How vision-of-chosen/iac-drift-multi-agents works
Project type and purpose
• This is an open-source, terminal-based, multi-agent orchestration system written in Python.
• Its job is to detect Terraform infrastructure drift, analyze its severity, then automatically generate and apply remediation plans.
• It uses the strand-agents SDK and AWS services (Terraform CLI, AWS APIs, Anthropic Claude via AWS Bedrock).
File structure
• Top-level scripts and configuration
• terraform_drift_system.py – main entry point
• chat_interface.py – terminal chat UI
• config.py / prompts.py – global settings, system prompts
• shared_memory.py – blackboard/shared memory implementation
• requirements.txt, .env.example, setup.sh
• Agents directory
• orchestration_agent.py – central coordinator
• detect_agent.py – drift detection specialist
• drift_analyzer_agent.py – analysis & assessment
• remediate_agent.py – automated remediation
• Third-party SDKs and tools
• sdk-python/ – local install of strand-agents SDK
• tools/ – custom toolset wrappers (strand-tools, use_aws, file writers)
• Terraform workspace
• terraform/ – user’s .tf, .tfstate, variables, outputs
• terraform/*.tf scripts invoked by agents
• Infrastructure-as-code orchestration
• terraform_drift_system.py + terraform_drift_system.py calls into terraform/ directory
README insights
• Four specialized agents collaborate via a shared-memory (blackboard) pattern.
• Workflow:
User → Chat UI → OrchestrationAgent → {DetectAgent → SharedMemory → DriftAnalyzerAgent → SharedMemory → RemediateAgent} → OrchestrationAgent → User
• Integrations: Terraform CLI, AWS CLI/SDKs, Anthropic Claude (via AWS Bedrock), strand-tools.