Preamble¶

What people who monitor the market really lack is not “more data,” but “answering the right questions at the right time.” The traditional approach is to search through thirty open tabs, which scatters attention and easily leads to missing critical moments.

The @lisonevf/dsh-stock-panel introduced below is a Web plugin mounted inside DeepSeek Harness (DSH). It does not provide additional market data; instead, it transforms the right-side view slot of the DSH session page into an A-share monitoring execution console organized by a time-based workflow.

Plugin Introduction¶

What It Is¶

This is a DSH Web plugin maintained by LisonEvf. Its core purpose is to turn the AI session window into an A-share monitoring execution console, with the interface organized around a trading timetable: post-market review → next-day call auction comparison → answering only three questions during trading → finalizing positions near the close.

Core Features¶

The plugin includes the following functional modules:

1. Market Dashboard
   Default entry point. Displays advancing/declining counts, index charts, limit-up tiers, rankings, and market anomalies. The whole page does not scroll, while individual blocks scroll internally. The number of columns adapts automatically to container width (three-column tier at 1440px, two-column tier at 780px).

2. Seven-Step Post-Market Review
   During the post-market phase, the plugin provides a seven-step review process, outputs a next-day expectation list (≤5 items / 7 fields), and archives it.

3. Self-Discovered Sectors
   Uses unsupervised co-movement clustering + LLM naming and categorization. Instead of relying on labels, it looks at capital flows to identify sectors that official classifications may lag on recognizing. Each category includes an intensity bar and source capsules, and supports renaming.

4. Individual Stock Workbench
   Clicking any row in the left panel enters that stock’s view: quote header + daily K-line/intraday chart + capital flow/tick-by-tick/call auction + right-panel AI decision card.

5. Stock Screening
   Supports snapshot condition filtering and MA signal filtering, and includes AI ranking functionality (with an AI#n badge).

6. Monitoring Alerts
   Pushes Toasts and badges through AlertWatcher, remains active resident, and does not occupy the main area.

7. One-Click AI Judgment
   The model reads market context and outputs a decision card (stance / confidence score / supporting points / risk), then fills in key price levels and rankings. Results are always read-only.

Design Principles¶

• Read-only Results: AI-generated conclusions (such as decision cards, rankings, and sectors) are never automatically written into your list. You must manually click “+” to write an entry into the list. Items already in the list are marked rather than inserted again.
• Density Management: The interface uses strict layout conventions such as no full-page scrolling, internal scrolling, and equal-width grids, ensuring controllable information density.

Installation and Enablement¶

Before installing, make sure DSH is already installed (npm install -g @deepseek-ai/dsh).

1. Run the installation command:

    dsh plugin --profile web add https://github.com/LisonEvf/dsh-stock-panel/releases/download/v1.6.0/lisonevf-dsh-stock-panel-1.6.0.tgz

2. After installation is complete, restart dsh web.
3. Hard refresh the page in the browser (Ctrl+Shift+R).
4. An A-Share Workbench tab will appear on the session page.

Typical Usage¶

• Post-market Review: Complete the seven-step process in the review view to generate the next-day expectation list.
• Next-day Call Auction: Compare list items one by one against the call auction results.
• Intraday Operation: Answer only three key questions and make decisions based on the workbench data.
• Final Position Setting Near Close: Decide position actions based on that day’s performance.

Applicable Scenarios and Notes¶

• Data Sources: The plugin includes 19 built-in data tools (market quotes, K-lines, intraday charts, tick-by-tick data, call auctions, sector constituents, capital flow, etc.) and 12 persistent tables. You configure the data endpoints yourself. The plugin does not build in or distribute any market data.
• Functional Positioning: It is a data statistics and presentation tool, not a full-featured quant platform (it does not include vectorized backtesting, financial data pipelines, etc.).
• Disclaimer: The plugin does not recommend stocks, predict price movements, provide buy/sell points, display win rates, or promise returns. All analysis results are for reference only. Users must make independent judgments and bear investment risk.

Summary¶

@lisonevf/dsh-stock-panel is suitable for users who want to establish a standardized A-share monitoring workflow within the DSH ecosystem. Through its time-scheduled view design and strict read-only AI interactions, it helps users escape information overload and focus on execution-level decisions.

For complete documentation and configuration details, see: GitHub Repository