🎯 YouTube Comments Spam Filter — Backend Service

Overview

This is a high-performance backend service written in Rust designed to detect and classify Indonesian YouTube comment spam, particularly spam linked to illegal online gambling operations. It leverages AI-based inference engines (DeepSeek / Gemini) and an optional Redis-based caching layer for optimal speed and accuracy.

---

🔧 Key Features

• ✅ Real-time classification of YouTube comments as spam / not spam
• ✅ AI-powered analysis using DeepSeek and Gemini language models
• ✅ Redis-based caching layer for reduced latency and cost
• ✅ Designed with extensibility and horizontal scalability in mind

---

🧱 Technology Stack

Layer	Technology
Language	Rust (async runtime)
Cache Layer	Redis (or compatible key-value store)
AI Integrations	DeepSeek, Gemini (via HTTP APIs)
API Interface	JSON over HTTP (REST)
Observability	(TBD) Tracing, Metrics, Logs

---

🏗 Architecture

     +---------------------+
     |     Client API      |
     +---------+-----------+
               |
               v
     +---------------------+
     |     API Server      |  <--- Written in Rust
     +---------------------+
               |
      +--------+--------+
      |                 |
      v                 v
 +--------+       +------------+
 |  Redis |       |  AI Models |
 +--------+       +------------+
                      |
             DeepSeek / Gemini

• API Server: Exposes endpoints to classify comments.
• Cache Layer: Speeds up repeated classifications using Redis.
• AI Module: Connects to either DeepSeek or Gemini APIs to perform classification.

---

📡 API Endpoints (Planned)

Method	Endpoint	Description
POST	/analyze	Accepts a YouTube comment and returns spam classification (CSV format)
GET	/health	Health check
GET	/cache/:keyword	Inspect cache hit for a normalized gambling keyword

Note: Full OpenAPI spec will be available post-v1.

---

🧪 AI Integration Examples

🧠 DeepSeek

curl --request POST \
  --url https://api.deepseek.com/chat/completions \
  --header 'authorization: Bearer {DEEPSEEK_TOKEN}' \
  --header 'content-type: application/json' \
  --data '{
    "model": "deepseek-chat",
    "messages": [
      {
        "role": "system",
        "content": "You are an AI classifier detecting Indonesian YouTube comment spam related to illegal gambling..."
      },
      {
        "role": "user",
        "content": "Aku adalah pemenang, dan ❄️ KYT4D ❄️ adalah keberuntunganku!"
      }
    ],
    "stream": false,
    "max_tokens": 50
  }'

🤖 Gemini

curl --request POST \
  --url 'https://generativelanguage.googleapis.com/v1beta/models/gemini-2.0-flash-lite:generateContent?key={GEMINI_TOKEN}' \
  --header 'content-type: application/json' \
  --data '{
    "system_instruction": {
      "parts": [
        {
          "text": "You are an AI classifier detecting Indonesian YouTube comment spam related to illegal gambling..."
        }
      ]
    },
    "contents": [
      {
        "parts": [
          {
            "text": "Aku adalah pemenang, dan ❄️ KYT4D ❄️ adalah keberuntunganku!"
          }
        ]
      }
    ],
    "generationConfig": {
        "stopSequences": [
          "\n"
        ],
        "temperature": 0.2,
        "maxOutputTokens": 50,
        "topP": 0.5,
        "topK": 3
      }
  }'

📤 Response Format

The API response (from either DeepSeek or Gemini) will return a CSV-formatted string:

S,K,C

Columns:

• S: Spam classification (1 = SPAM, 0 = NOT SPAM)
• K: Normalized keyword (e.g., KYT4D, AXL777, GenericGambling, or N/A)
• C: Confidence score (0.00–1.00)

Sample Responses:

1,KYT4D,0.95
0,N/A,0.03
1,GenericGambling,0.80

---

🚀 Getting Started

1. Install Rust:

curl https://sh.rustup.rs -sSf | sh

2. Install Redis (or a compatible key-value store)

3. Clone the Repository:

git clone https://github.com/your-org/yt-comments-filter-be.git
cd yt-comments-filter-be

4. Build & Run the Service:

cargo build --release
./target/release/yt-comments-filter-be

🤝 Contributing

We welcome external contributions and collaborative improvements. To contribute:

1. Fork the repository to your own GitHub account.
2. Create a new branch for your feature or fix.
3. Make your changes, following Rust best practices and existing code structure.
4. Test your code thoroughly.
5. Submit a Pull Request (PR) with a clear explanation of your changes and rationale.

Note: All contributions must pass formatting (cargo fmt) and linting (cargo clippy) checks before being reviewed.

---

📌 Roadmap

• Admin dashboard (browser-based) for cache visibility and spam analytics
• WebSocket support for real-time comment feed ingestion
• Direct integration with YouTube Data API (v3) for automated moderation
• Expand language model support beyond Indonesian (e.g., Tagalog, Vietnamese)

---

🧠 Strategic Use Cases

This backend is not just a spam filter — it’s a foundation for multiple verticals:

• 🔍 Brand safety tools for marketing agencies targeting SEA/Indonesia
• 🛡 Anti-abuse layers for influencer management platforms
• 📊 Real-time spam analytics for YouTube creators and partners
• 💼 Licensable SaaS IP — ready for resale via platforms like Acquire.com

---

📜 License

Distributed under the MIT License. See the LICENSE file for full legal terms.

---

👨‍💻 Maintainer

Hendri Marcolia
LinkedIn: linkedin.com/in/hendri-marcolia-847ba0190

For enterprise collaborations, investment, or acquisition inquiries — connect via LinkedIn or direct business channels.

---