virnavi_ai_agent_compose 0.0.3

Ties Flutter UI to MCP tool results. Widgets rebuild reactively when AI agents invoke tools — no setState needed.

0.0.3

New features

• Nested model propagation — McpComposeBinding now accepts an optional McpSummary. When provided, after every successful tool call the binding walks the result model's nestedExtractors and stores each nested model's JSON in McpResultStore under its own model ID. Recurses for arbitrarily deep nesting. This means @McpView widgets bound to nested models (e.g. Dimensions inside Product) react automatically without a separate tool call.

---

0.0.1

Initial release.

• McpToolState — sealed state type: McpIdle, McpLoading, McpSuccess(data), McpError(message).
• McpResultStore — ChangeNotifier that holds per-tool state; updates reactively as AI agents invoke tools.
• McpComposeBinding — wraps ToolDefinition handlers to drive McpResultStore through idle → loading → success | error transitions automatically.
• McpResultBuilder — Flutter widget backed by ListenableBuilder that rebuilds whenever a specific tool's state changes.
• McpViewDefinition — pairs a model ID with a Widget Function(Object?) builder; attached to McpSummary via McpSummaryViewsX extension.
• McpSummaryViewsX extension on McpSummary — bindViews(), views, buildView() using an Expando to avoid circular dependencies.
• McpView annotation re-exported from virnavi_ai_agent_mcp — app code only needs to import compose.
• List<ToolDefinition>.registerWith(bridge, binding) convenience extension.