StockBot

A tool for collecting and analyzing financial news about companies using Yahoo Finance API, with integrated support for the DeepSeek LLM and investment report generation.

Features

• Collects financial news articles from Yahoo Finance API via yfinance library
• Stores the information in a persistent vector database (ChromaDB)
• Enables semantic search for company-specific financial information
• Extracts structured data from news articles including:
  • Article titles
  • URLs
  • Publishers
  • Content/summaries
  • Publication dates
• Integrates with DeepSeek-r1 1.5B model via Ollama
• Generates Warren Buffett-style investment reports for any ticker

Installation

1. Clone this repository
2. Ensure you have Python 3.10+ installed
3. Install dependencies using uv:

uv sync

4. Install Ollama (if not already installed):

curl -fsSL https://ollama.com/install.sh | sh

Usage

Stock News Collection

Run the main stockbot script:

uv run python src/stockbot.py

When prompted:

1. Enter a company ticker symbol (e.g., AAPL, MSFT, TSLA)
2. The system will fetch recent news about the company from Yahoo Finance
3. The articles will be stored in a vector database
4. You can then search for specific financial information using natural language queries

DeepSeek Model Integration

Run the DeepSeek model installer:

uv run python src/deepseek_test.py

This script will:

1. Verify Ollama is running
2. Pull the DeepSeek-r1 1.5B model
3. Test the model with a sample prompt
4. Display information about the model

Investment Report Generation

Run the investment report generator:

uv run python src/investment_report.py AAPL

This script will:

1. Retrieve relevant news articles for the specified ticker from ChromaDB
2. Format the articles into a context document
3. Send a prompt to DeepSeek with the context and ticker information
4. Generate a detailed investment report with Warren Buffett-style analysis
5. Save the report to [TICKER]_investment_report.md

Investment Report Format

Reports follow a structured format defined in src/improved_prompt.md with:

• Company Overview
• Financial Analysis
• Competitive Position
• Management Assessment
• Risk Factors
• Valuation
• Investment Recommendation (Buy/Hold/Sell)

Customizing the Prompt

You can modify the investment report generation by editing:

• src/improved_prompt.md - Primary prompt template with structured format
• company-info.md - Fallback prompt template

The system will first try to use the improved prompt, then fall back to the original if needed.

Example Searches

After loading news for a company, you can search for:

• "lawsuit" - Find news about legal issues
• "financial results" - Find articles about earnings
• "new products" - Find announcements about products
• "management changes" - Find news about executive shuffles

Technical Implementation

The system consists of several main components:

1. YahooFinanceNewsCollector: Retrieves news articles for a given ticker symbol using yfinance API
2. VectorDatabase: Stores and indexes articles using ChromaDB for semantic search
3. NewsArticle: Data structure for representing financial news articles
4. DeepSeek Integration: Connects to Ollama for accessing DeepSeek-r1 1.5B model capabilities
5. InvestmentReportGenerator: Creates detailed investment analysis reports based on collected news

Data Storage

The vector database is stored in a directory called chroma_db in the project root. This allows for persistent storage and retrieval of financial news data across multiple runs.