The Stock Analysis Assistant is a sophisticated AI-powered application designed for stock analysis and investment recommendations. Leveraging advanced Agentic Retrieval-Augmented Generation (RAG) architecture, the system extracts insights from financial news and provides actionable recommendations tailored to user queries.
The project is implemented using the following:
- LangChain for building agents and handling LLM interactions.
- Chroma for document retrieval and semantic search.
- Streamlit for an interactive user interface.
- Finnhub API for fetching real-time financial news and data.
-
app/
main.py: Entry point for the Streamlit-based application.data/: Stores input data and vectorized documents.
-
notebooks/
agentic_rag.ipynb: Jupyter notebook for experiments and testing the architecture.
-
src/
backend.py: Implements the RAG-based architecture and the agent functionalities.__init__.py: Initialization file for thesrcpackage.
-
utils/
- Contains helper scripts for configuration and shared utilities.
-
tests/: Folder for unit tests to ensure code reliability.
The application employs an Agentic RAG workflow for stock analysis, as outlined in the diagram above:
- Connects to the Vector Database (Chroma) to retrieve relevant documents based on the user's query.
- Embedding model:
granite-embedding.
- Evaluates the relevance of retrieved documents to the query using an LLM.
- Generates binary relevance scores ("yes" or "no") based on the semantic overlap with the query.
- When retrieved documents are not relevant, the agent fetches the latest financial news using the Finnhub API.
- Extracts company ticker symbols and retrieves recent news articles.
- Processes relevant documents or news articles to generate investment recommendations.
- Utilizes prompt-based LLMs for actionable insights.
- Embeds data into a vectorized format using Ollama Embeddings.
- Efficient search of documents relevant to stock-related questions.
- Grades the retrieved documents using a custom ChatPromptTemplate.
- Filters irrelevant data before passing information downstream.
- Leverages Finnhub API to fetch real-time financial news based on the query.
- Ensures the recommendations are grounded in the latest market trends.
- Uses the Recommender Agent to suggest actionable investment strategies.
- Examples: Buy/Sell/Hold decisions based on recent news insights.
- Python 3.9+
- Finnhub API Key (Add it to a
.envfile)
- Clone the repository:
git clone https://github.com/mlengineershub/Stock-Analysis-Assistant cd Stock-Analysis-Assistant - Install dependencies:
pip install -e . pip install -r requirements.txt - Make sure you have llama3.2:latest:
ollama run llama3.2:latest
- Run the Streamlit application:
streamlit run app/main.py
- Launch the application using Streamlit.
- Enter a query such as:
- "What are the latest updates on Apple?"
- "Should I invest in Tesla?"
- The system processes the query through the Agentic RAG architecture and provides:
- Relevant financial insights.
- Actionable investment recommendations.
src/backend.py- Implements the complete RAG pipeline including retrieval, grading, search, and recommendation nodes.
.env- Stores sensitive API keys for external services like Finnhub.
- LangChain: RAG pipelines and agent orchestration.
- Chroma: Vector database for document retrieval.
- Ollama Embeddings: High-quality embeddings for semantic search.
- Finnhub API: Financial news and market data.
- Streamlit: Interactive front-end for user queries and recommendations.
- AIT SAID Azzedine
- DJERFAF Ilyes
- KESKES Nazim
This project is licensed under the MIT License. See the LICENSE file for details.
- Integration of advanced sentiment analysis for financial news.
- Support for multiple financial APIs.
- Historical performance tracking of recommendations.
Feel free to open issues or contribute via pull requests. For inquiries, contact us :).