2.0.36

[acharneski]

This pull request introduces several improvements and refactorings across the codebase, primarily focusing on agent and task orchestration. The most significant change is the addition of the new ParsedImageAgent class, which enables parsing structured responses from image-and-text inputs. The refactoring also standardizes access to default and parsing chat interfaces within tasks, improving code maintainability and clarity. Minor updates include a library version bump and enhanced logging for Gemini SDK requests.

New agent functionality

• Added the ParsedImageAgent class to support parsing structured responses from image-and-text inputs, including robust JSON response handling and integration with type describers. (jo-penai/src/main/kotlin/com/simiacryptus/cognotik/agents/ParsedImageAgent.kt)

Refactoring and standardization of chat interface usage

• Introduced defaultChatter and parsingChatter properties to AbstractTask, and refactored all task classes to use these properties for consistent model selection and child client creation. [1] [2] [3] [4] [5] [6] [7] [8] [9]

Task type registration

• Registered the constructor for GenerateSpriteSheetTask in the TaskType registry, enabling dynamic instantiation of this task type. (webui/src/main/kotlin/com/simiacryptus/cognotik/plan/TaskType.kt)

Logging improvements

• Enhanced Gemini SDK request logging to include the system prompt, aiding debugging and traceability. (jo-penai/src/main/kotlin/com/simiacryptus/cognotik/chat/GeminiSdkChatClient.kt)

Version update

• Updated the library version from 2.0.35 to 2.0.36 in gradle.properties.