mcp_remote_daily.py

#!/usr/bin/env -S uv run
# /// script
# requires-python = ">=3.10"
# dependencies = [
#     "datasets",
#     "pandas",
# ]
# ///

"""
Analyze daily proportion of sessions that include mcp-remote.

This script analyzes the evalstate/hf-mcp-logs dataset to calculate what proportion
of sessions included "mcp-remote" in the client name on a daily basis.

Usage:
    uv run mcp_remote_daily.py [--limit N]
"""

import sys
import re
import argparse
from datetime import datetime
from collections import defaultdict
from datasets import load_dataset


def main():
    parser = argparse.ArgumentParser(description="Analyze daily mcp-remote usage")
    parser.add_argument("--limit", type=int, default=None, help="Limit processing to N rows (for testing)")
    args = parser.parse_args()

    print(f"[{datetime.now().isoformat()}] Loading dataset from evalstate/hf-mcp-logs...", file=sys.stderr)

    # Load the source dataset (streaming for large datasets)
    ds = load_dataset('evalstate/hf-mcp-logs', 'sessions', streaming=True)
    sessions_ds = ds['sessions']

    # Pattern to detect mcp-remote suffix (any version)
    MCP_REMOTE_PATTERN = re.compile(r'\(via mcp-remote')

    # Track by day
    daily_stats = defaultdict(lambda: {'total': 0, 'with_mcp_remote': 0})

    total_rows = 0
    start_time = datetime.now()

    print(f"[{datetime.now().isoformat()}] Processing sessions...", file=sys.stderr)

    for batch in sessions_ds.iter(batch_size=10000):
        batch_len = len(batch['time'])
        total_rows += batch_len

        # Progress indicator
        if total_rows % 100000 == 0:
            elapsed = (datetime.now() - start_time).total_seconds()
            rate = total_rows / elapsed if elapsed > 0 else 0
            print(f"[{datetime.now().isoformat()}] Progress: {total_rows:,} rows processed "
                  f"({rate:.0f} rows/sec)", file=sys.stderr)

        if args.limit and total_rows >= args.limit:
            print(f"[{datetime.now().isoformat()}] Reached limit of {args.limit:,} rows", file=sys.stderr)
            break

        for i in range(batch_len):
            time_str = batch['time'][i]
            name = batch['name'][i]

            # Skip if name is None
            if name is None:
                continue

            # Extract date from time string (ISO format: 2025-11-19T...)
            date_str = time_str.split('T')[0] if 'T' in time_str else time_str

            daily_stats[date_str]['total'] += 1

            # Check if name contains mcp-remote suffix
            if MCP_REMOTE_PATTERN.search(name):
                daily_stats[date_str]['with_mcp_remote'] += 1

    elapsed = (datetime.now() - start_time).total_seconds()
    rate = total_rows / elapsed if elapsed > 0 else 0

    print(f"[{datetime.now().isoformat()}] Processing complete: {total_rows:,} rows", file=sys.stderr)
    print(f"Rate: {rate:.0f} rows/sec, elapsed: {elapsed:.1f} seconds", file=sys.stderr)
    print(f"Found {len(daily_stats)} days of data", file=sys.stderr)
    print()

    # Print daily statistics
    print("Daily mcp-remote Usage")
    print("=" * 70)
    print(f"{'Date':<12} {'Total Sessions':>15} {'mcp-remote':>12} {'%':>10}")
    print("-" * 70)

    # Sort by date
    for date in sorted(daily_stats.keys()):
        stats = daily_stats[date]
        total = stats['total']
        with_mcp = stats['with_mcp_remote']
        pct = (with_mcp / total * 100) if total > 0 else 0
        print(f"{date:<12} {total:>15,} {with_mcp:>12,} {pct:>9.1f}%")

    print("=" * 70)

    # Overall summary
    overall_total = sum(s['total'] for s in daily_stats.values())
    overall_with_mcp = sum(s['with_mcp_remote'] for s in daily_stats.values())
    overall_pct = (overall_with_mcp / overall_total * 100) if overall_total > 0 else 0

    print(f"{'TOTAL':<12} {overall_total:>15,} {overall_with_mcp:>12,} {overall_pct:>9.1f}%")
    print()


if __name__ == '__main__':
    main()