Here is the corrected content in a clean, copy-pasteable Markdown format. I’ve structured it with proper headers and bullet points to ensure it’s readable and professional.

# Asana RL Environment - Seed Data Generator

This project generates realistic, high-quality seed data for an Asana reinforcement learning environment. The data simulates a B2B SaaS company with 5,000–10,000 employees using Asana for product development, marketing, and operations.

## Overview
The generator creates a complete enterprise Asana workspace with:
* 1 Organization: A simulated B2B SaaS company.
* 25 Teams: Including Engineering, Product, Marketing, Sales, Customer Success, and Operations.
* 5,000–10,000 Users: Featuring realistic names from US Census data.
* 100+ Projects: Spanning different team workflows.
* 2,000+ Tasks: Using realistic naming patterns based on GitHub issues and Asana templates.
* Metadata: Includes comments, custom fields, tags, and attachments.

## Features

1. Data Realism
* Task Naming: Follows real-world patterns (e.g., "Implement OAuth 2.0 authentication flow" instead of "Task 1").
* Name Distribution: First and last names are pulled from US Census data.
* Research-Backed Distributions:Task Completion Rates: Based on project type (70–85% for sprints, 40–50% for ongoing tasks).
    * Due Date Patterns: 25% within 1 week, 40% within 1 month, 10% with no due date.
    * Assignment Rates: Approximately 15% unassigned, per Asana benchmarks.
* Temporal Consistency: Tasks are completed after creation; comments are spread across the task lifetime.
* Department-Specific Patterns: Engineering tasks follow a `[Component] - [Action]` pattern, while Marketing follows `[Campaign] - [Deliverable]`.

## Setup Instructions

1) Prerequisites
* Python 3.8 or higher
* pip

2) Installation
1. Clone or download the repository.
2. Install dependencies:
   ```bash
   pip install -r requirements.txt

3. (Optional) Set up API for LLM-generated content:

cp .env.example .env

Note: LLM integration is optional. The generator uses template-based generation by default, which produces equally realistic data without requiring an API key.

Running the Generator

python src/main.py

The script will:

1. Create a new SQLite database at output/asana_simulation.sqlite.
2. Generate all data with progress logging.
3. Display statistics on completion. Expected runtime: 2–5 minutes for the full dataset.

Configuration

Edit the CONFIG dictionary in src/main.py to adjust settings:

CONFIG = {
    'employee_count': 7500,        # Number of users (5000-10000)
    'output_db': 'output/asana_simulation.sqlite',
    'schema_file': 'schema.sql',
    'start_date': '2024-07-01',    # Start of data history
    'end_date': '2026-01-06',      # Current date
}

Database Schema

The schema follows Asana's entity model with proper foreign key relationships:

Core Tables

• organizations: Top-level workspace.
• teams: Groups within the organization.
• users: Team members with census-based names.
• team_memberships: Many-to-many user-team relationships.
• projects: Collections of tasks.
• sections: Subdivisions within projects (To Do, In Progress, Done).
• tasks: Core unit of work (supports subtasks via parent_task_id).

Metadata Tables

• comments: Task discussions and updates.
• custom_field_definitions: Project-specific field settings.
• custom_field_values: Actual values for custom fields on tasks.
• tags: Cross-project labels.
• task_tags: Task-tag associations.
• attachments: Simulated file metadata.

Key Design Decisions

• Custom Fields: Handled with separate tables for definitions (per-project) and values (per-task), allowing flexible field types (dropdown, text, number, date, checkbox).
• Task Hierarchy: Uses a single tasks table with a self-referential parent_task_id. Subtasks reference parents via this field (NULL for top-level tasks).
• Temporal Consistency: All timestamps are validated. Tasks cannot be completed before creation, and due dates respect business days (85% avoid weekends).

Data Quality Checks

Run these queries to verify data quality:

Check temporal consistency:

SELECT COUNT(*) FROM tasks WHERE completed_at < created_at; -- Should be 0

Verify assignment distribution:

SELECT ROUND(COUNT(CASE WHEN assignee_id IS NULL THEN 1 END) * 100.0 / COUNT(*), 2) as unassigned_pct FROM tasks; -- Should be ~15%

Verify due date distribution:

SELECT CASE 
    WHEN due_date IS NULL THEN 'No due date' 
    WHEN due_date < DATE('now') THEN 'Overdue' 
    WHEN due_date < DATE('now', '+7 days') THEN 'Within 1 week' 
    ELSE 'Other' END as cat, COUNT(*) FROM tasks GROUP BY cat;

Troubleshooting

• Inconsistent Timestamps: Check your system clock and verify start_date < end_date in CONFIG. Ensure end_date is not in the future.

License

All code is provided for evaluation purposes.