This release adds two improvements to the platform's AI chat surface. Chat requests now include structured context — query intent, the active database schema, session parameters, and backend metadata — giving agent models the precise data they need to reason about application state without constructing context from untyped text. Responses render Markdown, so agent replies that include code blocks, numbered steps, or formatted tables display correctly in the chat interface rather than as raw markup.
Structured request context
- Typed context objects. Each chat request carries a well-typed context block containing the query under construction, the active schema, the user session, and database connection metadata. Agent models receive this as structured input rather than extracting it from the conversation thread.
- Database metadata in agent configuration. The AI agent receives database-specific parameters at initialisation — engine type, active schema, connection constraints — enabling responses that reference the actual database context rather than generic SQL patterns.
- Schema-aware reasoning. With schema metadata present, agent responses can reference real table names, column types, and relationships from the live data model.
Markdown rendering
- Response rendering. Agent responses that include Markdown — code blocks, headers, bold labels, bullet lists — render visually in the chat interface rather than displaying raw markup syntax.
- Code block support. SQL snippets, script fragments, and configuration examples in agent replies appear in formatted monospace code blocks with correct line breaks.
Both changes improve the utility of AI-assisted query building and data exploration: the agent now responds with accurate, data-model-aware answers and presents them in a format that is readable directly in the interface without further editing.