Amazon Bedrock

The F1 Root Cause Analysis (RCA) assistant is a sophisticated multi-agent system built on Amazon Bedrock Agents that implements an agentic Retrieval-Augmented Generation (RAG) architecture. The system operates through a network of specialized agents that handle different aspects of troubleshooting and system monitoring. At its core, it uses foundation models to break down user queries into actionable steps and create orchestration plans.

The system architecture consists of multiple integrated layers:
1. User Interface Layer: Accepts natural language queries from engineers
2. Orchestration Layer: Bedrock Agents decompose complex queries into subtasks
3. Integration Layer: Connects with multiple systems including:
  - Database systems (for metrics and connectivity checks)
  - AWS services (EC2 instance monitoring)
  - External tools (Datadog for application monitoring, Jira for ticket management)
  - Amazon Bedrock Knowledge Bases (for contextual information)
4. Security Layer: Implements strict security protocols including controlled SQL queries and principle of least privilege

The workflow typically follows this pattern:
1. Engineer submits a natural language query
2. Agents create an execution plan
3. Multiple agents simultaneously:
  - Query relevant databases
  - Check system metrics
  - Pull logs from knowledge bases
  - Verify network connectivity
4. Results are consolidated and presented to the user
5. If needed, automated ticket creation in Jira