A simple dashboard to track, log, and visualize daily download counts for the portable version of ComfyUI.

This project fetches release and asset download data from the ComfyUI GitHub repository, stores it in a local SQLite database, aggregates daily download totals, and presents interactive charts through a Next.js dashboard.

• Automated Data Fetching: Retrieve GitHub release and asset download counts via a Node.js script.
• Local Persistence: Store raw data in SQLite (data/downloads.db) and maintain a daily_summary table for easy trend analysis.
• Interactive Dashboard: Visualize download metrics with a responsive React interface using Recharts.
• Scheduled Updates: Leverage GitHub Actions to run the fetch script daily at 10:49 UTC and commit updates.
• No External Dependencies: Everything runs locally without requiring external database services.

• Next.js 15 (App Router) + React + TypeScript
• Tailwind CSS for styling
• Recharts for data visualization
• Better-SQLite3 for database interactions
• GitHub Actions for scheduled data updates

• Node.js v22 or newer (tested with v22)
• pnpm as package manager (v10 recommended)

git clone https://github.com/Comfy-Org/download-data.git
cd download-data
pnpm install

Run the data fetching script to create or update the SQLite database:

pnpm getdata
# or: node scripts/fetch-data.mjs

This populates data/downloads.db and updates daily summaries.

Start the development server:

pnpm dev

Open http://localhost:3000 to view the dashboard.

pnpm build
pnpm start

The asset_daily_stats table tracks daily download counts for each GitHub asset, including flags for draft and prerelease status.

CREATE TABLE IF NOT EXISTS asset_daily_stats (
  asset_id INTEGER NOT NULL,
  asset_name TEXT NOT NULL,
  date TEXT NOT NULL,
  download_count INTEGER NOT NULL,
  draft INTEGER NOT NULL,
  prerelease INTEGER NOT NULL,
  fetch_timestamp TEXT DEFAULT (strftime('%Y-%m-%dT%H:%M:%fZ', 'now')),
  PRIMARY KEY (asset_id, date)
);

The daily_summary table stores the daily delta (net new downloads) across all assets.

CREATE TABLE IF NOT EXISTS daily_summary (
  date TEXT PRIMARY KEY,
  downloads_delta INTEGER NOT NULL,
  fetch_timestamp TEXT DEFAULT (strftime('%Y-%m-%dT%H:%M:%fZ', 'now'))
);

Records when the fetcher backfills missing days.

CREATE TABLE IF NOT EXISTS backfill_events (
  id INTEGER PRIMARY KEY AUTOINCREMENT,
  start_date TEXT NOT NULL,
  end_date TEXT NOT NULL,
  strategy TEXT NOT NULL,
  lookback_days INTEGER,
  trend_window INTEGER,
  noise_scale REAL,
  total_delta INTEGER NOT NULL,
  created_at TEXT DEFAULT (strftime('%Y-%m-%dT%H:%M:%fZ', 'now'))
);

If the fetch process misses days, the script evenly distributes the total change between the last available snapshot and today across all missing days. This avoids one-day spikes when resuming after a gap.

Environment variables:

• BACKFILL_STRATEGY: even (default), pattern, stochastic, or none.
  • even: distribute evenly across the gap.
  • pattern: replay recent variability and scale to the gap’s total.
  • stochastic: synthesize a series using recent trend and noise, then scale to match the gap’s total (deterministic with seed).
  • none: disable backfill; use only the previous day’s diff.
• BACKFILL_MIN_GAP_DAYS: minimum number of missing days to trigger backfill (default: 2).
• BACKFILL_LOOKBACK_DAYS: number of days to learn the pattern shape from for pattern (default: 30).
• BACKFILL_PATTERN_FALLBACK: behavior when pattern is unavailable or invalid; even (default) or fail.

Stochastic options:

• BACKFILL_TREND_WINDOW: moving-average window for trend estimation (default: 7).
• BACKFILL_NOISE_SCALE: multiplier for noise standard deviation derived from recent daily changes (default: 1).
• BACKFILL_RANDOM_SEED: optional fixed seed; defaults to a deterministic seed derived from dates.
• BACKFILL_STOCHASTIC_FALLBACK: even (default), pattern, or fail.

A workflow in .github/workflows/fetch-data.yml is configured to:

1. Run the fetch script daily at 10:49 UTC.
2. Commit changes to data/downloads.db if new data is fetched.

The dashboard shades backfilled date ranges and shows a short note above the chart. Data is served via GET /api/backfill/events?timeframe=... and rendered as subtle ReferenceArea overlays.

Data is assigned to a canonical daily “bucket” defined by a UTC cutoff time.

• DAILY_CUTOFF_UTC: HH:mm UTC string (default: 10:49). If a run occurs before this time, data is bucketed to the previous date; otherwise to the current date.
• BUCKET_WRITE_MODE: once (default) or replace.
  • once: if asset_daily_stats already has rows for the bucket date, the script skips writing asset rows again. This prevents manual, late-day runs from inflating the current bucket and reducing the following day’s delta.
  • replace: always overwrite the bucket’s asset rows.
• SUMMARY_WRITE_MODE: once (default) or replace.
  • Controls writes to daily_summary. In once, INSERT OR IGNORE preserves the first summary written for a bucket.

Tip: Keep BUCKET_WRITE_MODE=once and SUMMARY_WRITE_MODE=once to maintain consistent day-over-day deltas even if you trigger manual runs mid‑day.

Contributions, bug reports, and feature requests are welcome! Please open an issue or submit a pull request.

This project is licensed under the GPL-3.0 license.

• Built with Next.js, React, Tailwind CSS, and Better-SQLite3.