SemiMarketAgent

This project combines a MeTTa knowledge graph and Fetch.ai uAgents framework to allow users to analyze semiconductor markets with institutional-grade intelligence instantly. The agent is designed to aggregate news from multiple sources (NewsAPI, Google News, Yahoo Finance), filter articles using LLM reasoning, and synthesize real-time stock data with structured investment knowledge. Behind the scenes, it uses ASI:One for natural language processing and yfinance for live market data, delivering comprehensive analysis through an autonomous chat interface on Agentverse. The system processes complex queries like 'Analyze NVIDIA based on recent news' by simultaneously fetching articles, stock prices, and knowledge graph insights, then combining them into professional investment reports with full source attribution.

This project uses the Fetch.ai uAgents framework behind the scenes to connect to NewsAPI and Yahoo Finance APIs. We used MeTTa symbolic reasoning to design the knowledge graph and the backend is built in Python.

The architecture combines three core technologies in a unique way. MeTTa's hypergraph structure stores semiconductor company data as symbolic atoms, enabling complex relationship reasoning between market trends, company fundamentals, and investment factors. Unlike traditional databases, MeTTa allows us to express relationships like (company NVIDIA (segment "AI chips") (risk "cyclical")) and perform pattern matching queries directly.

Fetch.ai uAgents runs our agent autonomously on Agentverse with ASI-compatible chat protocols. The agent processes natural language queries, classifies intents (news analysis, stock tracking, company research), and coordinates parallel data collection threads. This eliminates the need for traditional REST APIs or web servers.

ASI:One LLM handles the heavy lifting for natural language understanding and response synthesis. We implemented intelligent news filtering where the LLM ranks 50+ articles and selects the top 15 most relevant ones, significantly reducing noise.

Partner technology benefits: MeTTa's symbolic reasoning lets us encode investment logic as executable patterns rather than hardcoded rules. Fetch.ai's agent framework provides built-in networking, message passing, and autonomous execution capabilities we'd otherwise need to build from scratch.

Notable hacks: We implemented robust JSON parsing fallbacks when LLM responses aren't properly formatted. The system tries multiple parsing strategies - full JSON, regex pattern extraction, and finally raw number extraction - ensuring reliability even with unpredictable AI outputs. We also use threading to parallelize news fetching from multiple sources (NewsAPI, Google News RSS, Yahoo Finance) while maintaining real-time performance.

The result is a Wall Street-grade analyst that synthesizes multi-source intelligence in under 30 seconds, operating 24/7 without human intervention.