Sie

SIE is an open source inference engine from Superlinked that serves the models an AI agent depends on from one cluster in the customer's own cloud. Rather than paying per token to a frontier lab, a team runs open models for embeddings, reranking, extraction, document parsing and language tasks on its own infrastructure, so prompts and documents never leave that environment. The engine serves 138 open models and exposes an OpenAI compatible endpoint, so an existing agent can be pointed at it by changing one base URL. Integrations are listed for the OpenAI Agents SDK, with Hermes, LangGraph and CrewAI marked as coming soon, and partner vector stores such as Chroma, LanceDB, Qdrant and Weaviate are quoted on the site. Worked examples include self hosted product search with attribute extraction, a multimodal wine recommender with OCR, private compliance retrieval and a comparison tool for embedding models. It runs on AWS, GCP, Azure or fully air gapped, and the vendor reports SOC2 Type 2. The product is aimed at engineering teams that want lower cost, private and self hosted inference for retrieval and agent pipelines. The vendor cites comparisons such as a multilingual embedding model against a hosted API at about 50 times lower cost. Self hosting is always free under the Apache 2.0 license, with Docker, Helm and Terraform options. A managed service and an agent plugin are listed as upcoming, and free hosted capacity is offered to selected projects through an inference grant application. No paid prices are published.