AI Droplet

你的每日 AI 情报站 · Your Daily AI Intelligence

A static aggregation site covering AI news, models, tools, MCP servers, and agent skills — updated daily and hosted on GitHub Pages.

Live site → https://hisbanliu.github.io/AI_droplet/

---

Contents

Section	Source	Update
Daily News	10 RSS feeds → Gemini scoring, Top 10	Daily
Models	HuggingFace API + curated commercial models	Daily
MCP Servers	punkpeye/awesome-mcp-servers README (1 294 servers)	Weekly
Agent Skills	Anthropic — curated	Manual
AI IDEs	Curated (8 IDEs)	Manual
AI Tools	Curated (32 tools, 8 categories)	Manual

Languages: EN / 中文 toggle, preference saved to localStorage.

---

Quick Start

# 1. Clone
git clone https://github.com/hisbanliu/AI_droplet.git
cd AI_droplet

# 2. Virtual environment
python -m venv .venv
.venv\Scripts\activate       # Windows
# source .venv/bin/activate  # macOS/Linux

# 3. Dependencies
pip install -r requirements.txt

# 4. API keys
cp .env.example .env
# Edit .env — add GEMINI_API_KEY and GITHUB_TOKEN

# 5. Run
python run_daily.py
# Opens http://localhost:8000 is NOT needed — just open docs/index.html in a browser,
# or push to GitHub and view the Pages URL.

run_daily.py accepts flags:

Flag	Effect
--push	git add, git commit, git push automatically
--skip-news	Skip news fetch (save API quota)
--skip-models	Skip HuggingFace fetch
--skip-mcp	Skip MCP README fetch

---

Architecture

Python scripts (scripts/)
    ├── fetch_news.py     →  data/news.json        (RSS + Gemini scoring)
    ├── fetch_models.py   →  data/models.json       (HuggingFace + curated)
    ├── fetch_mcp.py      →  data/mcp.json          (GitHub README parse)
    └── generate_site.py  →  docs/index.html        (Jinja2 render)
                             docs/data/*.json        (copied for fetch())

GitHub Pages (docs/)
    └── index.html ← 41 KB shell, loads data at runtime via fetch()

Templates are split for maintainability:

templates/
  index.html.j2          ← skeleton (head / nav / hero / footer / scripts)
  sections/
    news.html.j2          ← // 01 Daily News
    models.html.j2        ← // 02 Models
    mcp.html.j2           ← // 03 MCP Servers
    skills.html.j2        ← // 04 Agent Skills
    ides.html.j2          ← // 05 AI IDEs
    tools.html.j2         ← // 06 AI Tools

---

Adding Tools

Edit data/tools.json — append an entry:

{
  "name": "Tool Name",
  "category": "coding",
  "company": "Company",
  "description": "One-line description in English.",
  "url": "https://example.com",
  "pricing": "Free tier + $20/mo",
  "tags": ["tag1", "tag2"]
}

Existing categories: image-generation · video-generation · audio · ai-search · coding · productivity · ai-agents · writing

Adding a new category: add entries with a new category value, then add the i18n key tools.cat.<category> in both en and zh blocks inside templates/index.html.j2. The frontend derives tab buttons automatically from the data.

---

Environment Variables

# .env (local only — never committed)
GEMINI_API_KEY=your_key_here
GITHUB_TOKEN=your_pat_here    # optional — raises API rate limit 60 → 5000 req/h

---

Tech Stack

• Runtime: Python 3.12, feedparser, httpx, Jinja2, google-generativeai
• Frontend: HTML + Tailwind CSS CDN + Alpine.js CDN
• Fonts: Fira Code + Inter (Google Fonts)
• Hosting: GitHub Pages (docs/ directory, same branch)
• AI: Gemini 2.0 Flash (news scoring & summarization)

---

Roadmap

See PLAN.md for the full v1–v4 upgrade roadmap and v2 architecture design (manifest-driven section registry, i18n file extraction, GitHub Actions automation).