Digikala-Smart-Search

A powerful web-based application designed for Persian-speaking users to search for products on Digikala, Iran’s leading e-commerce platform. By entering a product name in Persian, users can leverage AI-driven category classification and web automation to perform seamless searches, with results displayed directly in a browser.

Features

• Persian-Friendly Interface: Built with Streamlit, featuring right-to-left (RTL) text support and Persian font styling for a native user experience.
• AI-Powered Category Classification: Uses LangChain with Ollama to intelligently determine the appropriate product category based on the user’s query.
• Automated Web Search: Employs Selenium to navigate Digikala’s website, simulating human interactions to perform searches efficiently.
• Structured Workflow: Managed by LangGraph, ensuring a smooth sequence of tasks from query input to search execution.
• Real-Time Feedback: Provides visual feedback and error handling within the Streamlit interface for a reliable user experience.

Technologies Used

Technology	Description
Python 3.8+	Core programming language for the application.
Streamlit	Creates the interactive web-based user interface with RTL support.
LangChain	Framework for building applications with language models.
LangChain-Ollama	Integrates Ollama for natural language processing tasks.
LangGraph	Manages the workflow of query processing and search execution.
Selenium	Automates web interactions with Digikala’s website.

Prerequisites

• Python 3.8 or later: Download from python.org.
• Ollama: A local AI server for running the qwen2.5:latest model. Install from ollama.com.
• Google Chrome: Required for Selenium WebDriver automation. Download from google.com/chrome.
• ChromeDriver: Must match your Chrome version, available at chromedriver.chromium.org.
• Internet Access: Required for accessing Digikala and the font URL for styling (Vazir font).

Installation

1. Install Python:

   • Ensure Python 3.8 or later is installed. Verify with:

     python --version

2. Install Ollama:

   • Follow instructions at ollama.com.
   • Pull the required model:

     ollama pull qwen2.5:latest

   • Start the Ollama server:

     ollama serve

3. Install Google Chrome and ChromeDriver:

   • Install Google Chrome from google.com/chrome.
   • Download ChromeDriver matching your Chrome version from chromedriver.chromium.org.
   • Ensure ChromeDriver is in your system’s PATH or specify its path in the code if needed.

4. Clone the Repository:

   git clone https://github.com/armanjscript/Digikala-Smart-Search.git
   cd Digikala-Smart-Search

5. Install Python Libraries:

   pip install streamlit langchain-ollama langgraph selenium langchain

6. Verify Network Access:

   • Ensure access to Digikala and the font URL (Vazir font) for proper styling.

Usage

1. Run the Application:

   • Navigate to the project directory and start the Streamlit app:

     streamlit run main.py

   • Open your browser and go to http://localhost:8501 to access the application.

2. Using the Application:

   • In the Streamlit interface, you’ll see a title in Persian (e.g., "جستجوی هوشمند دیجی‌کالا").
   • Enter a product name in Persian (e.g., "گوشی موبایل") in the text input field.
   • Click the "جستجو" (Search) button.
   • The system will:
     • Analyze the query to determine the product category using the qwen2.5:latest model.
     • Navigate to Digikala using Selenium, select the appropriate category, and perform the search.
     • Open a browser window displaying the search results.
   • A success or error message will appear in the Streamlit app.

How It Works

The Digikala-Smart-Search integrates AI and web automation for a seamless experience:

1. User Input: Users enter a product name in Persian via the Streamlit interface.
2. Category Classification: The qwen2.5:latest model analyzes the query to determine the appropriate product category (e.g., "موبایل" for mobile phones).
3. Web Automation: Selenium WebDriver navigates to Digikala, selects the category, enters the search term, and submits the search, simulating human interactions.
4. Feedback: Streamlit displays a success message if the search completes or an error message if issues arise (e.g., network errors or unavailable categories).

The workflow is orchestrated by LangGraph, ensuring efficient task sequencing.

Limitations

• External Dependencies: Relies on Digikala and Ollama models, which may change or become unavailable.
• Web Automation Stability: Dynamic elements on Digikala may cause occasional failures (e.g., StaleElementReferenceException).
• Network Dependency: Requires a stable internet connection for web access and font loading.
• Category Accuracy: The AI model’s category classification depends on the quality of the input query and model performance.
• Browser Dependency: Limited to Google Chrome and ChromeDriver; other browsers are not supported.

Known Issues

• Selenium Errors: Dynamic web elements may cause automation failures.
• Timeouts: Fixed timeouts (e.g., 10 seconds) may fail on slow networks.
• Font Loading: The external font URL may be inaccessible, affecting styling.
• Error Handling: Limited handling for edge cases like network interruptions or LLM failures.

Future Improvements

• Enhance error handling and logging for better debugging.
• Support additional e-commerce platforms to reduce dependency on Digikala.
• Optimize Selenium for faster performance, possibly using headless mode.
• Improve UI/UX with progress indicators and input validation.
• Add unit tests for reliability.
• Expand language support and accessibility features.

Contributing

Contributions are welcome! Fork the repository, make changes, and submit a pull request. Ensure your code follows the project’s style and includes tests.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Contact

For questions or feedback, please contact [Arman Daneshdoost] at [[email protected]].