Generate text embeddings using ONNX Runtime C API, producing .bfvec files for semantic similarity and inference tasks.
Apache 2.0
457,293 tools. Updated 2026-08-14 12:47
"ONNX" matching MCP tools:
- Inspect ONNX model layers and extract specific components with layer-aware precision.Apache 2.0
- Import a public Hugging Face model's pre-built ONNX file into EdgeGate, registering it as an Artifact. Returns an artifact_id for edgegate_create_pipeline, with optional polling for completion.MIT
- Submit a genie compile job for a target device chipset, choosing one of three input formats: HuggingFace repo, ONNX artifacts, or precompiled bundle. Returns a job ID to poll for progress.MIT
- Execute a complete clustering pipeline to identify high-CPU SQL queries, extract performance metrics, and group similar queries for optimization prioritization.MIT
- Run an ONNX benchmark on any connected edge device (Jetson, Snapdragon, gateway) by name or ID, and get latency/memory results within seconds. Requires admin access.MIT
Matching MCP Servers
- AlicenseAqualityAmaintenanceLocal RAG system for Claude Code with hybrid search (semantic + BM25), cross-encoder reranking, markdown-aware chunking, and 12 MCP tools. Zero external servers, pure ONNX in-process.13252MIT
- AlicenseAqualityAmaintenanceLocal semantic recall MCP server for citation-integrity workflows. Enables semantic search over a text corpus, returning file paths and page numbers for verification, with 100% local processing using ONNX embeddings and LanceDB.21MIT
- Predict ONNX operator coverage on Qualcomm Hexagon NPU before spending AI Hub credits, returning compute-weighted coverage percentage, fallback ops, and fix recommendations.MIT
- Store and persist conversation memories for AI agents. Supports episodic, semantic, and procedural memory types with adjustable importance weights.MIT
- Generate a depth map from any 2D image or video TOP using Depth Anything v2, removing the need for Kinect or RealSense. Output depth TOP feeds displacement, pop-field, or silhouette tools.MIT
- Turn object/person detection into TouchDesigner control channels via WebSocket or CPU ONNX inference, outputting presence, count, and normalized bounding boxes as a Null CHOP ready for binding.MIT
- Builds a TouchDesigner YOLO tracking scaffold with input source, detection DAT, stable tracks CHOP, annotated TOP, and configurable backend (WebSocket, ONNX, NDI, file watch).MIT
- Extract text from images using offline OCR. Reads text from screenshots, documents, and photos via local file paths or base64 data.MIT
- Turn object/person detection into TouchDesigner control channels via WebSocket or ONNX CPU inference, providing presence, count, and bounding boxes as CHOP channels.MIT
- Convert any 2D image or video into a depth map for TouchDesigner using Depth Anything v2, eliminating the need for Kinect or RealSense hardware.MIT
- Executes the complete SQL query clustering pipeline to identify high-CPU queries for optimization. Extracts, embeds, and clusters queries to surface optimization candidates.MIT
- Sets up a TouchDesigner scaffold for YOLO-style object tracking, creating source input, detection DAT, stable track CHOP, and annotated TOP. Routes live detections via WebSocket, ONNX, NDI, or file-watch backends.MIT
- Create a semantic knowledge base layer from PDFs. Index files into a named database and optionally force re-index of already indexed files.AGPL 3.0
- Index PDFs into a named knowledge database layer using local embeddings, creating a semantic knowledge base for persistent research memory. Supports rebuilding existing databases.AGPL 3.0
- Pre-download the embedding model and optional reranker to prevent sync delays. Run before initializing a project.
- Execute a complete SQL query clustering pipeline to identify high CPU usage queries for optimization analysis, extracting logs, calculating performance metrics, generating embeddings, and grouping similar queries.MIT