FreeAPIHub：自动维护的免费 API 目录，基于 GitHub Actions 每日发布

A self-maintaining directory of free and freemium APIs. Set it up once, let it run every day.

FreeAPIHub discovers useful public APIs from legitimate sources, verifies them against their official websites and documentation, researches them with an LLM (never inventing facts), and publishes 0–4 quality pages per day to a fast static website - fully automatically, on free infrastructure (GitHub Actions + GitHub Pages).

Every listing is checked before publication. Days with no quality candidates publish zero pages. Quality over quota.

What this project does

Once per day a GitHub Actions workflow runs this pipeline:

03:17 UTC
   ↓
Discover API candidates from public sources (curated lists, open directories, GitHub)
   ↓
Normalize + deduplicate (deterministic - never the LLM)
   ↓
Verify: is the official site alive? HTTPS? does it really describe an API?
   ↓
Select only candidates scoring above the quality threshold
   ↓
LLM research: summarize, categorize, extract only source-supported facts
   ↓
Generate API pages + update homepage, categories, search index, sitemap
   ↓
Validate everything (JSON, HTML, links, ads, config)
   ↓
Commit + deploy to GitHub Pages

A weekly Maintenance workflow re-checks existing listings, applies the status policy (verified → warning → temporarily_unavailable → dead), and never deletes a listing on a single failure. A manual Deploy Site workflow rebuilds after config changes.

The LLM is used only for summarizing verified candidates. Duplicate detection, URL checks, scoring, sitemaps, statistics and page rendering are all deterministic code - so running costs stay near zero (a handful of small LLM calls per day, at most).

Requirements

• A GitHub account (free plan is enough)
• An LLM API key (OpenAI-compatible: OpenAI, Groq, OpenRouter, DeepSeek, etc.)
• Advertisement code from your ad network (optional - the site works without it)
• Optional: a custom domain

No VPS. No database server. No paid anything.

One-time setup

1. Get the code

Fork or clone this repository to your own GitHub account, then clone locally if you prefer. Everything also works by just pushing the repository as-is.

2. Configure the site

Edit config/site.json (the only file you normally need to touch):

3. Add the LLM API key (GitHub Secrets)

Go to Settings → Secrets and variables → Actions → New repository secret (green button "New repository secret") and add:

LLM_API_KEY is never exposed to the browser, never written to HTML/JSON, and never logged. If you omit it the site still runs - new publications just use deterministic metadata only (stricter, so fewer pages pass).

4. Add your advertisement code

Open config/ads.json and paste your ad network's code into the slots you want to enable:

{
  "enabled": true,
  "slots": {
    "top": "<script ...></script>",
    "after_intro": "",
    "middle": "",
    "before_related": "",
    "sidebar": "",
    "footer": ""
  }
}

• Available slots on API pages: top, after_intro, middle, before_related, footer
• The homepage and category pages use top, after_intro and footer
• Your code is inserted verbatim - never modified, escaped or duplicated
• Leave a slot "" to disable it; delete nothing else

After editing, run Actions → Deploy Site → Run workflow to rebuild the pages.

5. Enable GitHub Pages

1. Repository Settings → Pages
2. Under "Build and deployment", set Source: GitHub Actions
3. Done - the workflows deploy automatically after every successful run

6. Enable GitHub Actions

If Actions are not already enabled: Settings → Actions → General → Allow all actions. The Daily Publisher runs at 03:17 UTC; you can also trigger it manually (see below).

7. (Optional) Custom domain

Follow the instructions inside CNAME.example in the repository root, then update site_url / site_url_path in config/site.json and re-run Deploy Site.

Running manually

You do not need to run anything by hand. But to test:

• Daily pipeline: Actions → Daily Publisher → Run workflow
• Maintenance: Actions → Maintenance → Run workflow
• Rebuild/deploy only: Actions → Deploy Site → Run workflow

Each run writes a human-readable summary (discovered / verified / published / failed) at the bottom of the workflow run page.

Running locally instead:

pip install -r requirements.txt
python scripts/seed.py        # one-time: verify + publish the curated starter APIs
python scripts/run_daily.py   # full daily pipeline
python scripts/maintenance.py # weekly maintenance
python scripts/validate.py --selftest  # failure-mode tests

Repository layout

config/       site.json (EDIT THESE VALUES) + ads.json (paste ad code)
data/         persistent JSON database (apis, queue, sources, history, run log)
scripts/      discovery, verification, LLM research, site build, sitemap,
              maintenance, validation - all plain Python + stdlib + requests
templates/    HTML templates and components
site/         the generated static website (committed, deploy-ready)
.github/      the three workflows + the API submission issue template

Submitting an API

Use the API submission issue template (Issues → New issue). Every submission - whether from the discovery pipeline or a human - goes through the same verification before publication. Nothing is auto-published without validation.

Troubleshooting

Honest expectations

This project optimizes for low cost, useful pages and long-term organic traffic. It does not promise revenue, does not fabricate popularity, ratings or testimonials, and does not manipulate ads. The site earns nothing automatically - it is the pages, over time, that create the conditions for advertising income.

License

MIT - see LICENSE. Third-party API names referenced by the directory belong to their respective owners.