You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
ONNX implementation of the BGE-M3 multilingual embedding model and tokenizer with native C#, Java, and Python implementations. Generates all three embedding types: dense, sparse, and ColBERT vectors.
A demonstration of hybrid search with reranking using Qdrant and BGE-M3 model. A showcase of dense and sparse retrieval combined with ColBERT reranking for optimal search results
A local-first AI reading application. Loss-bounded semantic compression with deterministic source traceability. Ingest PDFs/EPUBs and read faster, without losing context.
Does a TypeSafe Jev rerank beat embedding search? Graded relevance eval (9,831 pairs, 164 zh/en queries) over the Agent Skills Hub catalog, with the judge-circularity bias measured.
RAG for researchers: page-level citations from your personal library, LLM access via MCP. Ask your entire archive, get answers with the intelligence of leading AI.
Superseded by memindex: https://github.com/NikitaHerndlhofer/memindex — local archive of Super Whisper recordings, Cursor chats, and Claude Code sessions. This repo is no longer maintained.
Citation Oracle ✦ — an AI oracle for a PhD literature corpus: neural (bge-m3) semantic search + an interactive citation constellation, plus MIT-licensed Claude Code skills incl. a Gemini Deep Research prompt builder