nexus-orchestrator

Nexus Orchestrator acts as an intelligent routing layer for Large Language Models (LLMs), allowing users to mix local providers (like Ollama) with cloud APIs (OpenAI-compatible) within a single interface. It uses a lightweight router model to classify prompt intent into categories such as CODING, REASONING, CREATIVE, VISION, and SECURITY, automatically dispatching the request to the designated model pool for that category. Key capabilities include: - Hybrid Orchestration: Seamlessly switch between local and cloud models per category. - Tool Integration: Supports Model Context Protocol (MCP) servers and integrates with self-hosted SearXNG for web search and URL fetching via tool calling. - Multi-User Management: Includes per-user provider configurations, category mappings, and project management with an admin dashboard. - Advanced LLM Support: Handles vision and document uploads, and provides a dedicated "Thinking" toggle to display reasoning chains (e.g., for DeepSeek R1) in a collapsible UI section. - Security & Privacy: Designed for privacy-first operation with 100% local routing options, featuring built-in CORS, rate limiting, and SSRF protection for URL fetching.