// the find
bklieger-groq/stockbot-on-groq
StockBot powered by Groq: Lightning Fast AI Chatbot that Responds With Live Interactive Stock Charts, Financials, News, Screeners, and More. Powered by Llama3-70b on Groq, Vercel AI SDK, and TradingView Widgets.
A Next.js chatbot that wraps TradingView widgets and Llama3-70b on Groq — you ask about a stock, it picks the right TradingView component and renders it inline. The two-call architecture (intent classification → widget selection) is the core mechanic. Aimed at developers who want a working example of generative UI with tool calling, not at anyone who needs actual trading functionality.
The widget-dispatch pattern is well-executed: the model classifies intent and returns a structured response that maps to a specific TradingView component, which is a clean demonstration of function calling. One-click Vercel deploy with a single env var makes it genuinely easy to fork and run. TradingView widgets handle all the hard data problems (live prices, charts, screeners) so the app never has to manage a data feed. The component library under `components/tradingview/` is modular enough to pull individual widgets into another project.
No persistence — every page refresh loses the conversation, which limits it to toy status. The model choice is hardcoded to Llama3-70b and the `model-selector.tsx` component in the tree suggests UI exists for switching models but it's unclear if it works end-to-end. No error handling when TradingView widgets fail to load or the Groq API rate-limits, so the chat just silently breaks. The whole thing is a Groq marketing demo, so the repo sees maintenance spikes around Groq announcements and then goes quiet — last push was December 2025 but the core logic hasn't changed meaningfully in over a year.