AlphaForge

AlphaForge is a modular, multi-agent AI research and trading workflow built as a submission for the Google Agent Development Kit Hackathon. This project demonstrates how to orchestrate multiple specialized AI agents using the Agent Development Kit and Google Cloud for robust, explainable, and automated financial decision-making.

---

Key Features

• Parallel Analyst Agents: Technical, Sentiment, News, and Fundamental analysts run in parallel to generate comprehensive market reports.
• Debate Simulation: Bullish and Bearish researchers debate using LangGraph, with a judge and synthesizer agent providing verdict and summary.
• Risk Synthesis: Aggressive, Neutral, and Safe risk debators generate plans, which are synthesized into a strategic risk recommendation.
• Trader Agent: Makes a final actionable trading decision (BUY/HOLD/SELL) based on all synthesized insights.
• Reflection Agent: Checks the trader's decision for hallucinations or unsupported claims, ensuring reliability.
• Automated Workflow: The entire process is orchestrated in a single pipeline, with feedback loops for correction if hallucinations are detected.

---

Workflow Overview

flowchart TD
    A[Start: Input Ticker] --> B[Parallel Analysts]
    B --> B1[Technical Analyst]
    B --> B2[Sentiment Analyst]
    B --> B3[News Analyst]
    B --> B4[Fundamental Analyst]
    B1 --> C
    B2 --> C
    B3 --> C
    B4 --> C
    C[Researcher Debate – Bull vs Bear]
    C --> D[Debate Judge]
    D --> E[Debate Synthesizer]
    E --> F[Risk Synthesizer]
    F --> G[Trader Agent]
    G --> H[Reflection Agent]
    H -- Hallucination Detected --> G
    H -- No Hallucination --> I[Final Decision Output]

Loading

---

Installation

1. Clone the repository:

   git clone https://github.com/SRINJOY59/Google-Agent-Development-Kit
   cd Google-Agent-Development-Kit

2. Install dependencies:

   pip install -r requirements.txt

3. Set up your .env file with the required API keys and environment variables for Google Cloud and Agent Development Kit.

---

Usage

Run the main workflow with your desired ticker:

python Agents/workflow.py

Run the streamlit UI

streamlit run app.py

You can customize the pipeline or run individual agents for testing and development.

---

Project Structure

Google-Agent-Development-Kit/
│
├── Agents/
│   ├── Researcher/
│   ├── Analyst/
│   ├── Risk_Management/
│   ├── Trader/
│   ├── Reflection/
|   |── custom
|   |── app.py
│   └── workflow.py
├── README.md
├── requirements.txt
└── ...

---

Customization

• Add or modify analyst agents for new data sources or asset classes.
• Extend the debate or risk synthesis logic for more complex scenarios.
• Integrate with real-time data feeds or trading APIs.
• Deploy on Google Cloud for scalable, production-grade agent orchestration.

---

License

MIT License

---

Submission for the Google Agent Development Kit Hackathon:
AlphaForge — Modular, explainable, and robust multi-agent AI for financial research and trading, built for the Agent Development Kit hackathon