Introduction¶
The core concept of DeepSeek Harness (DSH) is “everything is a plugin.” When you have Factory (Droid) subscription credits but want to call these models directly in DSH’s native environment, you need an adapter layer. dsh-factory-provider is built for this purpose. It integrates Factory models into DSH’s native LLM provider plugin system, allowing you to use your Factory API key to run inference directly in DSH.
Plugin Overview¶
- Core functionality: Connects Factory (Droid) subscription credits to the native LLM provider plugin in DeepSeek Harness (DSH).
- Maintainer: Dalizi2026
- License: MIT
- Current version: 1.3.7
Installation and Configuration¶
Use the DSH CLI to install it.
dsh plugin --profile desktop add link:<你 clone 的目录>
After installation, you must fully restart DSH for it to take effect. After restarting, go to the Factory plugin in the DSH settings page, paste the API key generated on the Factory official website, and save it.
Features¶
- Model groups: Includes five model groups: Core, Claude, GPT, Gemini, and Grok.
- Upstream fault tolerance: Supports switching in the order of API key routing, with successful lines kept per session.
- Model catalog: Includes the official Factory model table and thinking intensity levels.
- Credential management: API keys are stored locally only (permissions 0600), and are not written to settings files or logs.
- Subscription usage: The settings page displays 5-hour / weekly / monthly usage for the standard pool and Core pool.
- Token statistics: Supports cumulative input / output / cache read/write statistics by model.
Environment and Dependencies¶
- Supported DSH version: 0.2.0-rc.2 (verified).
- Node.js version: Requires Node.js >= 20.
- Plugin dependency:
@deepseek-ai/schemastery(^3.18.2).
Notes¶
- You must fully restart DSH after installing or uninstalling.
- API keys are stored locally only with permissions 0600, and are not uploaded or recorded in logs.
