wtf is vibe-research?

Vibe Research is a personal investment research dashboard covering Chinese A shares, along with US and Hong Kong stocks. It does not tell you what to buy or sell. Instead it gathers price data, valuation numbers, financial reports, company announcements, fund flows, and news, and lays it all out on a clean dashboard, leaving the final judgment to an AI model that you connect yourself. The dashboard brings together several pieces in one place: a daily market recap covering both the overall index and individual stocks you are watching, a news radar that pulls from over a hundred public feeds across a dozen categories and lets AI summarize the day's key points, detailed pages for individual stocks with valuation history, financial highlights, and fund flow data, a watchlist you can paste stock codes into, a sector overview, a record of your current and past holdings, a place to upload your own research reports, and a running log of your research notes and AI conversations you can look back on later. One notable feature is a bull versus bear debate mode. The backend first pulls together a set of objective facts about a stock, then has separate bullish and bearish AI researchers form independent arguments from that same data, optionally challenging each other, before a neutral moderator summarizes where they agree, where they genuinely disagree, and what data is still missing. It is designed to produce a clearer picture of the debate, not a buy or sell recommendation. All of the underlying market data comes from open source data toolkits that are included directly inside the repository, so nothing extra needs to be downloaded or connected separately to get started. You can connect your own AI in three ways: by using a coding assistant CLI you are already logged into, such as Claude Code, Codex, Qwen Code, or DeepSeek CLI, without needing a separate API key, by entering your own API key for a provider like DeepSeek, OpenAI, or OpenRouter, or by exposing the backend as an MCP server that an agent like Claude Code can call directly. The backend is written in Python with FastAPI, and the frontend uses React 19 with TypeScript and Tailwind. It is released under the MIT license.