Project: Synthetic Data Generator - Phase 1

Phase 1 Objective: Establish the foundational collaborative workspace where a User and an AI Agent can seamlessly co-create and edit Entity-Relationship (ER) database schemas in real-time.

1. Overview

Before generating synthetic data, the system needs a defined database schema. Phase 1 focuses entirely on building a "Live Sync" schema designer. The unique value proposition is the shared state: the user can build the diagram manually using a GUI, or instruct the AI via chat to build/modify it for them. Both operate on the exact same underlying data structure.

2. User Journey & UI Layout

1. Onboarding / Context Entry: The user lands on the application, enters their details (user_id), and defines the project (project_id, name, description). This creates the Context Packet.
2. Main Workspace: The user is taken to the main interface, split into two main sections:
   • Left Pane (Interactive Canvas): A visual ER diagram builder.
     • Users can manually add/remove tables.
     • Define column names and data types.
     • Establish Primary Key (PK) and Foreign Key (FK) relationships.
   • Right Pane (AI Chat Panel): A conversational interface.
     • Users can type natural language prompts (e.g., "Create an e-commerce schema with users, orders, and products" or "Add a 'status' column to the orders table").

3. Core Architecture & "Live Sync"

To achieve the seamless interaction between the User UI and the AI Agent, the system will use a centralized database as the single source of truth.

• Database: SQLite (for Phase 1 simplicity).
• State Management (schema_json): The entire visual representation and structure of the ER diagram is stored as a JSON object inside the schema_json column of the database.
• The Interaction Loop:
  • Manual Edit: User updates the canvas -> App updates the schema_json in SQLite -> Canvas re-renders.
  • AI Edit: User sends a chat message -> AI Agent processes intent -> AI calls an update_schema tool/function to modify the schema_json in SQLite -> UI detects the change ("Live Sync") and the Canvas updates automatically.

4. Database Schema (Phase 1)

Based on the architecture, the core SQLite table required to manage the shared workspace is the Projects Table:

5. Out of Scope for Phase 1 (Deferred to Phase 2)

• Generating synthetic rows using LLMs.
• Seeding bulk data using SDV (Synthetic Data Vault) or Faker.
• Exporting data to external databases.

---

Note: The primary technical challenge for Phase 1 will be establishing the real-time or near-real-time synchronization loop between the SQLite database updates (performed by the AI) and the frontend React/Vue canvas.