A TypeScript/Node.js service that schedules and executes waifu2x-ncnn-vulkan image upscaling jobs with PostgreSQL-backed caching, filesystem storage, configurable concurrency, web page window, and a simple test client.

Mainly used for custom processImage in venera.

• Express 5 API exposing /upscale and /download endpoints with rate limiting, auth hooks, and structured log output.
• Task orchestration powered by a PostgreSQL table (waifu2x_tasks), a file cache, and a dynamic p-queue scheduler honoring FIFO/LIFO strategies.
• Cache deduplication that reuses completed results and in-flight jobs using deterministic cache keys.
• Runtime configuration hot-reloaded from config.json, covering waifu2x parameters, scheduler tuning, logging, and filesystem paths.
• Real-time log page served at /log, including optional token auth, live streaming output, and client-side persistence for gated access.
• Visual Upscaling Studio hosted at /studio, providing drag-and-drop uploads, parameter tuning (scale/noise/model/format/tile/gpu/threads, TTA), queue status, and auto-download flows.
• Test utilities for exercising the API (test/processImage.js, test/test_database.js) and resetting the database (test/reset_database.js).

• Node.js 18+
• pnpm 8+
• PostgreSQL 14+ running and reachable via the configured connectionString
• waifu2x-ncnn-vulkan binary (or compatible) accessible at the path defined in config.json

1. Install dependencies

   pnpm install

2. Configure the service
   • Copy config.example.json to config.json.
   • Adjust:
     • api host/port and optional maxInputSizeMB.
     • postgres.connectionString (and ssl if needed).
     • waifu2x executable path and default upscale parameters.
     • scheduler settings (concurrency, queueStrategy, maxQueueSize, taskTimeoutMs).
     • cache.ttlSeconds, filesystem paths.
3. Prepare PostgreSQL
   • Ensure the target database exists and the connectionString user has privileges.
   • Initialize/reset the schema with (auto-initialize on startup):

     node test/reset_database.js

4. Build dist

   pnpm build

5. Run the service With pm2 (recommended):

   pnpm start

   Or directly:

   node .

6. Interact with the API
   • POST binary data to http://<host>:<port>/api/upscale with required query params (sourceKey, cid, etc.).
   • Poll .../download/<taskId> until the result is ready (legacy .../download?taskId=... remains supported).
   • Use test/processImage.js as a reference client for venera.

The custom logger (log4js + chalk) emits color-coded, bracketed tags to differentiate subsystems. Log level hot-reloads when config.json changes.

• Visit http://<host>:<port>/log to watch structured logs live; append ?token=<config.api.logToken> (and set the token in config.json) to gate access. The page remembers your token in localStorage for future reloads.
• Five consecutive invalid tokens from the same IP will push that IP into config.api.blackIPs, so share URLs carefully.

• Open http://<host>:<port>/studio to access the drag-and-drop interface. It supports file validation, parameter tweaking (scale/noise/model/format/tile/gpu/threads), TTA toggles, queue status visualization, and automatic polling with download + copy-link actions.
• Just for fun. It is quite cool.

• test/processImage.js demonstrates how to call the API in Venera.
• test/test_database.js confirms DB connectivity and prints query timing.
• test/reset_database.js safely rebuilds the schema.
• test/test_waifu2x.js test script for waifu2x upscaling functionality.