Developer builds LLM trading bot using visual analysis and debate prompts

A developer details their journey in creating a trading bot that leverages large language models (LLMs) for financial analysis. Initially, the bot failed by relying on text-based LLM outputs, but the developer found success by feeding candlestick chart images directly to Google Gemini 3.6 Flash for visual inference. The system was further optimized with custom, JIT-compiled indicator engines and a novel prompt structure that forces the LLM to debate both bullish and bearish cases simultaneously, improving analytical rigor and efficiency. AI

IMPACT Demonstrates novel applications of LLMs in visual analysis for specialized domains like financial trading.

RANK_REASON Developer blog post detailing a custom tool built using LLMs and other libraries.

Read on dev.to — LLM tag →

• Bollinger Bands
• Google Gemini 3.6 Flash
• Ichimoku
• LLM
• Numba
• NumPy
• Relative strength index
• stochastic
• TTM Squeeze
• volume

• product
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →