May 2026 monthly summary focusing on delivering stability and value across Pulsar (apache/pulsar) and Spring AI (spring-projects/spring-ai). The work emphasized reliability for long-running sessions, robustness of data operations, and enabling clearer model configuration for future iterations. Key features delivered: - DeepSeek v4 Chat Models: Added constants for the new DeepSeek v4 chat model options and documented usage, enabling smoother integration and configurability for chat features. Major bugs fixed: - Pulsar Client Idle Connection Release Safety: Fixed premature idle connection releases by validating session types before releasing connections, ensuring ongoing V5 watch sessions are not disrupted. Commit: 0a6b6f90bb68d1df426fc301c609c1d4751fe159. - Redis Vector Store Delete Robustness: Fixed delete operation by paginating search results to delete all matching documents, not just the first page. Commit: ec1e05282ddb6cb2a7875a685fca9a9df4d5eb9d. Overall impact and accomplishments: - Improved runtime stability for long-lived watch sessions and data integrity during bulk deletions. - Reduced risk of unintended session disruptions and partial data deletions. - Strengthened model configuration capabilities for DeepSeek v4, enabling faster experimentation and deployment. Technologies/skills demonstrated: - Java and Pulsar client programming - Redis vector store operations and pagination logic - Documentation and model configuration for DeepSeek v4 - Change validation and signing practices demonstrated by commit messages