dsh-openai-gateway

Integrating a DeepSeek Harness Agent into existing OpenAI-based workflows can be challenging, as the native Web UI is not designed for programmatic access. The dsh-openai-gateway solves this by exposing your DeepSeek Harness as a standard OpenAI-compatible API server. This allows any OpenAI SDK client—be it IDE plugins, automation scripts, or tools like Raycast—to seamlessly drive your Agent with just a base URL and API key.

At its heart, the plugin maps each API request to a real Agent session complete with tools and a workspace, rather than acting as a mere LLM proxy. It supports both streaming and non-streaming responses and offers flexible session modes: a stateless default for quick, isolated calls, and sticky sessions for maintaining context across requests, ideal for complex, multi-turn tasks. Authentication and listening are handled independently, ensuring secure separation from the dsh Web UI and allowing you to expose endpoints as needed.

This gateway is perfect for AI application developers, research engineers, or tool enthusiasts who want to infuse their workflows with intelligent Agent capabilities. It caters especially to those familiar with the OpenAI ecosystem who wish to leverage the unique toolchain and session management of DeepSeek Harness. With minimal configuration, you can bridge a powerful local Agent to the global API standard, unlocking broader integration possibilities.