Local-first AI that turns A-share chatter into research notes

What it does

easy-stock is a local-first desktop app built with a Go backend and an Electron/React frontend that pulls A-share market data from sources like East Money and Sina into a single research pane. It uses a built-in browser and scheduling agent to auto-collect posts from finance influencers across Xueqiu, Taoguba, and WeChat, then cleans, deduplicates, and archives them locally so an AI model can summarize cross-author consensus and map sentiment cycles. The tool also runs portfolio health checks for concentration risk, style mismatch, and correlation across up to ten positions.

The interesting bit

Most retail trading tools either dump raw quotes or push black-box buy signals; easy-stock treats research as an audit trail, tagging every AI-generated theme score and consensus conclusion with its original URL, update time, and data-source latency. The app even classifies individual stocks into paths like “emotional consecutive board” or “trend growth” before generating customized reports, effectively forcing the AI to declare its reasoning framework upfront.

Key highlights

• Aggregates market data (indices, K-lines, capital flow, limit-up ladders) and influencer posts into one local-first desktop workspace.
• Auto-collects and cleans “Big V” commentary via a built-in browser agent, then generates cross-author consensus reports with verifiable source links.
• Maps short-term sentiment cycles through limit-up promotion rates and emotional time-series, alongside longer-term theme-radar views.
• Runs portfolio AI inspections for up to ten holdings, checking concentration, correlation, style match, and per-stock action conditions.
• Stores all articles, analysis records, and AI sessions locally; carries a non-commercial license.

Caveats

• The project is strictly A-share specific; every data source and terminology is rooted in China’s domestic market.
• Output quality depends on upstream data availability and cache freshness, and the README notes that scores, themes, and risk parameters fluctuate with source latency and degradation.
• License is explicitly non-commercial, so integrating it into a paid product or service is off the table.

Verdict Worth a look if you are a retail trader or quant-curious developer researching China’s A-share market and want your AI to cite sources. Skip it if you need multi-market coverage, commercial deployment, or a fully open license.