How I Built a Personal AI‑Powered Asset Tracker: Prompts, Features, and Workflow

Initially I asked AI a simple question—"I want a software to display currency prices, what should I use?"—to get direction, then refined the request to include Binance Alpha, crypto and US stocks, establishing the three core asset categories for the tool.

Prompt Development

By iteratively adding concrete requirements such as "show Binance Alpha currencies, crypto, and US stocks," the AI generated a first version of the page layout, demonstrating that starting with the core need and letting AI flesh out the details is more effective than writing a perfect prompt from the start.

Integrating Personal Record Keeping

The initial price‑only view proved insufficient; I needed to relate assets to my own purchase records. I added fields for buy quantity and buy amount, letting the system compute the average price: 买入均价 = 买入金额 ÷ 买入数量 Each asset row now displays current price, 24h change, buy quantity, buy amount, average price, market value, profit and profit rate, turning the page into a true holdings workbench.

Low‑Key, Direct UI Design

To keep the panel unobtrusive, I removed large titles and visual flourishes, arranging the three sections vertically (Alpha, Crypto, US) with individual search boxes that include an inline add button, ensuring the interface is quick to understand and use.

Feature Iteration Driven by Specific Prompts

Concrete feedback prompts guided functional additions, for example:

排序的时候按照买入金额从大到小展示。

This sorted assets by purchase amount, placing the most invested items at the top. Another prompt added a "追" (add) function that automatically accumulates new purchase quantities and amounts into the existing totals, eliminating manual calculations.

K‑Line Handling

Attempting to embed a K‑line chart proved unsatisfactory, so I changed the requirement to provide a button that links to external chart providers such as TradingView or Dexscreener, leveraging mature services instead of reinventing the wheel.

Remark Functionality

Notes became essential for tracking why an asset was bought, its source, logic, risk points and follow‑up observations, turning the tool into a personal investment diary that aids later review.

Local JSON Storage

Instead of a database, all selections, holdings and remarks are saved in data/local-state.json. This keeps data on the local machine, avoids cloud dependencies, allows direct file inspection, simplifies migration and suits a lightweight personal tool.

Running the Tool

The dashboard runs as a local service; after starting the server, it is accessed via http://localhost:4173/. Directly opening index.html is insufficient because price APIs and file synchronization require the local server.

Effective Prompt Patterns

The most useful prompts were simple, scenario‑specific statements such as "我填入的是买入数量和金额，不填均价" or "页面分成上中下排列，Alpha、Crypto、US". These concrete cues guided the AI to produce exactly the needed adjustments.

Conclusion

The resulting tool is not a commercial product but a highly personalized, low‑profile local panel that demonstrates the value of AI‑assisted development: start with a minimal prompt, iterate based on real‑world usage, and let concrete feedback shape the final solution.