[{"data":1,"prerenderedAt":4},["ShallowReactive",2],{"readme:sie":3},"\u003Cdiv align=\"center\">\u003Cpicture>\n  \u003Csource srcset=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F65dce6831bf9f730421e2915\u002F66ef0317ed8616151ee1d451_superlinked_logo_white.png\" media=\"(prefers-color-scheme: dark)\">\u003C\u002Fsource>\n  \u003Cimg width=\"320\" src=\"https:\u002F\u002Fcdn.prod.website-files.com\u002F65dce6831bf9f730421e2915\u002F65dce6831bf9f730421e2929_superlinked_logo.svg\" alt=\"Superlinked logo\" \u002F>\n\u003C\u002Fpicture>\u003Ch1>SIE: Superlinked Inference Engine\u003C\u002Fh1>\u003Cp>\u003Cstrong>Self-hosted inference for agents. Every open model your agents call, served from one cluster in your cloud.\u003C\u002Fstrong>\u003C\u002Fp>\u003Cp>\n  \u003Ca href=\"https:\u002F\u002Fsuperlinked.com\u002Fdocs\u002F\" rel=\"nofollow ugc noopener\">Docs\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fsuperlinked.com\u002Fdocs\u002Fquickstart\u002F\" rel=\"nofollow ugc noopener\">Quickstart\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fsuperlinked.com\u002Fdocs\u002Freference\u002Fapi\u002F\" rel=\"nofollow ugc noopener\">API Reference\u003C\u002Fa> |\n  \u003Ca href=\"https:\u002F\u002Fsuperlinked.com\u002Fmodels\" rel=\"nofollow ugc noopener\">Models\u003C\u002Fa>\n\u003C\u002Fp>\u003Cp>\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsuperlinked\u002Fsie\u002Fblob\u002FHEAD\u002FLICENSE\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002Flicense-Apache%202.0-blue?style=flat-square\" alt=\"License\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fpypi.org\u002Fproject\u002Fsie-sdk\u002F\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fpypi\u002Fv\u002Fsie-sdk?style=flat-square\" alt=\"PyPI\" \u002F>\u003C\u002Fa>\n\u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsuperlinked\u002Fsie\u002Fstargazers\" rel=\"nofollow ugc noopener\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fgithub\u002Fstars\u002Fsuperlinked\u002Fsie?style=flat-square\" alt=\"GitHub stars\" \u002F>\u003C\u002Fa>\u003C\u002Fp>\n\u003C\u002Fdiv>\u003Ch2>About\u003C\u002Fh2>\n\u003Cp>SIE is an open-source inference engine that runs the models behind every agent task through one API: search and retrieval, document-to-markdown conversion, structured output, content safety, and the agent loop itself. It replaces the patchwork of a separate model server per task with one system that serves 100+ models, loading each on demand.\u003C\u002Fp>\n\u003Cul>\n\u003Cli>OpenAI-compatible API for drop-in migration: \u003Ccode>\u002Fv1\u002Fembeddings\u003C\u002Fcode>, \u003Ccode>\u002Fv1\u002Fchat\u002Fcompletions\u003C\u002Fcode>, \u003Ccode>\u002Fv1\u002Fcompletions\u003C\u002Fcode>, \u003Ccode>\u002Fv1\u002Fresponses\u003C\u002Fcode>\u003C\u002Fli>\n\u003Cli>Pre-configured model catalog: Stella, SPLADE, Qwen3, GLiNER, SigLIP, and more; embedding and retrieval models benchmarked on MTEB\u003C\u002Fli>\n\u003Cli>Serves multiple models simultaneously with on-demand loading and LRU eviction\u003C\u002Fli>\n\u003Cli>Ships the full production stack: load-balancing gateway, KEDA autoscaling, Grafana dashboards, Terraform for GKE, EKS, and AKS\u003C\u002Fli>\n\u003Cli>Integrates with LangChain, LlamaIndex, Haystack, DSPy, CrewAI, Chroma, Qdrant, Weaviate, and LanceDB\u003C\u002Fli>\n\u003C\u002Ful>\n\u003Ch2>Tasks\u003C\u002Fh2>\n\u003Cp>One SIE cluster runs the inference behind a whole agent. Each task is a handful of swappable models; browse \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsuperlinked\u002Fsie\u002Ftree\u002Fmain\u002Fpackages\u002Fsie_server\u002Fmodels\" rel=\"nofollow ugc noopener\">\u003Ccode>packages\u002Fsie_server\u002Fmodels\u002F\u003C\u002Fcode>\u003C\u002Fa> for the full set.\u003C\u002Fp>\n\u003Ctable>\n\u003Cthead>\n\u003Ctr>\n\u003Cth>Task\u003C\u002Fth>\n\u003Cth>What it does\u003C\u002Fth>\n\u003Cth>Models\u003C\u002Fth>\n\u003C\u002Ftr>\n\u003C\u002Fthead>\n\u003Ctbody>\u003Ctr>\n\u003Ctd>\u003Cstrong>Search\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Embed, match, and rerank to retrieve the right context.\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>bge-m3\u003C\u002Fcode>, \u003Ccode>splade-v3\u003C\u002Fcode>, \u003Ccode>colbertv2\u003C\u002Fcode>, \u003Ccode>qwen3-reranker\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Document to markdown\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>PDFs, Office files, and scans become clean markdown.\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>glm-ocr\u003C\u002Fcode>, \u003Ccode>mineru\u003C\u002Fcode>, \u003Ccode>paddleocr-vl\u003C\u002Fcode>, \u003Ccode>docling\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Structured output\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Schema-valid JSON, extracted or generated.\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>gliner2\u003C\u002Fcode>, \u003Ccode>nuner-zero\u003C\u002Fcode>, \u003Ccode>qwen3.6-27b\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Guard content\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>A safety verdict with a probability you threshold.\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>granite-guardian-2b\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003Ctr>\n\u003Ctd>\u003Cstrong>Run the agent loop\u003C\u002Fstrong>\u003C\u002Ftd>\n\u003Ctd>Plan steps and call tools with an open LLM, streaming included.\u003C\u002Ftd>\n\u003Ctd>\u003Ccode>qwen3.6-27b\u003C\u002Fcode>\u003C\u002Ftd>\n\u003C\u002Ftr>\n\u003C\u002Ftbody>\u003C\u002Ftable>\n\u003Ch2>Quickstart\u003C\u002Fh2>\n\u003Cp>Prefer a notebook? \u003Ca href=\"https:\u002F\u002Fgithub.com\u002Fsuperlinked\u002Fsie\u002Fblob\u002FHEAD\u002Fexamples\u002Fquickstart.ipynb\" rel=\"nofollow ugc noopener\">\u003Ccode>examples\u002Fquickstart.ipynb\u003C\u002Fcode>\u003C\u002Fa> runs this same flow, on your machine or a free Colab GPU.\u003C\u002Fp>\n\u003Cp>\u003Cstrong>1. Start the server\u003C\u002Fstrong>\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-bash\"># macOS (Apple Silicon) or Linux, native (requires Python 3.12)\npip install \"sie-server[local]\" && sie-server serve\n\n# Linux, NVIDIA GPU\ndocker run --gpus all -p 8080:8080 \\\n  -v sie-hf-cache:\u002Fapp\u002F.cache\u002Fhuggingface \\\n  ghcr.io\u002Fsuperlinked\u002Fsie-server:latest-cuda12-default\n\n# Linux, CPU\ndocker run -p 8080:8080 \\\n  -v sie-hf-cache:\u002Fapp\u002F.cache\u002Fhuggingface \\\n  ghcr.io\u002Fsuperlinked\u002Fsie-server:latest-cpu-default\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cpre>\u003Ccode class=\"language-bash\"># in a second terminal\ncurl http:\u002F\u002Flocalhost:8080\u002Freadyz   # expect: ok\n\u003C\u002Fcode>\u003C\u002Fpre>\n\u003Cp>The server speaks the OpenAI API out of the box, embeddings and generation alike (the cluster gateway serves \u003Ccode>\u002Fv1\u002Fchat\u002Fcompletions\u003C\u002Fcode>, \u003Ccode>\u002Fv1\u002Fcompletions\u003C\u002Fcode>, and \u003Ccode>\u002Fv1\u002Fresponses\u003C\u002Fcode>). Your first call needs nothing but curl:\u003C\u002Fp>\n\u003Cpre>\u003Ccode class=\"language-bash\">curl http:\u002F\u002Flocalhost:8080\u002Fv1\u002Fembeddings \\\n  -H 'Content-Type: application\u002Fjson' \\\n  -d '{\"model\": \"sentence-transformers\u002Fal\n\u003C\u002Fcode>\u003C\u002Fpre>\n",1784240408053]