AI-Hub: One-Stop Multi-Provider AI Conversation Platform Integrating Nine Model Providers and RAG Memory System

AI-Hub is an open-source project developed with Next.js and TypeScript. As a one-stop multi-provider AI conversation platform, it integrates nine model providers: OpenRouter, NVIDIA NIM, OpenAI, Anthropic, Google AI, Groq, Together AI, and Ollama, enabling a unified conversation interface. Key features include real-time sentiment analysis, multimodal input (voice + image), RAG knowledge base retrieval, long-term memory management, budget tracking, and human takeover functionality. Local key storage ensures security.

With the rapid development of large language models, different AI providers have varying API interfaces, pricing strategies, and functional features. Users need to flexibly switch models in a single interface while maintaining conversation continuity. AI-Hub addresses this pain point through a unified API interface layer, supporting seamless use of models from nine mainstream providers.

AI-Hub uses the Next.js 16 App Router frontend architecture, TypeScript 5 for type safety, Tailwind CSS 4 + shadcn/ui for the interface, Zustand for state management, and Prisma ORM for SQLite local storage operations. Keys are stored in a local database, and API calls are made on the server side to ensure sensitive information is not exposed.

It supports cloud services like OpenRouter and NVIDIA NIM as well as local Ollama deployment. Each provider is adapted for streaming responses (Server-Sent Events). The built-in model browser displays model metadata, and the key verification function automatically tests validity to avoid configuration errors.

It supports image upload (paste/drag-and-drop/file selection) and voice input (converted to text via Web Speech API), automatically detects over 15 languages and adjusts response strategies. Real-time sentiment analysis based on the AFINN dictionary marks emotions and adjusts the AI's response style to enhance the humanization of conversations.

The RAG feature supports automatic indexing of uploaded documents, retrieving relevant fragments and injecting them into prompts during conversations. Long-term memory saves key facts and injects them into system prompts across conversations to achieve a personalized experience. The two focus on document knowledge and user persistent information respectively.

It provides operations like message copying, editing, and resending. A context window indicator shows token usage, and a summary prompt is automatically triggered when there are more than 20 messages. Budget tracking allows setting monthly limits, calculating token costs in real time, and sending alerts, making it suitable for high-frequency API scenarios.

AI-Hub provides developers and users with a unified, secure, and scalable AI access layer. It is open-source under the MIT license to support extensions. For production deployment, it is recommended to encrypt API keys and configure data retention policies (30 days/90 days/1 year/permanent). In the future, such platforms will lower the threshold for AI usage and improve development efficiency.