March 2026 monthly summary for sxjeru/lobe-chat focusing on improving robustness of tool type resolution and test coverage for suffix-stripped model names. Key feature delivered: ToolNameResolver now recovers correct tool types from manifests even when model names have suffixes stripped, with an accompanying regression test validating the behavior. Major bug fix: context-engine now reliably recovers tool types from manifests when models strip suffixes (edge cases including GLM models). Impact: higher reliability of tool resolution, reduced runtime errors, and better compatibility with diverse model naming conventions. Technologies demonstrated: manifest parsing, ToolNameResolver, context-engine, regression testing, test-driven development. Business value: fewer manual interventions, smoother deployments, and improved end-user tooling.