deepsproxy
A local proxy that lets OpenAI-format apps and SDKs talk to DeepSeek models instead.
DeepsProxy is a local proxy server that lets any application built for OpenAI's API communicate with DeepSeek's language models instead. The README is written in Portuguese. The core problem it solves is compatibility: many AI tools, SDKs, and coding assistants are designed to talk to OpenAI's specific API format. This proxy sits in between, accepting requests in that OpenAI format and forwarding them to DeepSeek models, so you can swap providers without changing your existing application code.
It implements the standard OpenAI chat completions endpoint at a local URL, meaning you simply point your OpenAI SDK or tool at http://localhost:3000 instead of OpenAI's servers. It also includes a browser automation layer powered by Playwright, which is a headless browser control library, allowing the proxy to handle tool calls that require interacting with web pages. Login sessions with DeepSeek are saved so you do not need to authenticate repeatedly.
Someone would use this if they are building or using an AI-powered application and want to switch to DeepSeek models without rewriting their integration code, or if they want to route requests through a local proxy for privacy or cost reasons. The tech stack is TypeScript running on Node.js with the Hono web framework and Playwright for browser automation. Docker Compose is available for easy deployment.
Where it fits
- Point an existing OpenAI SDK integration at DeepSeek models without rewriting code
- Run a local chat completions endpoint for cost or privacy reasons
- Handle tool calls that need browser interaction using the built-in Playwright layer
- Deploy the proxy with Docker Compose for a repeatable local setup