rust-embedding

What it does

Graphs and embeddings: A Rust-based CLI tool that creates or imports knowledge graphs, generates or ingests text embeddings, and manipulates entities and relationships. It supports formats JSON-LD, RDF/Turtle, and CSV for import/export, and can query and visualize the graph. It can generate embeddings from text using embedded NLP models via Rust backends (Rust-BERT or ONNX Runtime), and can ingest precomputed vectors from CSV or JSON formats. It provides commands like embed, import, export, add-entity, add-rel, update, delete, query, and visualize. It reads model weights on first use and writes embedding outputs to stdout or a file. It auto-detects embedding file formats on import and supports batch embedding. It mentions workflow steps from installation to building and using the tool, including example commands for embedding, importing data, and updating the graph.

How it works

• Install via cargo and ensure Rust environment; run cargo install graphembed-cli and use graphembed help.
• Use subcommands for functionality: embed to generate embeddings, import/ingest-vec for vectors, add-entity and add-rel to build the graph, update and delete for CRUD, query for pattern or SPARQL queries, and visualize to output Graphviz DOT.
• Embedding generation uses a chosen model via -m/--model (example given: all-MiniLM-L6-v2) with backends Rust-BERT or onnx depending on features, and outputs a vector of floats to stdout or a file.
• Import supports JSON-LD, RDF/Turtle, and CSV; auto-detects by file extension unless overridden with --format.
• Ingested embeddings attach to entities as properties; entities and relationships are identified by labels or IDs, with possible URI mapping for RDF contexts.
• Visualization outputs a DOT format for Graphviz, and can focus on subgraphs.
• Workflow shows steps from setup to embedding, entity relationship management, and querying.

When to use it

• When you need a CLI to generate and manage embeddings and a knowledge graph in one tool.
• When you require support for multiple graph formats (JSON-LD, Turtle, CSV) and basic visualization.
• When batch embedding or ingesting precomputed vectors is needed.

What it can touch

• Embedding generation uses embedding models via Rust backends (Rust-BERT or ONNX Runtime) chosen by CLI flags or features.
• Import/Export handles JSON-LD (.jsonld), RDF/Turtle (.ttl or .rdf), and CSV (.csv) formats.
• Operations affecting graph entities and relationships through commands like add-entity, add-rel, update, delete.
• Embedding data is attached to entities as properties within the graph.

Caveats

• Model weights download occurs on first request if not present; an internet connection is required for downloads.
• SPARQL support may depend on compile-time feature flags; may not be available in all builds.
• Large graphs may consume significant memory since the tool primarily holds data in memory.
• Visualization relies on Graphviz or DOT output; external tools may be needed to render images.
• The skill is MIT-licensed.