dhaneswara/ai-stocks-news-analysis

📈 MarketCortex

A personal web app for swing/long-term US-stock decision support. It pulls price data, fundamentals, and recent news for a stock, computes simple technical indicators, and asks an LLM (Anthropic / OpenAI / Gemini / DeepSeek / local Ollama) to produce a structured analysis — a plain-language summary, a read on the news, and buy/sell signals drawn directly on an interactive chart with the reasoning shown on the page. It can also send scheduled buy/sell alerts to Telegram, rank opportunities across the S&P 500 — or focus on just your own portfolio (your watchlist plus the companies in your active knowledge graph), scored the same way but in seconds — map inter-company relationships as a knowledge graph, run an agentic Deep Analysis that pulls evidence on demand, and grade how accurate its own past calls turn out to be — for every signal source it produces (fast LLM, deep LLM, technical screen, network-blended). Or just chat with it — a multi-turn assistant that reasons step-by-step across any stock or theme and pulls whatever the question needs (prices, news, geopolitics, the knowledge graph, your portfolio).

⚠️ Decision support, not financial advice. The LLM can be confidently wrong, and the on-chart/alert signals are mechanical heuristics + retrospective reasoning — not a backtested, validated strategy. This project does not execute trades.

---

Features

• Interactive dashboard — candlestick chart (TradingView Lightweight Charts) with a timeframe selector (1M / 3M / 6M / 1Y / 2Y / 5Y), SMA50/200 overlays, and LLM-drawn buy ▲ / sell ▼ markers. Click a marker — or a row in the dedicated Signals list — to read that signal's reasoning. Your last analysis restores automatically when you reopen a ticker — the reasoning, the Signals list, and the chart's buy/sell markers reappear, tagged as of its call date — a pure read that costs no tokens and never affects evaluatio