create_datasets.py

import pandas as pd
from huggingface_hub import HfApi, login
import os

# --- CONFIGURATION ---
HF_USERNAME = "gberseth"  # <--- REPLACE THIS
HF_TOKEN = os.environ.get("HF_TOKEN")

# Define Repo IDs
REQUESTS_REPO = f"{HF_USERNAME}/rl-leaderboard-requests"
RESULTS_REPO = f"{HF_USERNAME}/rl-leaderboard-results"

# Authenticate
# login(token=HF_TOKEN)
api = HfApi()

def create_dataset(repo_id, csv_filename, columns):
    print(f"Processing {repo_id}...")
    
    # 1. Create the Repository on Hub (if it doesn't exist)
    try:
        api.create_repo(repo_id=repo_id, repo_type="dataset", exist_ok=True)
        print(f"✅ Repo '{repo_id}' created (or already exists).")
    except Exception as e:
        print(f"❌ Error creating repo: {e}")
        return

    # 2. Create a local empty CSV with headers
    # We create a dummy row to ensure columns are recognized, then drop it if preferred,
    # but an empty dataframe with columns is standard.
    df = pd.DataFrame(columns=columns)
    df.to_csv(csv_filename, index=False)
    
    # 3. Upload the CSV
    try:
        api.upload_file(
            path_or_fileobj=csv_filename,
            path_in_repo=csv_filename, # e.g. "requests.csv"
            repo_id=repo_id,
            repo_type="dataset"
        )
        print(f"✅ Initial '{csv_filename}' uploaded to {repo_id}.")
    except Exception as e:
        print(f"❌ Error uploading file: {e}")
    
    # Cleanup local file
    if os.path.exists(csv_filename):
        os.remove(csv_filename)

# --- EXECUTION ---

# 1. Setup Requests Dataset
# Columns used by Frontend and Backend to track queue
create_dataset(
    repo_id=REQUESTS_REPO, 
    csv_filename="requests.csv", 
    columns=["model_id", "status", "created_at", "submitted_by"]
)

# 2. Setup Results Dataset
# Columns used by Backend to store scores
create_dataset(
    repo_id=RESULTS_REPO, 
    csv_filename="results.csv", 
    columns=["model_id", "mean_reward", "run_time", "status", "completed_at"]
)

print("\n🎉 Setup complete! You can now link these datasets in your app.py files.")