Alexandria Felicia Seanne | feli

[AFsenna]

I have developed a Multi-Agent Financial Research Workflow that analyzes the conglomerate holdings of Prajogo Pangestu on the IDX.
Workflow Pattern: I implemented a Sequential Chain Pattern.
Agent Roles:

• Agent 1 (Screener): Dynamically identifies the relevant tickers from the exchange based on a natural language query.
• Agent 2 (Researcher): Takes the output from the Screener to perform a deep-dive financial analysis, retrieving real-time market caps and 52-week price ranges.

Project Link: https://colab.research.google.com/drive/1odsJy5Xxv6DxGqigLzD_3h22KH1TE_Jf?usp=sharing

[onlyphantom]

This is a really good effort. Thank you for submitting your work.

A small feedback is that even your matplotlib tool can be turned into an agent tool, like the sample implementation below:

@function_tool
def create_dashboard(data_json: str) -> str:
    data = json.loads(data_json)
    df = pd.DataFrame(data['data'])
    ...
    
@function_tool
def display_summary_table(data_json: str) -> str:
    ...

visualization_reporter_agent = Agent(
    name="Visualization Reporter",
    instructions="""You are a financial visualization expert. Use create_dashboard and display_summary_table to create visualizations and summary tables based on the research data.""",
    tools=[create_dashboard, display_summary_table]
)

Will mark this as a Passing grade!