Finn.no Computer Deals Analyzer

Overview

The Finn.no Computer Deals Analyzer is a Python-based automated system designed to scour Finn.no for computer listings, evaluate their specifications, and identify the best deals. By leveraging GPT-3.5 for data parsing and decision-making, this tool ensures that only the most relevant and cost-effective options make it to your final review.

Key Features:

• Automated Finn.no Listings Scraper: Pulls raw listings of computers from Finn.no.
• AI-Powered Data Parsing: Utilizes GPT-3.5 to transform raw listing data into a standardized JSON format of computer specifications.
• Smart Filtering: Removes irrelevant or low-quality listings based on your criteria.
• Tournament-Style Evaluation: Puts listings against each other in brackets, where GPT-3.5 determines the better deal in each round, ultimately selecting five finalists.
• Final Review: Presents the top 5 options for manual evaluation to ensure satisfaction.

---

Installation

1. Clone the Repository:

   git clone https://github.com/filip-rs/finn-scraper.git
   cd finn-scraper

2. Set Up the Environment:

   • Ensure Python 3.9+ is installed.
   • Create and activate a virtual environment:

     python3 -m venv env
     source env/bin/activate # On Windows: env\Scripts\activate

3. Install Dependencies:

   pip install -r requirements.txt

4. Configure API Keys:

   • Obtain an OpenAI API key
   • Create a .env file in the root directory and add:

     API_KEY=xx-xxxxxxxxxxxxxxxxxxxxxxxxx

---

Usage

1. Run the Script:

   python main.py

2. Process Flow:

   • Step 1: The script scrapes Finn.no for computer listings based on your search criteria.
   • Step 2: GPT-3.5 processes raw data into structured JSON specifications.
   • Step 3: Listings are filtered to remove irrelevant or low-value options.
   • Step 4: A tournament-style elimination determines the best deals.
   • Step 5: The final 5 candidates are displayed for manual review.

3. Output:

   • The script generates a detailed report of the finalists in both JSON and plain-text formats for easy access.

---

Configuration

To customize the behavior:

• Runtime: Certain options are available at runtime, such as search query.
• Additional customisation: Most of the programs logic is located in main.py and can be modified there.

---

Example Output

JSON Output

{
    "PRICE": $$$$,
    "CPU": "AMD Ryzen 7 2700X",
    "GPU": "MSI Radeon RX 5700 XT 8GB",
    "MEMORY": "Corsair Vengeance 3600 mhz 16 RGB",
    "STORAGE": "1 TB M.2 NVMe",
    "POWER_SUPPLY": "Sharkoon WPM 600 Bronze",
    "WIFI_CARD": "ROG CROSSHAIR VII HERO (WI-FI)",
    "url": "https://www.url.com"
}

Finalists

1. Dell XPS 15 2022 - $1200 - {url}
2. HP Spectre x360 - $1100 - {url}
3. Lenovo ThinkPad X1 Carbon - $1300 - {url}
4. Apple MacBook Pro 14" - $2000 - {url}
5. ASUS ROG Zephyrus G14 - $1400 - {url}

---

Dependencies

• Python Packages:
  • requests==2.32.3: For interacting with APIs.
  • beautifulsoup4==4.12.3: Scraping Finn.no listings.
  • openai==1.56.1: OpenAi API integration.
  • python-dotenv==1.0.1: Managing environment variables, more specifically the API key.

Install them with:

pip install -r requirements.txt

---

Contributing

Contributions are welcome! If you’d like to improve the system or add features, feel free to fork the repository and submit a pull request.

---

Limitations

• The script relies on GPT-3.5, which does cost money. It is however reasonably priced but beware.
• Performance may vary based on the quality of the Finn.no data and GPT-3.5's parsing capabilities. GPT can and will make mistakes.

---

License

---

Acknowledgments

• OpenAI GPT-3.5: For its text parsing and decision-making capabilities.
• Contributors: So far it's only been me, myself and I.