DeepClaude

DeepClaude is a Python-based API middleware that chains two AI models in sequence: a reasoning model (DeepSeek R1) generates a chain-of-thought response, and a second generation model then uses that reasoning to produce the final output. The project ships three pre-configured pairings -- DeepClaude (DeepSeek R1 + Claude Sonnet 4) for coding tasks, DeepGeminiPro (DeepSeek R1 + Gemini 2.5 Pro) for content creation, and DeepGeminiFlash for lighter daily use -- all exposed through an OpenAI-compatible REST API with both streaming and non-streaming support. It connects to Claude via the Anthropic API directly or through relay providers such as OpenRouter and OneAPI. Deployment runs through a single Docker command, and a browser-based configuration UI at /config replaces manual environment-variable setup. A concrete benchmark cited in the README shows the R1-plus-Sonnet pairing achieving 64% on the aider polyglot benchmark at 14x lower cost than the previous o1 result. Developers self-hosting LLM infrastructure or building on top of OpenAI-compatible tooling are the primary audience.

What people ask about DeepClaude