Governed AI gateway · production
One scoped key. Every model, metered.
SansaAI is the single policy-enforced entry point for model inference, embeddings, isolated browser work, and video for Sansa Vision products. Consumers authenticate with a dedicated workload key and are authorized and metered per call.
Text + reasoning
sansa/auto · sansa/glm-4.7-flash · sansa/glm-5.3-flash · sansa/llama-3.3-70b
OpenAI-compatible chat completions and responses with strict JSON support.
Embeddings
sansa/qwen3-embedding · 32 dimensions
Vectorize-compatible vectors for search and retrieval workloads.
Video
sansa/minimax-h3-max · 480P / 768P · 5–15s
Official MiniMax H3 Max generation with async job receipts.
Browser isolation
OxyBlink worker-native extraction with Browser Run fallback. No direct provider calls from consumers.
Metering
LimitFuse policy, budgets, and per-operation receipts. Current mode: real.
Integrate
POST /v1/responses · POST /v1/embeddings · /mcp
Discover through the Sansa Auth catalog, authenticate through Auth or a workload bearer, stay inside allowed routes.