This is a submission for the Hermes Agent Challenge: Build With Hermes Agent
🛠️ What I Built
Navigating the noise of the stock market requires serious algorithmic filtering and rapid data synthesis. Hermes Quant is a fully autonomous, localized AI financial analyst designed to eliminate information overload for retail investors and algorithmic traders.
Instead of manually checking multiple screeners, technical charts, and news feeds, traders can deploy Hermes Quant to run deep, multi-step quantitative analyses on any global or domestic ticker (from AAPL to RELIANCE). The application instantly aggregates live technical indicators, quarterly fundamentals, and breaking news sentiment into a cohesive, actionable investment thesis.
Best of all? It runs entirely locally. 🔒 By leveraging Hermes 3, the platform provides institutional-grade AI equity research with zero API costs and total data privacy.
🎥 Demo
📸 The Agent in Action:
Generated Pdf
💻 Code
⚙️ My Tech Stack
- 🧠 Core Agent: Hermes 3 (running locally via Ollama for zero-cost inference)
- 🔗 Agent Orchestration: Python, LangChain / LangGraph
- 🖥️ Frontend: React / Tailwind CSS
- 📊 Market Data APIs: yfinance / AlphaVantage / NewsAPI
🤖 How I Used Hermes Agent
Hermes 3 is not just a text generator in this project; it is the core reasoning and orchestration engine driving the entire application. I leaned heavily into its exceptional tool-calling and ReAct (Reasoning and Acting) capabilities.
When a user requests an analysis, Hermes enters an autonomous, iterative execution loop:
- 🔍 Observation & Tool Calling: It decides which specific financial tools to call based on the asset. For example, it autonomously triggers
get_market_sentiment(AAPL)to parse headlines or fetches RSI/MACD indicators. - 🧠 Live Reasoning: It evaluates the returned JSON data in real-time, matching overbought/oversold signals against market sentiment to detect hidden risks.
- 📝 Synthesis: It stops the loop once it has sufficient data, formatting the complex financial metrics into a clean, easy-to-read "Risk Divergence Matrix" and Executive Summary.
Hermes was the perfect fit for this use case because financial analysis requires strict adherence to system prompts and highly reliable, deterministic JSON tool-calling without hallucinating metrics—capabilities where Hermes absolutely excels.
📊 Example Walkthrough: Analyzing AAPL
To see Hermes Quant's agentic capabilities in action, here is how the agent autonomously evaluated Apple Inc. (AAPL):
- 📉 Step 1 (Technicals): The agent fetched core indicators, identifying that while the price was above the SMA50, the RSI was sitting at 79.0, flagging a highly overbought condition.
- 📰 Step 2 (Sentiment Extraction): Hermes called the sentiment tool, parsed recent macro headlines, and computed a neutral aggregate sentiment score of 0.0928.
- ⚠️ Step 3 (Divergence Detection): The agent cross-referenced these data points, catching a critical macroeconomic mismatch: highly bullish medium-term price action versus completely neutral, mixed market sentiment.
- 📑 The Final Artifact: It successfully generated a comprehensive report advising a contrarian strategy, entirely on its own.
Built by Atharva Atal