🤖 Autonomous Financial Research Agent (v2.0)# Autonomous Financial Research Agent (AI Systems Architecture)

🏗️ System Architecture

Unlike traditional linear scripts (Data -> Model -> Plot), this system utilizes a Cyclic State Graph to reason, plan, and execute research tasks autonomously.

The agent autonomously loops between "Thinking" (Gemini) and "Acting" (Tools) until it resolves the user's query.

---

🚀 Project Overview

This project represents a strategic shift from "Model Training" to "AI Systems Architecture."

Instead of simply training a sentiment model, I engineered an Autonomous Agent capable of performing the role of a Junior Financial Analyst. The agent uses Neurosymbolic Reasoning to not just summarize news, but to actively look for "Logical Divergences" (e.g., when sentiment is positive, but the stock price is crashing).

💡 Key Innovations (The "2026 Standard")

1. Agentic Orchestration (LangGraph): The system uses a graph-based state machine to handle complex, multi-step reasoning loops.
2. Neurosymbolic Guardrails: Integrates neural intuition (FinBERT Sentiment) with symbolic logic (Price Trend verification) to hallucination-proof the output.
3. Governance & Auditability: Implements a custom Throttling Layer to ensure compliance with API rate limits (mimicking enterprise ISO 42001 operational standards).
4. Model Context Protocol (MCP): Tools are designed as modular, stateless functions ready for integration with larger multi-agent swarms.

---

🛠️ Tech Stack

Component	Technology	Role
Orchestration	LangGraph	The "Brain" ensuring cyclic reasoning and state management.
LLM	Google Gemini 1.5 Flash	The cognitive engine responsible for planning and synthesis.
Sentiment Engine	FinBERT (Hugging Face)	Specialized financial sentiment analysis model.
Data Tools	yfinance & NewsAPI	Live market data and breaking news fetchers.
Governance	Python (Custom Logic)	Rate-limiting and error handling protocols.

---

📂 Project Structure

Financial_Agent_2026/
│
├── agents/               # 🧠 The Cognitive Layer
│   ├── __init__.py
│   └── researcher.py     # LangGraph State Machine & Logic
│
├── tools/                # 🛠️ The Operational Layer (MCP)
│   ├── __init__.py
│   └── market_tools.py   # Stock, News, and Sentiment Tools (w/ Throttling)
│
├── .env                  # 🔒 API Keys (Not uploaded to GitHub)
├── main.py               # 🚀 Entry Point
├── requirements.txt      # Dependencies
├── README.md             # Documentation
└── Screenshot of outcome.png