antflydb/antfly

Antfly is a distributed search engine written in Go that goes well beyond basic keyword search. It combines three different ways of finding information in a single query: traditional word matching (called BM25), vector similarity search (which finds content that means the same thing even if worded differently), and graph traversal (which follows relationships between pieces of data). It can index and search not just text but also images, audio, and video, using AI models to understand what they contain. The system is built to scale horizontally across many machines using a consensus algorithm called Raft, the same technology behind the etcd database, which keeps data consistent and replicated. When you write data to Antfly, it automatically generates the embeddings (numerical representations of meaning), chunks long documents into searchable pieces, and extracts relationships for the graph index. Built-in RAG agents (retrieval-augmented generation) let you connect the search results directly to a large language model for conversational question-answering over your data. You can connect your own AI models via Ollama, OpenAI, Bedrock, or Google. Client libraries are available in Go, TypeScript, Python, and React. There's also a PostgreSQL extension that lets you run Antfly searches directly from SQL queries. A web dashboard called Antfarm provides visual playgrounds for all features. The core server uses the Elastic License 2.0, while the SDKs and tools are Apache 2.0.