666ghj/bettafish

BettaFish is a multi-agent public opinion analysis system that monitors Chinese social media platforms and generates structured research reports about trending topics, brand sentiment, and public discourse. Its Chinese name translates roughly to Micro Debate, a wordplay on the project's fish mascot. The problem it solves is that understanding what the general public thinks about a topic, event, or brand across dozens of social media channels is extremely time-consuming to do manually. BettaFish automates this by deploying multiple specialized AI agents that work in parallel to collect, analyze, and synthesize data. The system is organized into four main engines, each implemented as a separate Python module with its own agent logic. QueryEngine searches news and websites domestically and internationally. MediaEngine handles multimodal content such as short videos and images from platforms like Douyin and Xiaohongshu. InsightEngine digs into private or internal databases for deeper analysis. ReportEngine collects everything the other agents found and generates an interactive HTML report. The agents use a forum-style debate mechanism where a moderator model prompts the agents to argue their findings against each other through multiple rounds, intended to reduce the blind spots of any single model. The system calls large language models through an OpenAI-compatible interface, so it can work with various providers. It uses Flask as the web server and supports streaming responses. You would use BettaFish if you need to monitor Chinese social media sentiment for a brand, research public reaction to news events, or analyze opinion trends at scale. It is built entirely in Python without depending on heavyweight agent frameworks, making it relatively straightforward to customize with your own models and data sources.