Chatbot that answers stock questions with live TradingView widgets

What it does StockBot is a chat interface that interprets natural-language questions about markets and replies by embedding live TradingView widgets—candlestick charts, heatmaps, news feeds, screeners—directly into the conversation. It routes requests through Llama3 70b running on Groq, uses the Vercel AI SDK to manage the generative UI, and leans entirely on TradingView for the actual financial data and visualization. In short, it is well-polished glue code that makes a fast LLM act as a smart router for third-party market widgets.

The interesting bit The README highlights Groq’s inference speed as the critical ingredient: the pipeline makes two specialized API calls to interpret the query and shape the response, and the low latency keeps the interaction from feeling sluggish. That makes the project a useful reference for how “fast” changes user expectations when an LLM is driving dynamic, embedded interfaces rather than just returning text.

Key highlights

• Uses Llama3 70b on Groq for tool-calling and response generation.
• Dynamically selects from nine TradingView interfaces (market overview, ETF heatmap, stock screener, etc.) based on the prompt.
• Covers stocks, forex, bonds, and cryptocurrencies.
• Built atop the Vercel AI Chatbot template, so it deploys to Vercel by default.
• Explicitly disclaims accuracy: intended for entertainment and instructional use only.

Caveats

• The README warns that StockBot may provide inaccurate information and explicitly does not provide investment advice.

Verdict A solid reference if you’re prototyping generative UI chatbots and want to see how low-latency inference affects the feel of tool-calling. Avoid it as a trading research tool; the value is in the integration demo, not financial reliability.