HomeProjectsArticlesPK BoardAbout
Explore Projects

DeepReasoning

Stale
GitHub Rust MIT

Description

A high-performance LLM inference API and Chat UI that integrates DeepSeek R1's chain-of-thought reasoning traces with Anthropic Claude models.

Key Features

  • Combines DeepSeek R1 chain-of-thought reasoning with Claude's code generation in a single stream
  • High-performance Rust API delivering zero-latency dual-model inference
  • Bring Your Own Keys architecture for complete control over API keys and data privacy
  • Streaming and non-streaming modes with extensive configuration via request body
  • Self-hostable on your own infrastructure with Docker or source compilation
  • Open-source MIT licensed codebase with active community contributions

Use Cases

💡 Complex problem solving requiring deep reasoning followed by precise code generation
💡 Building AI-powered development tools that need both analytical and creative capabilities
💡 Creating conversational agents with metacognitive reasoning and self-correction
💡 Developing applications where data privacy and local inference are critical requirements
💡 Research projects exploring the synergy between reasoning and generation models

Strengths & Limitations

✅ Strengths

  • • High community interest (5.4k stars)
  • • Permissive open-source license (MIT)

⚠️ Limitations

  • • No updates in over 11 months

Tags

rust llm api framework

Categories

⚡ Agent Tools

Quick Start

git clone https://github.com/winfunc/deepreasoning.git && cd deepreasoning && cargo build --release
Visit GitHub

Related Projects

mistral.rs

7.6k · Rust
Active A

Fast, flexible LLM inference engine built in Rust — supports multiple model architectures and quantization schemes for high-performance local LLM deployment.

rustllmtools +1
  • · Zero-Config Model Loading — auto-detect architecture, quantization format, and chat template from any Hugging Face model
  • · True Multimodality — text, vision, video, audio input, speech generation, image generation, and embeddings in one engine
  • · Smart Quantization — UQFF prebuilt quantizations and ISQ with automatic best-format selection at each level
View full details → View on GitHub

vLLM

90.8k · Python
Active A

A high-throughput and memory-efficient inference and serving engine for LLMs, featuring PagedAttention, continuous batching, and optimized KV cache management for production deployments.

llmpythonframework +2
  • · PagedAttention memory management - Efficiently manages attention KV cache via paging, dramatically reducing memory usage
  • · Continuous batching and chunked prefill - Supports continuous batching, chunked prefill, and prefix caching for high-throughput inference
  • · Multi-quantization support - Covers FP8, INT8, INT4, GPTQ/AWQ, GGUF, compressed-tensors, and more quantization formats
View full details → View on GitHub

Vision Agents

8.1k · Python
Active A+

Open Vision Agents by Stream. Build voice and vision agents quickly with any model or video provider, using Stream's edge network for ultra-low latency realtime interactions.

voiceagentpython +3
  • · Real-time video AI with YOLO, Roboflow integration combined with Gemini/OpenAI for live visual understanding
  • · Ultra-low latency audio/video under 30ms via Stream's edge network with 500ms join time
  • · Native SDK methods from OpenAI, Gemini, and Claude for always-latest LLM capabilities
View full details → View on GitHub

MiroThinker

8.4k · Python
Normal A

A deep research agent framework optimized for complex research and prediction tasks, with MiroThinker-1.7 and MiroThinker-H1 models achieving 74.0 and 88.2 on BrowseComp benchmark, supporting multi-step reasoning and information retrieval.

pythonagentllm +3
  • · Deep research agent framework optimized for complex research and prediction tasks
  • · MiroThinker-1.7 achieves 74.0 on BrowseComp, H1 model reaches 88.2
  • · Supports 256K context window and up to 300 tool calls per task
View full details → View on GitHub