Introduction¶
Developers using DeepSeek Harness usually encounter three types of problems: which model’s context length is sufficient, how much price difference exists between the same model on different providers, and whether there are ready-made plugins in the community to solve a specific task. The first two questions require regularly checking pricing pages, while the third relies on search engines or browsing directory sites. moldkey/dsh-deepseekbase integrates these queries directly into Harness’s tool layer, allowing developers to obtain price, spec, and plugin information with sources without leaving the session.
Below is an introduction to the positioning, features, installation method, and scope of applicability of this plugin.
What is it¶
moldkey/dsh-deepseekbase is a DSH plugin categorized under Model Inference. It is released by moldkey, the maintainer of the third-party indexing site deepseekbase.com, and provides four tools in Harness to answer questions related to Price, Plugin Search, Model Specs, and Usage Trends. The plugin itself does not store data; it makes HTTPS requests to deepseekbase.com to retrieve data at runtime.
deepseekbase.com is a third-party maintained DeepSeek ecosystem index and is not the official DeepSeek website. The README explicitly states that it is unrelated to DeepSeek, indexing a total of 283 pages and 21 open datasets, covering pricing, providers, derived models, plugins, etc.
Core Features¶
The plugin registers four tools:
| Tool | Question Answered |
|---|---|
deepseekbase_price |
What is the pricing of DeepSeek and which provider is the cheapest? |
deepseekbase_plugin |
What plugin is needed to complete task X, what is its popularity, and what are the risks? |
deepseekbase_specs |
What is the model scale, context length, and can it fit? |
deepseekbase_trend |
DeepSeek’s actual usage trends. |
The price query example given in the README:
You: compare DeepSeek V4.1 Flash prices across providers
Tool returns:
Official: off-peak in $0.15 / out $0.60 per 1M tokens Cheapest third party: Particle AI — in $0.025 / out $0.25
There is an approximately 6x price difference between the same model on different providers; obtaining this information quickly by manually checking pricing pages is difficult.
The plugin only makes HTTPS requests to deepseekbase.com at runtime; it does not read/write local files, credentials, or environment variables, nor does it execute shell commands. All implementations are concentrated in lib/index.js with no runtime dependencies.
The backend has two data endpoints:
| Endpoint | Purpose |
|---|---|
POST https://deepseekbase.com/mcp |
prices, plugins, specs |
GET https://deepseekbase.com/data/openrouter-tokens.json |
token usage trends |
Installation and Activation¶
Installation via GitHub, with the command written in the README:
dsh plugin --profile web add github:moldkey/dsh-deepseekbase
After installation, restart Harness, and the four tools will be available.
The dsh field in package.json specifies the packaging patch, and engines.dsh requires >=0.2.0-rc.2. This means installation may fail on Harness versions lower than this.
The README also provides an installation command in npm format but notes “once published”, meaning the npm package may not have been published yet:
dsh plugin --profile web add @liangjiehappy/dsh-deepseekbase
The actual package name is @liangjiehappy/dsh-deepseekbase, current version 0.1.1, license MIT, but the information does not provide a star count.
Suitable Scenarios and Notes¶
Suitable scenarios:
- Need to quickly compare DeepSeek model prices across different providers within a session without opening a webpage specifically for checking.
- Need to determine if a specific model can fit a current task based on context length.
- Want to search for DSH plugins in the community that do similar things.
Points to note:
- The plugin runs with the permissions of the current dsh process. Before installation, you should read the
lib/index.jssource code and confirm that the MIT license is acceptable. - Data such as prices and specs come from third-party sites. While they indicate data sources and verification dates, discrepancies between source site data and actual situations may still occur.
engines.dshrequires>=0.2.0-rc.2; users of older Harness versions should confirm the version first.- deepseekbase.com is a third-party index and is not the official DeepSeek website or Harness official app store. It is recommended to verify the content yourself before installing third-party plugins.
After installation, the plugin will initiate outbound HTTPS requests to deepseekbase.com. If running in an isolated network environment, you need to confirm that the domain is reachable first.
Conclusion¶
This plugin solves the narrow but frequent problem of “querying DeepSeek ecosystem information within a Harness session”. The four tools cover price, plugins, specs, and usage, respectively, with data coming from a third-party indexing site. The value lies not in a single feature, but in folding external queries into tool calls to reduce switching costs. If you frequently compare provider prices or need to select models based on context length, it is worth installing and trying.
Directory page: https://www.skillhub.cn/plugins/moldkey/dsh-deepseekbase
GitHub: https://github.com/moldkey/dsh-deepseekbase
