Provides a local Retrieval-Augmented Generation (RAG) server written in Go for semantic search and knowledge recall.
Sponsored
Goldie is a powerful, locally-run RAG server built in Go, designed to augment your AI interactions by providing context from your personal knowledge base. It allows you to index various forms of content, including local files, directories, and even past AI conversations, using either local MiniLM embeddings or Ollama models. By storing embeddings in a SQLite database, Goldie enables semantic search and intelligent knowledge recall, making your local data seamlessly accessible to tools like Claude Code, Claude Desktop, and OpenAI Codex for richer, context-aware responses.
Key Features
010 GitHub stars
02Multiple embedding backends: MiniLM (local via ONNX Runtime) or Ollama (supports various models)
03Local embeddings using all-MiniLM-L6-v2 model (384 dimensions)
04Persistent SQLite vector storage with sqlite-vec extension
05Automatic document chunking with overlap for optimal retrieval
06Semantic search, directory indexing, and intelligent knowledge recall
Use Cases
01Augmenting AI assistants (Claude Code, Claude Desktop, OpenAI Codex) with local, context-specific knowledge.
02Creating a searchable personal knowledge base from local documents, notes, and web content.
03Indexing and searching past AI conversation histories to find solutions, code snippets, and track project history.
