Introduction

When processing financial data in DeepSeek Harness (DSH), a common pain point is that the model needs to obtain market data, historical prices, news, and status information, but web content is often unstable, and parsing pages can easily introduce noise. DSH’s plugin mechanism allows mounting external capabilities to the host side, and dsh-finance-db is a plugin oriented towards reading market data.

It addresses a specific problem: when you ask about a stock, it allows DSH to call the data tool directly, rather than letting the model browse the web or piece together results on its own.

Below is an introduction to its positioning, capabilities, installation methods, and typical usage.

What is it

dsh-finance-db is a read-only market data plugin for DeepSeek Harness.

  • Repository path: zhiyaoli0221/dsh-finance-db
  • License: MIT
  • Source code repository: https://github.com/zhiyaoli0221/dsh-finance-db
  • Design goals: Support read-only analysis workflows; does not provide trading execution, fund movement, broker credential storage, or account access

Core Features

The plugin provides a set of explicit tool names for precise specification of the target to be called in the prompt.

Tool Purpose
finance_quote Get current price, change, volume, PE, and timestamp
finance_ohlcv Get historical open/high/low/close/volume data
finance_fundamentals Get valuation fields and provider metadata associated with the quote
finance_news Get recent financial headlines from NewsNow
finance_search_stocks Find ticker by code or name
finance_market_overview Get major indices and intraday changes
finance_health Check provider availability
finance_sentiment Sentiment entry point; requires mounting a local or agent provider

Installation and Enablement

Install to the DSH web profile using npm:

npx @deepseek-ai/dsh plugin --profile web add dsh-finance-db

This command adds dsh-finance-db to the DSH web profile. Once installed, you can invoke the corresponding capabilities in DSH by the tool name.

Typical Usage

The following examples are from verified use cases and are suitable as prompt templates.

Query Current Quote

Query only the current quote and require no web browsing:

Get the current quote for 600519. Use finance_quote only and do not browse the web.

This restricts the model to calling finance_quote, avoiding the introduction of additional web information.

Query Historical Range

Query 1-year OHLCV data and require a summary of the price range:

Call finance_ohlcv for AAPL over 1 year and summarize the price range.

Query News Headlines

Query recent news and list the five newest headlines:

Call finance_news for 600519 and list the five newest headlines.

Check Data Service Status

Check provider availability before using market data:

Run finance_health before using market data.

Applicable Scenarios and Notes

dsh-finance-db is suitable for read-only analysis, information organization, data verification, and automated workflows. It does not provide broker integration, execute trades, store broker credentials, move funds, or modify portfolios.

Notes to be aware of before use:

  • The plugin runs with the permissions of the current dsh process; check the source code and MIT license before installing.
  • Optional web search requires setting JINA_API_KEY.
  • finance_sentiment requires mounting a local or agent provider.
  • Requests are cached in memory, with distinct TTLs per capability.
  • If diagnosing connectivity or provider availability, use finance_health.
  • This plugin provides data capabilities, not investment, tax, accounting, or legal advice.

Conclusion

The value of dsh-finance-db lies in converging market data reading into explicit tool calls, reducing the uncertainty caused by the model parsing web pages on its own. The source code repository is:

https://github.com/zhiyaoli0221/dsh-finance-db