Internal RAG Chatbot

To assist Business Analysts and Citizen Developers who may not use external IDEs, Struktural integrates an internal Chatbot directly into the Studio interface. This chatbot relies on Retrieval-Augmented Generation (RAG) to provide contextual answers based on this Wiki and the current application schema.

Architecture

The internal Chatbot utilizes the RuntimeAiController and the IAiProvider factory configured in the platform's global settings (supporting Azure OpenAI, Anthropic, or Google Gemini).

  1. User Inquiry: A user asks a question in the Studio chat (e.g., "How do I configure a Master-Detail view for my Departments?").
  2. Intent Analysis: The AI provider analyzes the request using the AnalyzeRequirementsAsync system prompt to determine the complexity and intent of the question.
  3. Retrieval (RAG): The system invokes the internal MCP tools programmatically to fetch the relevant Markdown documentation from the Wiki (in this case, the advanced-ui.md content covering Tree and Linked Nested Forms).
  4. Context Injection: The raw Markdown content, along with the current JSON schema of the application, is injected into the LLM's context window.
  5. Generation: The LLM generates a highly specific, accurate response based strictly on the provided Struktural documentation, eliminating hallucinations regarding non-existent features.

Maintaining the Knowledge Base

Because the RAG Chatbot reads directly from the Markdown files located in the Wiki/ directory, any updates made to this documentation are instantly available to the AI.

If new features are added to the Struktural engine (e.g., a new Workflow Node type), administrators simply need to update the Wiki/workflows/node-dictionary.md file. The Chatbot will immediately begin utilizing the new information in its subsequent responses without requiring fine-tuning or vector database rebuilds.