April 2026 (2026-04) monthly update for langchain4j/langchain4j. Focused on improving Gemini usage visibility and reliability, with cross-model consistency and robust test coverage. Key features delivered: - Introduced GoogleAiGeminiTokenUsage as a dedicated TokenUsage subclass to expose Gemini-specific token usage metrics (cachedContentTokenCount and thoughtsTokenCount) for better visibility across chat, streaming chat, and image models. Major bugs fixed: - Resolved deserialization issues that dropped Gemini usage fields, ensuring the new token counts are surfaced consistently. This aligns with the intent to provide complete usage metadata (Closes #3219). Overall impact and accomplishments: - Provides accurate, actionable token usage data for Gemini models, enabling precise cost monitoring, usage analytics, and alerting. Cross-model consistency improves developer experience and reliability. - Maintained backwards compatibility with existing usage reporting while extending metadata coverage. Technologies/skills demonstrated: - Java class design and subclassing (TokenUsage), integration across multiple model types (chat, streaming chat, image). - Rigorous test coverage (unit/integration tests) validating new fields and end-to-end usage reporting across modules. - Adherence to PR hygiene and release readiness (no breaking changes, tests green across core/main modules).