BettaFish (微舆): A Multi-Agent System for Public Opinion Analysis

Core Capabilities and Strategic Positioning: A Decision Support System Beyond Sentiment Analysis

BettaFish (also known as “微舆”, WeiYu) is an open-source multi-agent system purpose-built for public opinion analysis. Its strategic positioning is not a simple personal tool, but rather a system designed to provide enterprise managers, researchers, and analysts with in-depth, comprehensive, and actionable decision support [[1,3]]. The core value of the project lies in precisely addressing three major pain points in the modern information environment: cognitive limitations caused by information cocoons, bias and one-sidedness resulting from single-model analysis, and the information silo problem between public-domain opinion and private-domain data [[6,29]]. By building an automated, multi-dimensional analysis pipeline, BettaFish aims to break down information barriers, restore a true picture of public opinion, and predict future trends, thereby supporting critical decisions [[7]].

BettaFish's core capabilities go far beyond traditional text sentiment classification; it provides a complete end-to-end solution. First, at the data collection layer, the system uses an AI-driven crawler cluster to achieve 7x24-hour uninterrupted monitoring and data capture across more than 30 mainstream social media platforms both domestically and internationally (such as Weibo, Xiaohongshu, Douyin, Kuaishou, etc.) [[1,9]]. This breadth and depth of data source coverage is the foundation for generating comprehensive insights. Second, the system has powerful multimodal content understanding capabilities, capable of deep parsing of unstructured data such as short videos and images [[3,5]]. For example, the Media Agent can use Playwright to capture videos, leverage automatic speech recognition (ASR) to extract audio transcriptions, extract visual information from keyframes through optical character recognition (OCR) and scene analysis, and parse structured information cards returned by search engines, such as weather and stock data [[5,6]]. This capability is crucial for analyzing a modern social media ecosystem dominated by video. Finally, one of the most strategically valuable capabilities is its seamless fusion of public and private domain data [[1,12]]. Through the Insight Agent, users can securely connect to internal business databases, combining external public opinion trends with internal sales data, customer feedback, product performance metrics, and more, thereby enabling more precise brand reputation management, market research, and crisis early warning [[6]].

Ultimately, all analysis results are presented to users in the form of structured HTML reports [[3,9]]. This demonstrates that BettaFish's design philosophy is to free analysts from tedious data processing and directly output easy-to-consume decision evidence. The system includes a variety of built-in report templates and can automatically select the most appropriate template based on the analysis topic, further enhancing its value as a decision-making brain [[3,9]]. In summary, BettaFish is strategically positioned as a highly specialized, decision-enabling automated analysis platform, rather than a general-purpose Q&A bot or a simple code editor.

Technical Architecture Dissected: The Innovative Collaboration Mechanism of “Agent Forum”

The technical soul of BettaFish lies in its original “Agent Forum” collaboration engine, which is the key technical highlight that distinguishes it from most AI Agent frameworks on the market [[1,12]]. This architecture is not simply a chain of multiple tasks; rather, it simulates a discussion group composed of experts, reaching consensus through debate, reflection, and iteration, thereby significantly improving the depth, accuracy, and interpretability of analysis [[6,9]]. The entire system is written in pure Python, featuring a lightweight and highly modular design, making it easy to extend and maintain [[1,5,7]].