A unified LLM API client for Unity - Call OpenAI, Anthropic Claude, Google Gemini, Azure OpenAI, AWS Bedrock, and local LLMs (gguf models) with a single interface.ThinLLM is a unified LLM API client that enables seamless integration of multiple AI providers into your Unity projects with a single, consistent interface. It lets developers call OpenAI, Anthropic Claude, Google Gemini, Azure OpenAI, AWS Bedrock, and local LLMs (GGUF models) through one simple Completion() method, with no code changes required when switching providers.Core Features:・Unified API Interface: Single Completion("{provider}/{model-name}", messages) method works across all 6 supported LLM providers・Real-time Streaming: Token-by-token response streaming for responsive UI updates・Local LLM Support: Run GGUF models offline via LLamaSharp - no API costs, no internet required・Built-in Model Downloader: Download GGUF models directly from URL within Unity Editor・Zero External Dependencies: All required libraries are pre-bundled in the packageThinLLM is perfect for building intelligent NPCs, dynamic dialogue systems, AI assistants, or any application that needs flexible LLM access. The simple model format ("{provider}/{model-name}") makes it trivial to experiment with different providers and find the best fit for your project.A demo scene with a working chat interface is included so you can start testing immediately out of the box.Features:・Text Generation via Multiple LLM Providers (OpenAI, Anthropic, Gemini, Azure, Bedrock)・On-Device LLM Inference via LLamaSharp (GGUF models)・Streaming Response Support・Environment File-based API Key Configuration・Editor Window for Model DownloadsImportant/Additional Notes:The ThinLLM package integrates and extends several open-source projects.These projects remain under their original licenses.・LLamaSharp・SmolLM2 (bundled demo model)LLM was used to draft initial code implementations and documentation.All AI-generated content was carefully reviewed, tested, and modified by a human developer to ensure correctness, quality, and production readiness.Final implementation decisions were made by humans.
