Structured Output (Deep Scan)

When using Deep Scan mode, the model returns structured JSON responses with:

• Explanation — Detailed analysis of the code or graph
• Code Suggestions — Specific code changes with file paths and descriptions
• Severity — Assessment of the issue (Info, Warning, Error, Critical)
• Category — Classification (Architecture, Performance, Security, UX, etc.)
• Follow-up Prompts — Suggested next questions

This structured output is reconstructed into readable markdown for display. Streaming is disabled in this mode to ensure complete JSON parsing.

---