Show HN: Treyspace ─ Open Source Graph RAG on Your Excalidraw Canvas (github.com)

Treyspace is an open-source RAG (Retrieval-Augmented Generation) SDK that turns Excalidraw-style canvases into queryable knowledge graphs. It ingests canvas elements, mirrors them into a graph-vector store (Helix DB by default or an in-memory graph for development), and exposes LLM-driven analysis and streaming SSE endpoints to summarize, reason about, or augment diagrams. A hosted demo with an Excalidraw UI is available at treyspace.app. The project is MIT-licensed, development-focused (no built-in auth/rate-limits), and designed to be used either as a library or a standalone server. Technically, Treyspace provides an SDK (createHelixRagSDK, executeFullPipeline) and a façade with routes like /v1/responses, /api/ai/engine (SSE), /api/clusters and /api/mcp-bridge to proxy Helix for semantic, relational, and spatial clustering. It integrates with OpenAI (OPENAI_API_KEY required), supports Node.js >=18, and can run with Helix enabled via --enable_helix or fallback to an in-memory store for testing. Example scripts, end-to-end tests, and a clear pipeline guide let developers sync canvases, refresh clusters, and run full RAG pipelines that combine spatial structure with vector retrieval—making it immediately useful for building diagram-aware assistants, design-review tooling, and spatially grounded multimodal applications.