arthrod/Reasoner4All

Reasoner4All

Reasoner4All is a simple demo that allows you to supercharge your favourite LLMs by integrating advanced reasoning capabilities from the latest open-source thinking models. Whether you're working with OpenAI, Mistral, Groq, or Gemini, this framework seamlessly enhances their decision-making with an additional reasoning layer. Leverage models like DeepSeek R1 Distill LLaMA-70B to improve your AI's problem-solving skills and contextual understanding!

Useful Links 🔗

• —AiCore Repo my custom wrapper around multiple providers which uniforms streaming, json outputs and reasoning integration.

Features

Supports Multiple LLM Providers

• —OpenAI
• —Mistral
• —Groq (including DeepSeek R1 Distill LLaMA-70B for reasoning)
• —Gemini
• —and more!

Reasoner Integration

• —Uses an additional (reasoning) LLM to enhance reasoning capabilities
• —Support for Deepseek R1 Distill LLaMA-70B (Groq Hosted) and Deepseek R1 (Nvidia Hosted)

Conversation History

• —For context reasons the converstation history up to latest 4096 tokens is preserved and passed to the llms as context

User Interface

Settings

• —Provider and Model selection
• —Reasoner Option and Model selection
• —Support for System Propmpt and Reasoner System prompt specification

Profiles

• —Reasoner4All uses a series of pre-set API Keys for demo purposes
• —OpenAi allows you to connect to your models with support for api key if required don't worry I am not logging your keys anwyehre but you can check the src code of this space to be sure :)

Chat

• —Reasoning steps appear inside a step which can be expanded to visualize the process