Content Digital Employee Platform v5.0 Design Proposal

Product Positioning

A universal digital content management platform for individuals/enterprises — integrating 3 years of self-media experience with an AI-driven content production system

Core Architecture (Reference AIHOT)

``

┌─────────────────────────────────────────────────────────────┐

│ Content Digital Employee Platform │

├─────────────────────────────────────────────────────────────┤

│ │

│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │

│ │ Source Management │───▶│ AI Processing Center │───▶│ Content Output │ │

│ └─────────────┘ └─────────────┘ └─────────────┘ │

│ │ │ │ │

│ ▼ ▼ ▼ │

│ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐ │

│ │ T1 Official Sources │ │ Pre-screening AI Relevance │ │ Curated Recommendations │ │

│ │ T1.5 Official Social Media │ │ Multi-dimensional Scoring │ │ Daily Report Generation │ │

│ │ T2 KOL/Media │ │ Event Clustering │ │ Multi-platform Publishing │ │

│ │ T3 Industry Vertical │ │ Translation + Summary │ │ CMS Integration │ │

│ └─────────────┘ └─────────────┘ └─────────────┘ │

│ │

└─────────────────────────────────────────────────────────────┘

`

I. Source Management System

1.1 Source Tiering (Inspired by AIHOT)

1.2 Integrated Sources

• ✅ Weibo Hot Search (T2)
• ✅ Zhihu Hot List (T2)
• ✅ Reddit Trending (T2-T3)
• ✅ Unitree Embodied AI Community (T3)
• ? OpenAI Blog (T1)
• ? Anthropic Blog (T1)
• ? Hugging Face (T1.5)
• ? 36Kr/Machine Heart (T2)
• ? Product Hunt (T2)

1.3 Source Configuration File

`json

{

"sources": [

{

"id": "openai-blog",

"name": "OpenAI Blog",

"tier": 1,

"type": "rss",

"url": "https://openai.com/blog/rss.xml",

"category": ["AI", "LLM"],

"enabled": true

},

{

"id": "weibo-hot",

"name": "Weibo Hot Search",

"tier": 2,

"type": "api",

"url": "https://weibo.com/ajax/side/hotSearch",

"category": ["General", "Social"],

"enabled": true

}

]

}

`

II. AI Processing Center

2.1 Processing Flow (Referencing AIHOT Version 11 Iteration Experience)

`

Raw Information ──▶ Pre-screening (AI Relevance) ──▶ Multi-dimensional Scoring ──▶ Weight Calculation ──▶ Selection Judgment

│ │ │ │ │

│ ▼ ▼ ▼ ▼

│ DeepSeek V3 DeepSeek V4 Code Formula Threshold Filtering

│ (Low-Cost Pre-Screening) (Five-Dimensional Scoring) (Controllable & Adjustable) (Classification Threshold)

│

└──▶ Translation + Summary ──▶ Event Clustering ──▶ Storage

`

2.2 Multidimensional Scoring System

Five-Dimensional Scoring (Handled by large model, 200-line Prompt):

| Dimension | Description | Weight |

|------|------|------|

| Novelty | Whether it's newly released/first reported | 20% |

| Importance | Impact scope, industry influence | 25% |

| Relevance | Relevance to target domain | 20% |

| Readability | Content quality, information density | 15% |

| Virality | Topic potential, discussion potential | 20% |

Final Weight Calculation (Handled by code, quantifiable and adjustable):

`javascript

finalScore = (

baseScore tierWeight // Source weight

noveltyScore * 0.20 + // Novelty

importanceScore * 0.25 + // Importance

relevanceScore * 0.20 + // Relevance

readabilityScore * 0.15 + // Readability

viralScore * 0.20 // Virality

) * categoryMultiplier // Category coefficient

`

2.3 Event Clustering System

Problem: Same event reported by multiple sources → duplicates on featured page

Solution:

1. Embedding-based semantic similarity calculation
2. Time-window aggregation (within 24 hours)
3. Principle of authoritative source priority: Official website > Official social media > KOL > Media

`javascript

// Event cluster structure

{

"clusterId": "evt_20260507_gpt55",

"mainItem": "OpenAI Blog Post", // Most authoritative source

"relatedItems": [

"36Kr Report",

"Machine Heart Interpretation",

"Sam Altman's Tweet"

],

"category": "Model Release",

"score": 89

}

`

2.4 Pre-screening Strategy (Cost Optimization)

`javascript

// Phase 1: Rule-based Pre-screening (Free)

if (isSpam(item) || isDuplicate(item) || isIrrelevantCategory(item)) {

return; // Discard directly

}

// Phase 2: AI Pre-screening (Low-cost Model)

const isAIRelated = await preScreenWithAI(item); // DeepSeek V3

if (!isAIRelated) {

store(item); // Store in database but do not score

return;

}

// Phase 3: Deep Scoring (High-intelligence Model)

const scores = await scoreWithAI(item); // DeepSeek V4 Pro

`

III. Content Output System

3.1 Featured Recommendations

• Filter by category thresholds (different thresholds for different categories)
• Real-time updated timeline
• Supports user-defined interest tags

3.2 AI Daily Digest

• Automatically generated daily at 8:00
• Five sections: Model Releases, Product Updates, Industry News, Paper Research, Opinions & Tips
• Processed data aggregated directly, no secondary AI call required

3.3 Multi-platform Publishing

IV. Data Storage

4.1 Core Table Structure

`sql

-- Source Table

CREATE TABLE sources (

id VARCHAR(32) PRIMARY KEY,

name VARCHAR(100),

tier TINYINT,

type ENUM('rss', 'api', 'html', 'twitter'),

url VARCHAR(500),

category JSON,

enabled BOOLEAN DEFAULT TRUE,

last_fetch DATETIME

);

-- Content Table

CREATE TABLE contents (

id VARCHAR(32) PRIMARY KEY,

source_id VARCHAR(32),

title VARCHAR(500),

summary TEXT,

original_url VARCHAR(500),

publish_time DATETIME,

fetch_time DATETIME,

-- AI Processing Results

is_ai_related BOOLEAN,

scores JSON, -- Five-dimensional Scores

final_score DECIMAL(5,2),

is_featured BOOLEAN, -- Whether Featured

cluster_id VARCHAR(32),

-- Category tags

category VARCHAR(50),

tags JSON,

INDEX idx_score (final_score DESC),

INDEX idx_featured (is_featured, publish_time DESC)

);

-- Event clusters table

CREATE TABLE event_clusters (

cluster_id VARCHAR(32) PRIMARY KEY,

main_content_id VARCHAR(32),

category VARCHAR(50),

score DECIMAL(5,2),

created_at DATETIME

);

`

5. API Interfaces

5.1 Core APIs

`

GET /api/featured # Featured content list

GET /api/daily # Today's daily report

GET /api/trending # Trending topics

POST /api/sources # Add a source

GET /api/sources # Source list

POST /api/score # Manual scoring

GET /api/search?q=xxx # Content search

`

5.2 Webhooks

`

POST /webhook/new-featured # New featured notification

POST /webhook/daily-ready # Daily report generation completed

``

6. Technology Stack

VII. Development Plan

Phase 1: Core Framework (1 week)

• [ ] Data Source Configuration System
• [ ] Collection Scheduler
• [ ] AI Scoring Module
• [ ] Database Table Structure

Phase 2: AI Processing (1 week)

• [ ] Pre-screening Prompt Design
• [ ] Five-dimensional Scoring Prompt
• [ ] Event Clustering Algorithm
• [ ] Daily Report Generator

Phase 3: Output Layer (1 week)

• [ ] Curated API
• [ ] Daily Report API
• [ ] Web Management Interface
• [ ] Webhook Notification

Phase 4: Integration & Optimization (1 week)

• [ ] CMS Integration
• [ ] PLM Data Integration
• [ ] Performance Optimization
• [ ] Monitoring & Alerting

VIII. Cost Estimation

AI Call Cost (Daily 500 items):

• Pre-screening: 500 × 0.5K tokens × ¥0.001/K = ¥0.25
• Scoring: 250 × 2K tokens × ¥0.01/K = ¥5.00
• Translation & Summarization: 250 × 1K tokens × ¥0.01/K = ¥2.50
• Daily Cost: ¥8 / day ≈ ¥240 / month

Optimization Strategies:

1. Local Model for Pre-screening (Free)
2. Batch Processing to Reduce Call Count
3. Cache Scores for Similar Content

Core Philosophy:

Use code for control, let AI do what it excels at (understanding semantics), and calculate everything else with explicit formulas.

Controllable, Adjustable, Quantifiable, Back-testable.