PanSou Cloud Drive Resource Search API

PanSou is a high-performance cloud drive resource search API service. It supports searching via Telegram (TG) channels and custom plugins. Key features include:

• High-Performance Search: Concurrent execution of multiple TG channels and asynchronous plugin searches using a worker pool design.
• Cloud Drive Classification: Automatically identifies and categorizes various cloud drive links.
• Intelligent Sorting: A multi-dimensional algorithm based on plugin levels, time freshness, and priority keywords.
• Asynchronous Plugin System: Allows extending search sources via plugins with a "respond quickly, process continuously" model to handle slow search sources.
• Two-Level Caching: Uses a combination of sharded memory and sharded disk caching to improve repeat query speed and concurrency performance.

The SuSu search plugin is a component of the PanSou network disk search system. It is specifically designed to search and extract network disk resource links from the SuSu website (susuifa.com).

Key features include:

• Asynchronous Search: Implements a "respond quickly, process continuously" pattern based on BaseAsyncPlugin.
• Multi-level Caching: Reduces redundant network requests and computations.
• Concurrency Control: Uses goroutines and semaphores for efficient parallel searching.
• Smart Filtering: Filters irrelevant results early in the process.
• Resilience: Includes an exponential backoff retry strategy and fault-tolerant design that returns partial results even if some requests fail.

The PanSou system is organized into several functional layers:

• HTTP Service Layer (api/): Handles routing (router.go), request processing (handler.go), and middleware like logging and CORS (middleware.go).
• Search Service Layer (service/): Contains the core search logic and the mergeSearchResults function used to combine results from different sources.
• Plugin System Layer (plugin/): Defines the AsyncSearchPlugin interface and provides the BaseAsyncPlugin base class. Individual plugins reside in their own subdirectories (e.g., jikepan, pan666).
• Utility Layer (util/): Provides foundational services including the Two-Level Cache System (cache/), the Worker Pool for concurrency control (pool/), and various HTTP and parsing tools.

To correctly process Bixin forum data into PanSou standard format, follow this sequence:

1. Search Request: Send a request to the discussions API to get the list of discussions and included replies.
2. Parse Discussion Data: Extract metadata like title and time.
3. Associate Replies: Use IDs to link discussion posts with their corresponding replies (found in the included array).
4. Clean HTML: Use a cleanHTML function (similar to pan666) to remove HTML tags and entities, leaving only pure text.
5. Extract Links: Extract Mobile Cloud Drive links (caiyun.139.com) from the cleaned text.
6. Match Passwords: Use a position-based association strategy to match passwords/access codes with the extracted links.
7. Validate: Filter out invalid or broken links.
8. Format: Convert the final data into the PanSou standard result format.

When authentication is enabled (AUTH_ENABLED=true), all API endpoints except /api/auth/login and /api/health require a valid JWT Token.

To authenticate:

1. Call the login endpoint to receive a token.
2. Include the token in the Authorization header of subsequent requests using the Bearer <token> format.
3. Re-authenticate if the token expires.

# Example of an authenticated request
curl -X POST http://localhost:8888/api/search \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9..." \
  -d '{"kw":"速度与激情"}'

When implementing a scraper for Fox4K, adhere to these technical requirements:

• URL Construction:
  • Search: https://www.4kfox.com/search/{keyword}-------------.html (Keyword must be URL encoded).
  • Detail: https://www.4kfox.com/video/{ID}.html.
• Image Handling: The site uses lazy loading; always target the data-original attribute instead of src to get the actual image URL.
• Anti-Scraping: Set appropriate User-Agent and request headers to avoid being blocked.
• Encoding: The website uses UTF-8.
• ID Extraction: Use the regex /video/(\d+)\.html to parse the unique resource ID from URLs.

When implementing the daishudj plugin, ensure the following logic is applied:

• Search Logic: Request the HTML list and parse .item-jx for base info.
• Detail Logic: Scrape .article-body for <a> links and parse naked URLs from text.
• Extraction Code: Use keyword matching (提取码/密码/pwd/code) in the vicinity of the link.
• Date Formatting: Convert Chinese date formats (e.g., 2025年11月16日) to YYYY-MM-DD.
• Deduplication: Use the article ID (e.g., from /1047/) as part of the UniqueID to prevent duplicate results.