OPS Agent

OPS Agent will help you to analyze and debug ops (live operations) issues.

Services

• LocalStack: Emulates AWS services locally
  • SQS queues and DLQs defined in dlqs.json
• OpenSearch: Search and analytics engine
• OpenSearch Dashboards: UI for interacting with OpenSearch
• mcp-server-sse: Server-Sent Events implementation for MCP communication
• mcp-api-server: API server for handling requests
• ops-ui: React-based frontend application

Getting Started

Prerequisites

• Docker and Docker Compose
• AWS CLI (optional for manual testing)
• Ollama - for local LLM processing

Setting Up Ollama

1. Download and install Ollama from ollama.ai
2. Start the Ollama service
3. Pull the required model:

ollama pull qwen3:8b

Running the Environment

1. Build and start all services with a single command:

docker-compose up --build

This command will:

• Build all required Docker images
• Start the containers
• Initialize SQS queues and populate them with sample data via the migration service
• Load sample logs into OpenSearch
• Set up all necessary infrastructure for the OPS Agent to work

1. Check if services are running:

docker-compose ps

1. Access the UI at http://localhost:3000 and execute queries

Accessing Services

• LocalStack AWS services: http://localhost:4566
• OpenSearch: http://localhost:9200
• OpenSearch Dashboards: http://localhost:5601
• Ops UI: http://localhost:3000
• MCP API Server: http://localhost:3002
• MCP SSE Server: http://localhost:3001

Working with the Agent

OPS Agent provides two operational modes to analyze and solve problems:

Agent Without Thinking Mode

In the standard mode, the agent processes your queries and returns the final results without showing the internal reasoning process:

This mode is optimized for quick answers and a cleaner interface.

Agent With Thinking Mode

When the thinking mode is enabled, OPS Agent shows its complete reasoning process, displaying each step of its analysis:

The thinking mode reveals:

• Internal thought processes
• Tools being used
• Step-by-step reasoning
• Intermediate results

Agent Thinking Process

The agent uses a structured approach to solve problems:

This includes:

1. Parsing and understanding your query
2. Planning the steps to address the issue
3. Selecting and using appropriate tools
4. Analyzing data from multiple sources
5. Formulating a response based on collected information

Agent Internal Process with Tools

The agent leverages various tools to gather and process information:

These tools include:

• Log retrieval and analysis
• Queue inspection
• Trace correlation
• Pattern recognition
• Data summarization

Working with SQS Queues

The following queues and Dead Letter Queues (DLQs) are automatically created from dlqs.json:

• order-queue → order-queue-dlq
• order-cancellation-queue → order-cancellation-queue-dlq

Testing SQS Locally

# List queues
aws --endpoint-url=http://localhost:4566 sqs list-queues

# Send a message to a queue
aws --endpoint-url=http://localhost:4566 sqs send-message \
  --queue-url http://localhost:4566/000000000000/order-queue \
  --message-body '{"orderId": "123456", "status": "created"}'

# Receive messages from a queue
aws --endpoint-url=http://localhost:4566 sqs receive-message \
  --queue-url http://localhost:4566/000000000000/order-queue

Working with OpenSearch

Sample logs have been preloaded into the OpenSearch instance. You can access them via the OpenSearch API or using OpenSearch Dashboards.

Sample Query

# Search all logs
curl -X GET "http://localhost:9200/cwl-logs/_search" -H 'Content-Type: application/json' -d '{
  "query": {
    "match_all": {}
  }
}'

# Search for error logs
curl -X GET "http://localhost:9200/cwl-logs/_search" -H 'Content-Type: application/json' -d '{
  "query": {
    "match": {
      "level": "ERROR"
    }
  }
}'

Using the Ops UI

The Ops UI provides a user-friendly interface for interacting with all services:

1. Access the UI at http://localhost:3000
2. Use the interface to:
   • Monitor SQS queues and DLQs
   • Search logs in OpenSearch
   • Perform operations through the MCP API
   • Toggle between standard and thinking modes

Troubleshooting

If you encounter connectivity issues with SQS services:

• Ensure the initialization script has been run to create the queues
• Check LocalStack logs with docker logs ops_agent-localstack-1
• Verify the queue URLs are correctly formatted

If Ollama isn't connecting:

• Ensure Ollama is running on your host machine
• Check that the qwen3:8b model has been pulled
• Verify that host.docker.internal is accessible from containers

Shutting Down

To stop all services:

docker-compose down

To remove all data (including volumes):

docker-compose down -v