Skip to main content

Release Summary

This release delivers significant system-wide improvements, focusing on enhanced scalability, resource management, and reliability across multiple services. Notable updates include advanced cost scanning orchestration, optimized data ownership in algorithmic processing, and robust handling of large prompt payloads in report generation. Several critical bug fixes and operational enhancements have been implemented to ensure improved performance and stability.

helmChart​

  • 🔧 Improved cost scanner scalability by integrating KEDA for dynamic scaling and increased RabbitMQ consumer timeout.
  • 🔧 Configured KEDA HTTP trigger to use RabbitMQ management API over TLS, leveraging existing secrets without embedding credentials.
  • 🔧 Enabled selection between legacy CronJob and KEDA Deployment via legacyCronjob.enabled, ensuring only one is active at a time.
  • 🔧 Added new configuration keys for batch size, pagination limits, and timing controls to enhance environment variable management.
  • 🔧 Adjusted terminationGracePeriodSeconds to allow for graceful shutdown of RabbitMQ and MongoDB connections.
  • 🔧 Provided operational guidance to manually clear and declare the cost-scan queue before deployment.
  • 🔧 Updated RabbitMQ consumer_timeout to align with scanner timeouts, preventing premature channel closures and ensuring broker-wide consistency.
  • 📣 Merged latest development changes.
  • 🔼 Updated dependencies.

python-algo​

  • 🔧 Scoped subscription algorithm runs to operate only on their respective data, reducing resource contention and improving performance for large tenants.
  • 🔧 Centralized ownership logic in utils/owner_scope.py and enhanced cleanup routines for orphaned relationships and node deletions.
  • 🔧 Optimized configuration reads to minimize memory usage and improved edge count accuracy by recalculating from source/target counts.
  • 🔧 Ensured secrets and gateway targets are restricted to account-owned nodes for improved security.
  • 🔧 Moved algorithm execution to a worker thread for improved message handling and heartbeat reliability.
  • ✨ Enabled cost scan triggering through the cost-scan queue, publishing persistent messages per account for asynchronous processing.
  • 🔧 Ensured idempotency in relationship and cost scan writers, allowing safe concurrent updates.
  • 📣 Documented deployment and rollback procedures, including queue management and edge count backfills.
  • 📣 Merged latest development changes.
  • 🔼 Updated dependencies.

reports-generator​

  • 🐛 Fixed Bedrock token overflow and out-of-memory crashes by splitting oversized prompts and merging results with robust fallback logic.
  • 🔧 Refined prompt handling to dynamically adjust token budgets and prevent data loss during report generation.
  • 🔧 Ensured that incomplete merges result in visible warnings and are excluded from caching to maintain data integrity.
  • 🔧 Standardized cost data sourcing from dailyaccountcostexplorers for AWS and Azure, ensuring accurate and consistent reporting.
  • 🔧 Transitioned budget strategy and documentation to updated models and removed deprecated agents.
  • 🔧 Updated README to reflect new split/merge logic and Bedrock defaults.
  • 🔧 Modified credential fetching to only request Azure credentials when necessary, preventing unnecessary failures on the Bedrock path.
  • 🔧 Implemented fail-fast token ceilings and concurrency caps to address memory pressure and improve reliability.
  • 🔧 Introduced truncation-based degrade paths for rightsizing operations, avoiding hard aborts.
  • 🔧 Cleaned up unused fields from LLM payloads and performed a comprehensive dependency update.
  • 📣 Merged latest development changes.
  • 🔼 Updated dependencies.

Disclaimer: The release notes are generated by OpenAI ChatGPT and may not be accurate. Please contact our support team for more information.

2026-Oct-03T21:39:02