Add guardrail tracing

[KyMidd]

Human notes

• Add guardrail tracing option, surfaces why guardrail blocked things to the user in the slack response
• Fix system prompt injection

Robot notes beep boop

This pull request introduces several updates to the python/devopsbot.py file to enhance functionality, improve error handling, and refine AI interaction. Additionally, there are changes to the Terraform configuration in lambda_receiver.tf to standardize IAM role naming. Below is a categorized summary of the most important changes.

Enhancements to AI Interaction and Guardrails

• Introduced guardrailTracing to enable tracing options for guardrails (enabled, enabled_full, disabled) and included it in the AI request payload (def ai_request) to enhance security and debugging capabilities. [1] [2]
• Replaced model_guidance with system_prompt and updated the initialization process to use initial_model_system_prompt for better clarity and maintainability. [1] [2]
• Enhanced Slack responses to include detailed guardrail blocking information, such as type, confidence, and filter strength, when content is flagged by Veradigm's content filter.

Error Handling Improvements

• Added exception handling in ask_bedrock_llm_with_knowledge_base to catch and log errors, including specific handling for Aurora auto-pausing scenarios, with user-friendly Slack messages prompting retries. [1] [2]
• Added debug logging for raw and structured knowledge base responses when VERA_DEBUG is enabled.

Refactoring for Consistency

• Renamed url fields to source in knowledge base responses to better represent diverse data sources (e.g., Confluence, S3). Updated all related references in rerank_text and handle_message_event. [1] [2] [3]
• Updated AI request conversation structure to append assistant responses instead of user responses for knowledge base citations. [1] [2]

Terraform Configuration Updates

• Standardized IAM role naming in lambda_receiver.tf by renaming Ue1TiDevOpsBotReceiverRole to DevOpsBotReceiverRole across all related resources for improved clarity and consistency. [1] [2] [3] [4]

These changes collectively improve the bot's robustness, maintainability, and user experience while aligning infrastructure naming conventions.