simonlin1212/vibe-astock

What it solves

Vibe-Astock automates the tedious process of short-term stock market review (fupan) for the A-share market. Instead of manually tracking limit-up pools, board sequences, and capital flows, it aggregates data and uses AI to synthesize a readable market judgment, saving traders from hours of repetitive manual data gathering.

How it works

The system separates hard data from AI analysis. It first calculates "derived sentiment indicators" (like profit effect and promotion rates) using pure computation from sources like akshare and Tencent Finance. Then, it employs a multi-agent architecture where five specialized analysts (Sentiment, Capital, Theme, Hot Money, and Leader tracking) process the data, and a "Judge" agent converges their findings into a structured market judgment.

Who it’s for

Short-term A-share traders who need a systematic way to review daily market sentiment, track their own trading logs and risk boundaries, and perform deep-dives into specific stocks without receiving biased investment advice.

Highlights

• Derived Sentiment Indicators: Focuses on medians and promotion rates rather than simple averages to accurately gauge market "vibe."
• Multi-Agent Analysis: Uses a pipeline of five analysts and a judge to turn raw data into a narrative.
• Local-First Trading Journal: A private, local-only ledger for tracking trades, risk-at-risk capital, and execution deviations.
• Flexible LLM Integration: Supports OpenAI-compatible APIs or local CLI subscriptions (e.g., Claude) to avoid API key requirements.
• Customizable Analysis Logic: Allows users to replace the "Prompt Pack" to change the style and requirements of the AI analysts.
• Market Snapshotting: Automatically captures market state at specific time intervals (e.g., 09:25, 10:00) to track the intraday sentiment path.