude 3 Sonnet access:

# List available foundation models
aws bedrock list-foundation-models --region us-east-1

# Test model access (if needed)
aws bedrock invoke-model \
    --model-id anthropic.claude-3-sonnet-20240229-v1:0 \
    --body '{"messages":[{"role":"user","content":"Test"}],"max_tokens":100}' \
    --cli-binary-format raw-in-base64-out \
    response.json

SQS Trigger Configuration:

Ensure the SQS trigger is properly configured:

# List event source mappings
aws lambda list-event-source-mappings \
    --function-name AILambda

# Verify batch size and queue configuration
aws lambda get-event-source-mapping \
    --uuid [event-source-mapping-uuid]

🧪 Testing Your Deployment

After deployment, test the function thoroughly:

Test AI Processing:

# Send test message to source queue
aws sqs send-message \
    --queue-url https://sqs.us-east-1.amazonaws.com/{account-id}/AIQueue.fifo \
    --message-body '{
        "title": "Test Tender for Infrastructure Development",
        "description": "This is a test tender for AI processing",
        "tenderNumber": "TEST-001",
        "source": "eTenders"
    }' \
    --message-group-id "etenderscrape" \
    --message-deduplication-id "test-$(date +%s)"

# Monitor function execution
aws logs tail /aws/lambda/AILambda --follow

Test Parameter Store Integration:

# Invoke function to test prompt retrieval
aws lambda invoke \
    --function-name AILambda \
    --payload '{"Records":[]}' \
    response.json

# Check response for prompt loading logs
cat response.json

Expected Success Indicators:

✅ Function executes without errors
✅ CloudWatch logs show successful prompt retrieval from Parameter Store
✅ Bedrock integration working (AI summaries generated)
✅ Messages properly routed to TagQueue with AI summaries
✅ No timeout or memory errors
✅ Proper error handling for failed messages
✅ SQS batch processing functioning correctly

🔍 Monitoring and Maintenance

CloudWatch Metrics to Monitor:

Duration: Function execution time for AI processing
Error Rate: Failed AI generation operations
Memory Utilization: RAM usage during Bedrock calls
SQS Metrics: Message processing rates and queue depths
Bedrock Metrics: Model invocation success rates and latencies
Parameter Store: GetParameter API call rates

Log Analysis:

# View recent logs
aws logs tail /aws/lambda/AILambda --follow

# Search for AI processing statistics
aws logs filter-log-events \
    --log-group-name /aws/lambda/AILambda \
    --filter-pattern "AI-Enhanced"

# Search for prompt loading issues
aws logs filter-log-events \
    --log-group-name /aws/lambda/AILambda \
    --filter-pattern "Prompt loading"

# Monitor Bedrock integration
aws logs filter-log-events \
    --log-group-name /aws/lambda/AILambda \
    --filter-pattern "Bedrock"

🚨 Troubleshooting Deployments

.NET 8 Runtime Issues

Issue: Function fails to start or throws runtime errors

Solution: Ensure proper .NET 8 configuration:

Verify the handler path: Sqs_AI_Lambda::Sqs_AI_Lambda.Function::FunctionHandler
Check that all NuGet packages are compatible with .NET 8
Ensure the project targets net8.0 framework
Verify all dependencies are included in the deployment package

Bedrock Access Issues

Issue: Cannot access Claude 3 Sonnet model

Solution: Verify Bedrock configuration:

Ensure the IAM role has bedrock:InvokeModel permissions
Check that Claude 3 Sonnet is available in your region
Verify model ARN: anthropic.claude-3-sonnet-20240229-v1:0
Test model access using AWS CLI
Check Bedrock service quotas and limits

Parameter Store Access Failures

Issue: Cannot retrieve prompts from Parameter Store

Solution: Debug Parameter Store configuration:

Verify IAM role has ssm:GetParameter permissions on /TenderSummary/Prompts/*
Check that all required parameters exist
Test parameter retrieval using AWS CLI
Ensure parameter names match exactly (case-sensitive)
Verify parameter type is "String"

SQS Message Processing Issues

Issue: Messages not being processed or routed incorrectly

Solution: Debug SQS configuration:

Verify SQS trigger is configured with correct batch size (10)
Check message format matches expected JSON structure
Ensure FIFO queue attributes are properly set
Verify MessageGroupId routing logic
Monitor dead letter queue for failed messages

Memory and Performance Issues

Issue: Function runs out of memory or times out

Solution: Optimize function performance:

Increase memory allocation (current: 512 MB)
Optimize AI processing batch size
Review Bedrock API call patterns
Monitor prompt caching effectiveness
Consider implementing connection pooling

Environment Variables Missing

Issue: Function cannot access required queue URLs

Solution: Set environment variables using AWS CLI:

aws lambda update-function-configuration \
    --function-name AILambda \
    --environment Variables='{
        "FAILED_QUEUE_URL":"https://sqs.{region}.amazonaws.com/{account-id}/FailedQueue.fifo",
        "SOURCE_QUEUE_URL":"https://sqs.{region}.amazonaws.com/{account-id}/AIQueue.fifo",
        "TAG_QUEUE_URL":"https://sqs.{region}.amazonaws.com/{account-id}/TagQueue.fifo"
    }'

Workflow Deployment Fails

Issue: GitHub Actions workflow errors

Solution:

Check repository secrets are correctly configured
Verify .NET 8 SDK is properly installed in workflow
Ensure AWS Lambda Tools installation succeeds
Check that Sqs_AI_Lambda.csproj exists in repository
Verify target Lambda function exists in AWS
Monitor workflow logs for specific error messages

Choose the deployment method that best fits your development workflow and infrastructure requirements. SAM deployment is recommended for development environments, while workflow deployment excels for production systems requiring automated CI/CD pipelines with AI model integration.

🧰 Troubleshooting

Missing prompts
- KeyNotFoundException → Check Parameter Store paths
Bedrock permission denied
- AccessDeniedException → Update IAM role for Bedrock
Model not available
- ValidationException → Verify Claude 3 Sonnet availability in region
Throttling issues
- ThrottlingException → Monitor Bedrock quotas, retry logic handles automatically
DLQ growth
- Investigate error patterns → fix root cause → redrive

Helpful commands:

aws logs filter-log-events --log-group-name "/aws/lambda/AILambda" --start-time $(date -d '1 hour ago' +%s)000
aws ssm get-parameter --name "/TenderSummary/Prompts/System"
aws bedrock list-foundation-models --region us-east-1
aws sqs get-queue-attributes --queue-url https://sqs.us-east-1.amazonaws.com/211635102441/AIQueue.fifo --attribute-names All

🗺️ Roadmap

Enhanced AI: Fine-tuning models for even better summary quality
Smart tagging: Implement content-based tagging using AI insights
Multi-language: Support for Afrikaans and other SA languages
Analytics: Summary quality metrics and feedback loops

Built with love, bread, and code by Bread Corporation 🦆❤️💻

Name		Name	Last commit message	Last commit date
Latest commit History 48 Commits
.github/workflows		.github/workflows
Converters		Converters
Interfaces		Interfaces
Models		Models
Properties		Properties
Services		Services
.gitattributes		.gitattributes
.gitignore		.gitignore
AILambda (1).yaml		AILambda (1).yaml
Function.cs		Function.cs
LICENSE		LICENSE
Readme.md		Readme.md
Sqs_AI_Lambda.csproj		Sqs_AI_Lambda.csproj
Sqs_AI_Lambda.sln		Sqs_AI_Lambda.sln
aws-lambda-tools-defaults.json		aws-lambda-tools-defaults.json

License

Bread-Corp/Function-SQS-AI

Folders and files

Latest commit

History

Repository files navigation