This project demonstrates a reasoning agent that acts as an expert financial advisor, capable of breaking down complex investment questions and providing structured, step-by-step recommendations.
- Expert Financial Reasoning: The agent specializes in investment strategies, risk assessment, and portfolio recommendations.
- Step-by-Step Explanations: Each response is broken down into clear reasoning steps, showing the agent’s thought process.
- Tool-Enhanced Reasoning: Uses explicit tool calls to analyze and justify recommendations.
- Transparent Output: Shows intermediate reasoning, tool calls, and the final answer in a readable format.
The agent is powered by the agno framework and uses the Nebius Llama-3.3-70B-Instruct model. It is configured to:
- Break down questions into component parts
- State assumptions and key variables
- Develop structured reasoning paths
- Evaluate evidence and consider alternatives
- Provide well-justified, balanced investment advice
Running the agent with the prompt:
Recommend an investment strategy for a client with moderate risk tolerance
Produces a detailed, step-by-step reasoning process and a final recommendation, as shown below:
-
Install dependencies (ensure you have Python 3.8+ and agno):
pip install agno python-dotenv
-
Set your Nebius API key in a
.envfile:NEBIUS_API_KEY=your_api_key_here -
Run the agent:
python3 main.py
-
View the output: The agent will print its reasoning and final answer to the terminal.
- Edit the
instructionsinmain.pyto change the agent’s expertise or reasoning style. - Modify the prompt to ask different financial or investment questions.
main.py— Main entry point for the reasoning agent.README.md— Project documentation.demo.png— Example output screenshot.