winfunc/deepreasoning
Deepreasoning
DeepReasoning is a self-hosted AI gateway and chat interface that provides an LLM inference API. It functions as a bridge that merges reasoning traces from DeepSeek R1 with the generative capabilities of Claude models to facilitate complex problem solving.
The system is delivered as a dockerized application, allowing for deployment on private infrastructure. This architecture enables private LLM inference and secure local management of API keys and authentication tokens on user-controlled hardware.
The project covers multi-model orchestration by combining chain-of-thought reasoning and generative outputs into a single unified stream. It utilizes a stateless API gateway and server-sent events to deliver incremental model responses to the client interface.
Features
- Reasoning-Generative Bridges - Functions as a bridge merging DeepSeek R1 reasoning traces with Anthropic Claude models for complex problem solving.
- Self-Hosted Chat Interfaces - Provides a deployable, private web interface for interacting with various LLM backends via Docker.
- Complex Problem Solving - Solves complex tasks by integrating chain-of-thought reasoning with generative models for step-by-step logical processing.
- Chat and API Access - Offers both a web-based chat interface and a programmatic API for streaming combined reasoning and generative outputs.
- Multi-Model Orchestration - Coordinates the reasoning capabilities of DeepSeek R1 and the generative strengths of Claude models within a single workflow.
- Private LLM Inference - Provides a framework for running LLM APIs and interfaces on private hardware to maintain total environment control.
- Reasoning Step Outputs - Surfaces detailed intermediate reasoning steps from models alongside the final generative response in a single stream.
- Private Infrastructure Hosting - Allows deployment of the inference wrapper and chat interface on user-controlled hardware for full environment authority.
- Self-Hosted AI Infrastructure - Serves as a private gateway for managing LLM API keys and hosting inference interfaces on owned hardware.
- Self-Hosted API Deployments - Provides a configuration for deploying the API service on private, user-owned infrastructure.
- Self-Hosted Inference Servers - Enables running the AI inference engine on self-managed servers for complete control over the hardware environment.
- API Key Management - Provides secure local storage and management of LLM API keys and authentication tokens on private infrastructure.
- Model - Merges separate chain-of-thought reasoning traces and final generation streams into a unified response for the client.
- Model Response Streaming - Implements incremental delivery of AI model outputs to the user interface for improved responsiveness during inference.
- Dockerized LLM Chat Deployments - Ships as a fully containerized LLM chat application with an integrated API for easy deployment.
- Containerized Deployments - Provides standardized packaging via containers to ensure consistent application execution across diverse private infrastructure.
- Token Streaming - Utilizes real-time token streaming to deliver incremental model updates to the user interface.
- Local Secret Fallbacks - Stores authentication tokens and API keys in local encrypted files on private infrastructure to prevent cloud exposure.
- Container-Based Isolation - Uses container images to encapsulate the application and its dependencies for consistent execution across private hardware.
- Stateless Architectures - Implements a stateless architecture to route requests between the client and multiple LLM providers without server-side sessions.
