SuperAPI AI Directory
Partner with us中文AI newsFavorites
Skillsmodel-resources-collection

model-resources-collection

Coding & development

View repository

Updated 10-15 20:05

About this skill

Model resource collection

Use this skill when users need to find free or discounted AI model resources.

Trigger conditions

What users might say:

  • What cheap models do you recommend?
  • Help me find free AI APIs
  • What free models can I use?
  • Help me organize the current discounted models
  • Recommend some free model APIs
  • Whose models are the cheapest?
  • Help me search for what model discounts are available now

Workflow

Important principle: sources.md must be processed item by item

Each time a task is executed, the following steps must be followed:

  1. Read sources.md: First read completely references/sources.md, to obtain the list of all sources to be searched
  2. Create a table: Convert the source list into a table to be processed, including: source name, URL, status (to be fetched/success/failure), information obtained
  3. Search item by item: In order of priority, perform search and retrieval for each source
  4. Mark status: Each source must be marked with one of the following statuses:
    • ✅ Successfully obtained - detailed pricing information obtained
    • ⚠️ To be confirmed - partial information obtained but not complete, or further confirmation is needed
    • ❌ Failed - unable to obtain, page requires JavaScript or 404, etc.
  5. Checklist: After completion, it must be confirmed that each source has a corresponding processing record

Skipping any source is prohibited! Even if a source is expected to fail, retrieval must still be attempted and the status marked.

Step 1: Read and create a source checklist

As the first step of executing the task, sources.md must be read and a checklist in the following format must be created:

| 序号 | 信源类别 | 信源名称 | URL | 状态 | 获取信息摘要 |
|------|----------|----------|-----|------|--------------|
| 1 | 国内云厂商Coding Plan | 腾讯云 | https://cloud.tencent.com/act/pro/codingplan | ✅ | Pro 200元/月，首月39.9元 |
| 2 | 国内云厂商Coding Plan | 阿里云 | https://www.aliyun.com/benefit/scene/codingplan | ✅ | Pro 200元/月 |
...

Step 2: Online search - must be executed strictly in order!

Execution flow: Each source must be executed strictly in the order A→B→C→D→E!

  • Foreach source URLtry strategies A, B, C, D, E in sequence
  • Only when the current strategyclearly failscan you proceed to the next strategy
  • No strategy may be skipped!
  • Immediately update the status after completing all strategies for each URL

Strategy execution flowchart

URL1: 策略A → (失败?) → 策略B → (失败?) → 策略C → (失败?) → 策略D → (失败?) → 策略E
URL2: 策略A → (失败?) → 策略B → (失败?) → 策略C → (失败?) → 策略D → (失败?) → 策略E
...

Strategy failure criteria:

  • ❌ Returns "JavaScript required" or "Please enable JavaScript"
  • ❌ Returns 404 Not Found
  • ❌ Returned content is empty or incomplete (missing key information such as price, usage, etc.)
  • ⚠️ Baidu search is blocked by a verification mechanism

Strategy A - Directly access the main activity page

First access themain activity page URLin sources.md to obtain official pricing information.

Failure conditions (must proceed to Strategy B):

  • The page displays "Please enable JavaScript"
  • The page returns 404
  • Page content is incomplete, missing price/usage information

Strategy B - Visit help/documentation pages

Failure conditions (must proceed to Strategy C):

  • Page shows "Please enable JavaScript"
  • Page returns 404
  • Page content is incomplete

Help documentation URL references (from sources.md):

  • Tencent Cloud:https://cloud.tencent.com/document/product/1772/128947
  • Alibaba Cloud:https://help.aliyun.com/zh/model-studio/coding-plan
  • China Mobile Cloud:https://ecloud.10086.cn/op-help-center/doc/article/98320?marketLeoId=codingplan

Strategy C - Headless browser (Playwright) Must be executed!

Trigger conditions (must be used):

  • After both Strategy A/B fail
  • Page shows "Please enable JavaScript" or "We're sorry but...doesn't work properly without JavaScript"

Execution steps:

  1. Check whether Playwright is installed
# Windows环境检查
dir scripts/node_modules/playwright 2>nul || echo "NOT_INSTALLED"
  1. Install Playwright (if not installed) Enter the model-resources-collection directory to install:
# 安装依赖
npm install playwright

# 安装Chromium浏览器
npx playwright install chromium
  1. Modify the URL list in fetch-js-pages.js
# 编辑脚本文件，添加需要JS渲染的URL
notepad scripts/fetch-js-pages.js

Or directly modify the following array:

const urls = [
  { name: '移动云 Coding Plan', url: 'https://ecloud.10086.cn/portal/act/codingplan' },
  { name: '火山引擎 Coding Plan', url: 'https://www.volcengine.com/activity/codingplan' },
  { name: '京东云 Coding Plan', url: 'https://www.jdcloud.com/cn/pages/codingplan' },
  // 添加其他需要JS渲染的URL...
];
  1. Run the script
# 运行脚本获取页面
node scripts/fetch-js-pages.js
  1. Read the results and extract information
# 读取生成的JSON结果文件
type scripts/fetch-results.json
  • Extract price, package, usage, and other information from content
  • Update the status of the corresponding source from ❌ to ✅, and fill in the obtained information

Note: You must try to execute Playwright. If it cannot be executed due to environment limitations (such as the browser not being installed), mark it as ❌ and state "Strategy C execution failed (environment limitation)"

Strategy D - Baidu search

Use the websearch tool to search for keyword combinations:

  • 厂商名 + Coding Plan + 套餐 + 价格
  • 厂商名 + API + 免费额度 + 2026
  • 厂商名 + 模型 + 定价 + 用量限制

Failure conditions (must proceed to Strategy E):

  • No price/usage information was found in the search results
  • The search was blocked by a verification mechanism

Strategy E - Third-party information sources

When all the above strategies fail, search:

  • Technical blogs (InfoQ, Juejin, Zhihu)
  • Developer communities (V2EX, Reddit)
  • Review articles

1.3 Priority ranking

Priority 1 - Domestic cloud providers(Alibaba Cloud, Tencent Cloud, Baidu Cloud, JD Cloud, China Mobile Cloud, China Unicom Cloud, etc.)

Priority 2 - Domestic large model companies(DeepSeek, Zhipu, Moonshot AI, MiniMax, iFlytek, etc.)

Priority 3 - Free API services(Cloudflare Workers AI, Hugging Face, GitHub Models, etc.)

Priority 4 - Coding plans(GitHub Copilot, JetBrains, etc.)

Priority 5 - Developer communities(Search for the latest discussions and comparative reviews)

1.4 Information that must be obtained (important!)

Each vendor/packagemust include the following detailed information, and cannot simply say "refer to the official website":

  • Package name: The specific package name
  • Price: Specific price (e.g., 40 yuan/month, 200 yuan/month)
  • Supported models: List all supported model names
  • Usage limits: Specific request count limits (e.g., 1200 times/5 hours, 18000 times/month)
  • Supported tools: List of compatible AI coding tools
  • Purchase link: Link directly to the purchase page

Step 2: Organize information and call the script to generate the document

2.1 Extract key information (must be detailed!)

From the search resultsThe following detailed information must be extracted：

  • Package name: The specific package name (e.g., Coding Plan Lite, Coding Plan Pro)
  • Price: The specific price figure (e.g., 40 yuan/month, 200 yuan/month); first-month special offers must be noted
  • Supported models: List all supported models (e.g., GLM-5, Kimi-K2.5, MiniMax-M2.5, DeepSeek-V3.2)
  • Usage limits: The specific number of requests (e.g., 1200 requests/5 hours, 9000 requests/week, 18000 requests/month)
  • Quota recovery mechanism: How the quota recovers every 5 hours/weekly/monthly
  • Supported tools: Compatible AI programming tools (e.g., Claude Code, Cursor, Cline, etc.)
  • Package description: A brief description of the applicable scenarios
  • Notes: Special reminders (e.g., API calls prohibited, data authorization, etc.)
  • Purchase link: A link directly to the purchase page

Absolutely prohibitedSimply writing perfunctory descriptions such as "see the official website" or "refer to the official website"! If it cannot be obtained in one attempt, multi-level retrieval must be performed!

Tips for handling Baidu search results:

  • Baidu search results pages contain the latest news and price information
  • Prioritize results marked "official" or with the latest dates
  • Extract key price information from search snippets
  • If there are comparison review articles in the search results, refer to them first

2.2 Price comparison logic and sorting rules (important!)

Sorting rules for best value choices: arrange from low to high by average monthly price

  1. Prioritize displaying the package with the lowest average monthly price
  2. First-month special offer prices must be noted (e.g., 7.9 yuan for the first month, then 23.4 yuan/month)
  3. When prices are the same, sort by usage limits (higher quota first)
  4. All packages must include detailed information; "see the official website" must not appear

Example sorting:

  • Mobile Cloud: 23.4 yuan/month (7.9 yuan for the first month) → ranked first
  • Tencent Cloud Lite: 40 yuan/month → ranked second
  • Baidu Cloud Lite: 40 yuan/month → ranked third
  • Alibaba Cloud Pro: 200 yuan/month → ranked fourth
  • Tencent Cloud Pro: 200 yuan/month → ranked fifth

2.3 Prepare JSON data

Organize the extracted information into JSON format, with the following structure:

{
    "date": "2026-03-28",
    "cost_effective": [
        {
            "name": "厂商名称",
            "price": "价格信息（含首月特惠说明）",
            "models": "支持的模型1、模型2、模型3",
            "description": "套餐说明",
            "limit": "每5小时xx次请求，每周xx次请求，每月xx次请求",
            "recovery": "额度恢复机制说明",
            "tools": "支持的工具列表",
            "notes": "注意事项",
            "rating": "推荐指数",
            "link": "https://xxx"
        }
    ],
    "discount_plans": [...],
    "free_models": [...],
    "free_plans": [...]
}

Note: JSON data must be sorted from low to high by price!

2.3 Call the Python script to generate the document

The generate_report.py script in the scripts directory must be usedConvert the JSON data into a Markdown document.

Call command:

python scripts/generate_report.py --input data.json --template references/output-template.md --output model-resources-guide.md

Parameter description:

  • --input: The path to the input JSON data file
  • --output: The path to the output Markdown file
  • --template: Optional; specify the template file path (defaults to references/output-template.md)

Step 3: Final verification and saving

3.1 Final verification (must be performed!)

Before generating the document, the following verification must be completed:

  1. Open the checklist: Review the source checklist established in step 1
  2. Check the status of each item: Confirm that each source has a corresponding processing record
  3. Mark items pending confirmation: For sources that cannot be obtained, clearly mark "⚠️Pending confirmation" in the document and explain the reason
  4. Add missing items: If any sources are missing, obtain them immediately

Checklist example:

✅ 腾讯云 - 已获取详细价格
✅ 阿里云 - 已获取详细价格
✅ 百度云 - 已获取详细价格
✅ 火山引擎 - 已获取
⚠️ 京东云 - 需要JavaScript，待确认
⚠️ 移动云 - 需要JavaScript，待确认
⚠️ 智谱AI - 需要JavaScript，待确认
❌ 联通云 - 无帮助文档URL
...

3.2 Save the file

  • Save to the current directory:model-resources-guide.md
  • Mark the information retrieval date at the top of the document

Information sources

See references/sources.md for details(must refer to)

Output format

See references/output-template.md for details(information must be organized according to this template format)

Notes

  1. The information retrieval date must be marked in the document
  2. The user must be reminded that "specific discounts are subject to the latest official announcements of each platform"
  3. The configuration examples are for reference only; for the actual configuration, please refer to the official documentation
  4. Prioritize searching Chinese sources to ensure the information is of reference value to domestic users
  5. All sources in sources.md must be covered, and no item may be omitted
  6. Skipping any source is prohibited - Even if failure is expected, you must still try and mark the status
  7. Execute webfetch in parallel - Initiate multiple search requests simultaneously to improve efficiency
  8. Multi-level page crawling - When the main page has no information, you must look for subpages such as "User Guide" and "Learn More"
  9. Use Baidu Search as a supplement - When direct access fails, use Baidu Search to obtain information
  10. Perfunctory work is strictly prohibited - You cannot write "refer to the official website", "see the official website", etc.; you must provide specific information such as prices and usage limits
  11. User goal - The user needs to use this document to understand in detail the plans of each vendor, directly select the most cost-effective option, and click to purchase
  12. Final check - After completion, you must confirm that each source has a corresponding processing record, marked ✅/⚠️/❌
  13. Headless browser fallback - When all strategies fail, you must use Playwright to try to obtain JavaScript-rendered pages