Feature/uiux conversational editor

[DhairyaOG]

Pull Request Description

Title

feat: integrate Gemini 2.5 Flash conversational UI/UX editor with React Flow canvas

Overview

This PR completely overhauls the UI/UX Agent generation pipeline by ripping out the legacy static array responses and replacing them with a highly dynamic, interactive Conversational Editor powered by Gemini 2.5 Flash and React Flow.

Key Architectural Changes:

Visual React Flow Engine:

• Built a recursive React DOM rendering engine (UIRenderer.tsx) that translates Gemini's complex JSON component_tree layout directly into an interactive canvas using @xyflow/react.
• Injected a robust CSS Sanitizer to automatically catch and camelCase vendor-prefixed styles (like -webkit-background-clip) to prevent React crashes on premium designs.

Conversational Chat Proxy (POST /api/generate):

• Wired the floating ChatSidebar directly into a new FastAPI backend route to act as a secure proxy.
• Solved LLM "Amnesia" by continuously feeding the current_spec context directly back into Gemini's prompt, forcing it to modify the existing design rather than hallucinate a new one.

Data Persistence & Orchestration:

• Conversational edits are now actively written back to PostgreSQL via _save_artifact so that data is not lost on page refresh.
• Tightened the downstream SQL queries in Node (agentOrchestrator.js) with an explicit ORDER BY created_at ASC rule. Downstream agents (Security, Compliance) are now guaranteed to inherit the latest conversational UI/UX edit as their core context payload.

Resiliency:

• Completely rewrote the DEMO_MODE fallback in uiux_agent.py so that it returns a valid, visual React Flow error screen rather than crashing the canvas if an API key is missing.
• Confirmed GEMINI_API_KEY is securely injected via os.getenv() with absolutely zero hardcoded secrets.