fraudcrawler 0.7.4

Intelligent Market Monitoring

fraudcrawler

Fraudcrawler is an intelligent market monitoring tool that searches the web for products, extracts product details, and classifies them using LLMs. It combines search APIs, web scraping, and AI to automate product discovery and relevance assessment.

Features

• Asynchronous pipeline - Products move through search, extraction, and classification stages independently
• Multiple search engines - Google Search, Google Shopping, and more...
• Search term enrichment - Automatically find related terms and expand your search
• Product extraction - Get structured product data via Zyte API
• LLM classification - Assess product relevance using OpenAI API with custom prompts
• Marketplace filtering - Focus searches on specific domains
• Deduplication - Avoid reprocessing previously collected URLs
• CSV export - Results saved with timestamps for easy tracking

Prerequisites

• Python 3.11 or higher
• API keys for:
  • SerpAPI - Google search results
  • Zyte API - Product data extraction
  • OpenAI API - Product classification
  • DataForSEO (optional) - Search term enrichment

Installation

python3.11 -m venv .venv
source .venv/bin/activate
pip install fraudcrawler

Using Poetry:

poetry install

Configuration

Create a .env file with your API credentials (see .env.example for template):

SERPAPI_KEY=your_serpapi_key
ZYTEAPI_KEY=your_zyte_key
OPENAIAPI_KEY=your_openai_key
DATAFORSEO_USER=your_user  # optional
DATAFORSEO_PWD=your_pwd    # optional

Usage

Basic Configuration

For a complete working example, see fraudcrawler/launch_demo_pipeline.py. After setting up the necessary parameters you can launch and analyse the results with:

# Run pipeline
await client.run(
    search_term=search_term,
    search_engines=search_engines,
    language=language,
    location=location,
    deepness=deepness,
    excluded_urls=excluded_urls,
)

# Load results
df = client.load_results()
print(df.head())

Advanced Configuration

Search term enrichment - Find and search related terms:

from fraudcrawler import Enrichment

deepness.enrichment = Enrichment(
    additional_terms=5,
    additional_urls_per_term=10
)

Marketplace filtering - Focus on specific domains:

from fraudcrawler import Host

marketplaces = [
    Host(name="International", domains="zavamed.com,apomeds.com"),
    Host(name="National", domains="netdoktor.ch,nobelpharma.ch"),
]

await client.run(..., marketplaces=marketplaces)

Exclude domains - Exclude specific domains from your results:

excluded_urls = [
    Host(name="Compendium", domains="compendium.ch"),
]

await client.run(..., excluded_urls=excluded_urls)

Skip previously collected URLs:

previously_collected_urls = [
    "https://example.com/product1",
    "https://example.com/product2",
]

await client.run(..., previously_collected_urls=previously_collected_urls)

View all results from a client instance:

client.print_available_results()

Output

Results are saved as CSV files in data/results/ with the naming pattern:

<search_term>_<language_code>_<location_code>_<timestamp>.csv

Example: sildenafil_de_ch_20250115143022.csv

The CSV includes product details, URLs, and classification scores from your workflows.

Development

For detailed contribution guidelines, see CONTRIBUTING.md.

License

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

Architecture

Fraudcrawler uses an asynchronous pipeline where products can be at different processing stages simultaneously. Product A might be in classification while Product B is still being scraped. This is enabled by async workers for each stage (Search, Context Extraction, Processing) using httpx.AsyncClient.

For more details on the async design, see the httpx documentation.