ReSynap

ReSynap: Knowledge Graph AI Agent for MindMap Generation, Data Personalization and Data Automation

Welcome to ReSynap—an advanced Knowledge-RAG agent designed to generate dynamic mind maps from text, transform text chats into visual knowledge graphs, build a personalization engine, and automate a range of tasks that integrate with customer data platforms (CDPs) and more.

Features

1. Dynamic Mind Map Generation: Easily convert text into comprehensive mind maps that visually organize information and relationships.

2. Text-to-Knowledge Graphs: Transform unstructured text chats into interactive knowledge graphs, uncovering insights and making connections intuitive.

3. Customer Profile Classification: Automatically classify customer profiles within CDPs, allowing for personalized segmentation and enhanced targeting.

4. Data Reports and Insights: Generate detailed data reports with ease, helping teams stay informed with up-to-date insights.

5. Customer Segmentation: Segment customer data effectively based on behavior, demographics, or any custom metric that fits your business needs.

6. Real-Time Product Recommendations: Get tailored product recommendations for specific user profiles, enabling targeted marketing and enhancing user experience.

7. Automated Tasks:

   1. Send personalized emails at scale.
   2. Trigger Zalo notifications to keep customers engaged.
   3. Automate various customer-related workflows with seamless integration.

Use Cases

• Marketing Automation: Automate segmentation and customer communications.
• Customer Data Analysis: Deepen your insights into customer behavior and preferences.
• Product Recommendations: Serve real-time, relevant recommendations to improve engagement and conversion.
• Knowledge Management: Use knowledge graphs for easy data retrieval and management.
• Workflow Automation: Set up triggers for automating tasks across multiple platforms.

Getting Started

1. Installation: [TODO]
2. Usage: [TODO]
3. API Integration: [TODO]

Contributions

Contributions to ReSynap are welcome! Please follow the [contribution guidelines](link to contribution guidelines) and submit your pull requests.

---

With ReSynap, transform data and text into actionable, visual insights and automate your workflows seamlessly!

System requirements

• Python 3.10+
• FastAPI
• Pydantic
• Qdrant Vector Database. Refer to the Qdrant documentation for instructions: https://qdrant.tech/documentation/quick-start/
• Qdrant client library (install with pip install qdrant-client)
• ArangoDB for graph database
• Redis or Apache Kafka for data processing on queue
• Large language models (LLMs) likes Google Gemini or OpenAI

Installation

1. Clone the repository:

   git clone https://github.com/trieu/resynap

2. Create env

    python -m venv env
    source env/bin/activate

3. Install the dependencies:

   pip install -r requirements.txt

4. Start the Qdrant server:

   ./start_qdrant.sh

5. Next, create a .env file by coping the file sample.env or create your own file with content:

    QDRANT_HOST=localhost
    QDRANT_PORT=6333

    QDRANT_CLOUD_HOST=
    QDRANT_CLOUD_API_KEY=

    API_HOST=0.0.0.0
    API_PORT=8000

    REDIS_HOST=
    REDIS_PORT=0
    DEFAULT_AUTHORIZATION_KEY=personalization_test

Running the API

uvicorn main:api_personalization --reload 

This will start the API server. You can access the API documentation at http://localhost:8000/docs

API Endpoints

All API Endpoints must be called with the header: Authorization = [your_api_key]
The [your_api_key] must be in Redis. E.g: 127.0.0.1:6480> set personalization_test true

Profiles

• POST /add-profile/
  • Adds a single user profile to the database.
  • Request body: ProfileRequest object (see below)
  • Response: {"status": "Profile added successfully"}
• POST /add-profiles/
  • Adds multiple user profiles to the database in bulk.
  • Request body: List of ProfileRequest objects
  • Response: {"status": "All profiles added successfully"}
• POST /check-profile-for-recommendation/
  • Add or update a profile, then get real-time recommendations for the profile
  • Request body: ProfileRequest object
  • Response: List of recommended products (see example below)

Products

The product management API

• POST /add-product/
  • Adds a single product to the database.
  • Request body: ProductRequest object (see below)
  • Response: {"status": "Product added successfully"}
• POST /add-products/
  • Adds multiple products to the database in bulk.
  • Request body: List of ProductRequest objects
  • Response: {"status": "All products added successfully"}

Contents

The content management API

• POST /add-content/
  • Adds a single content item to the database.
  • Request body: ContentRequest object (see below)
  • Response: {"status": "Content item added successfully"}
• POST /add-contents/
  • Adds multiple content items to the database in bulk.
  • Request body: List of ContentRequest objects
  • Response: {"status": "All content items added successfully"}

Recommendations

The recommendation API to personalize product, contents and both.

• GET /recommend/{profile_id}
  • Gets personalized product recommendations for a given profile ID.
  • Path parameters:
    • profile_id: The ID of the user profile.
    • item_type: The item type to get recommendation: product, content, or both.
  • Query parameters:
    • top_n (optional): The maximum number of recommendations to return (default: 8).
    • except_product_ids (optional): Comma-separated string of product IDs to exclude from
    • except_content_ids (optional): Comma-separated string of content IDs to exclude from recommendations (e.g., "item_1,item_3").
  • Response: List of recommended product / content IDs (see example below)

Data Models

ProfileRequest:

{
    "profile_id": "crm_11",
    "page_view_keywords": [
        "car",
        "bike",
        "accessories"
    ],
    "purchase_keywords": [
        "helmet",
        "gloves"
    ],
    "interest_keywords": [
        "travel",
        "photography",
        "outdoors"
    ],
    "additional_info": {
        "age": 28,
        "location": "Germany"
    }
}

ProductRequest:

{
    "product_id": "item_1",
    "product_name": "Macbook Pro",
    "product_category": "Electronics",
    "product_keywords": ["tech", "work", "portable"],
    "additional_info": {"brand": "Apple", "price": 1200}
}

ContentRequest:

{
    "content_id": "item_1",
    "title": "Macbook Pro",
    "content_category": "Electronics",
    "content_keywords": ["tech", "work", "portable"]
}

Example Usage

Adding a Profile

curl -X POST \
  http://localhost:8000/add-profile/ \
  -H 'Content-Type: application/json' \
  -H 'Authorization: personalization_test' \
  -d '{
    "profile_id": "crm_15",
    "page_view_keywords": [
      "car",
      "bike",
      "accessories"
    ],
    "purchase_keywords": [
      "helmet",
      "gloves"
    ],
    "interest_keywords": [
      "travel",
      "photography",
      "outdoors"
    ],
    "additional_info": {
      "age": 28,
      "location": "Germany"
    }
  }'

Getting Recommendations

curl -X GET \
  -H 'Authorization: personalization_test' \
  "http://localhost:8000/recommend/crm_15?top_n=5&except_product_ids=item_1,item_6" 

Future Improvements

• Implement more sophisticated vectorization techniques for user profiles and product data.
• Add support for more advanced filtering and ranking options for recommendations.
• Integrate with a user interface for managing profiles and products.