deepreasoning

DeepReasoning is a Rust-based inference server and chat UI that chains DeepSeek R1 and Claude together in a single API call, using R1's chain-of-thought reasoning trace as a pre-processing step before handing off to Claude 3.5 Sonnet for final response generation. The server, built with Rust 1.75+, accepts requests at a local endpoint and authenticates separately against DeepSeek and Anthropic APIs via per-request headers, meaning users supply both API keys directly rather than routing credentials through a third party. It supports both standard and streaming responses, with configuration options for custom system prompts and per-model header and body overrides via a JSON request body. The project is self-hostable using a simple `config.toml` file specifying host, port, and pricing settings. Developers who want structured reasoning from R1 paired with Claude's code generation and conversational output in one request, without stitching together two separate API clients, are the primary audience.

What people ask about deepreasoning