Company Research Agent

Powered By

Llama 3 for advanced text generation • Tavily for web intelligence

Overview

This Python tool leverages the power of large language models LLaMA-3-8B and Tavily web search API to generate comprehensive company research reports. The agent automatically gathers information from multiple sources, synthesizes the data, and produces a professional Markdown report covering key business aspects of any company.

Key Features

• Comprehensive Research: Covers 7 critical business dimensions
• Source Verification: Includes all sources with confidence scoring
• Efficient Processing: Parallel search execution
• Professional Output: Ready-to-use Markdown report
• Optimized Performance: 4-bit quantization for GPU efficiency

Prerequisites

• Python 3.8+
• Tavily API key (free at app.tavily.com)
• NVIDIA GPU with CUDA support (recommended)
• Minimum 8 GB GPU RAM, 12 GB for optimal performance

Installation

1. Clone the repository:

git clone https://github.com/darkangrycoder/company-research-agent.git
cd company-research-agent

2. Install required dependencies:

pip install -r requirements.txt

Configuration

1. Get your Tavily API key from app.tavily.com
2. Open comp_research_agent.py in a text editor
3. Locate the API key section and insert your key:

# Get Tavily API key securely
TAVILY_API_KEY = "your_api_key_here"  # ← Replace with your actual key
tavily = TavilyClient(api_key=TAVILY_API_KEY)

Usage

Run the research agent with the company name as an argument:

python comp_research_agent.py --company "Company Name"

Example:

python comp_research_agent.py --company "Tesla"

Execution Notes

1. GPU Requirement: For optimal performance, run on GPU-enabled environment
   • Google Colab: Use "Runtime > Change runtime type > GPU"
   • Local: Ensure CUDA drivers are installed
2. Processing Time: Typically 3-6 minutes per report
3. Output:
   • Report printed in console
   • Markdown file saved as {Company_Name}_report.md

Report Sections

Each comprehensive report includes:

1. Overview: Founding, leadership, milestones
2. Financials: Revenue, funding, reports
3. Operations: Employees, locations
4. Market: Competitors, market share
5. Digital: Social media presence
6. Recent News: Latest developments
7. Employer: Company culture, ratings
8. All Sources: Complete reference list

Technical Architecture

graph TD
    A[User Input] --> B(Tavily Web Search)
    B --> C[Parallel Data Collection]
    C --> D[LLM Processing]
    D --> E[Report Generation]
    E --> F[Markdown Output]

Loading

Optimization Features

• 4-bit Quantization: Reduces model memory requirements by 4x
• Async I/O: Parallel web searches for faster data collection
• Content Truncation: Focuses on most relevant information
• HTML Sanitization: Clean text processing for better LLM comprehension

Troubleshooting

Common Issues:

• Slow Performance: Ensure you're using GPU acceleration
• API Errors: Verify Tavily API key is valid and has sufficient credits
• Incomplete Reports: Check network connection and API limits

Error Messages:

• TavilyAPIError: Invalid or expired API key
• CUDA Out of Memory: Reduce model size or use larger GPU
• ConnectionError: Check internet connection

Limitations

• Requires stable internet connection
• Tavily API has daily request limits
• Accuracy depends on web source reliability
• Complex company structures may require manual verification

Future Enhancements

• Multi-company comparison reports
• PDF export option
• Historical trend analysis
• Automated source credibility scoring
• Custom report templates

---

Get your Tavily API key: app.tavily.com
Llama 3.8 Web Modelhttps://huggingface.co/McGill-NLP/Llama-3-8B-Web Report Issues: GitHub Issues
Contribute: Fork and submit pull requests