The Universal Alliance
Freedom from vendor lock-in.
The Anahata ASI Platform is designed to adapt to any intelligence source. Our Universal Alliance adapters allow you to connect to Groq, Local Ollama, DeepSeek, vLLM, and any other platform that respects the standard API contracts.
OpenAI Compatible (The Swiss Army Knife)
The OpenAiCompatibleProvider is the ultimate fallback and universal connector. By simply dropping in a Base URL and an API Key, you can interface with hundreds of open-source and proprietary models instantly. It features automatic reasoning extraction for models that wrap their thoughts in XML-style tags.
NVIDIA NIM
OpenAI-compatible adapter connecting to NVIDIA NIM microservices (https://integrate.api.nvidia.com/v1). Provides access to 100+ frontier models (DeepSeek, Nemotron, Llama, Cosmos, GLM, Qwen, GPT-OSS).
NovaRouteAI
An OpenAI-compatible provider adapter offering seamless access to official Chinese models (Kimi, DeepSeek, Qwen, GLM, etc.) with pre-configured endpoints and affiliate onboarding. Learn more →
Mistral AI
Dedicated OpenAI-compatible adapter for Mistral AI, automatically discovering model capabilities (FIM, vision, audio, reasoning, and OCR) from Mistral's native /v1/models JSON endpoint.
Hugging Face (The Deep Inspector)
Our Hugging Face adapter doesn't just blindly send prompts. It runs a parallel "Deep Inspection" using the container's background executor to fetch config.json and tokenizer_config.json directly from the HF Hub. This allows it to dynamically learn the model's architecture, accurate token limits, and whether it uses <think> tags or reasoning_content fields!
Modal (Serverless GLM)
A sleek, pre-configured adapter that instantly taps into Modal's serverless GLM endpoints. It automatically parses GLM's native function calling mechanics and its specific reasoning_content fields for pristine UI rendering.
MiniMax (Anthropic-Compatible)
Demonstrating our extreme architectural flexibility, we mapped MiniMax's high-speed frontier models through our Anthropic-compatible adapter rather than OpenAI's. This provides native access to their lightning-fast agentic workflows while reusing our bulletproof Claude SSE streaming logic.
