LlamaIndex

Formerly called GPT Index, this Python framework links language models to private data and to external tools. Reader classes pull in files, APIs, and databases; the content is split into nodes, embedded, and stored in an index, and at question time retrievers and query engines pick the most relevant chunks and hand them to the model as context, the usual RAG loop. On top of that sit an event-driven Workflows system and agent classes for multi-step, tool-calling programs, plus helpers for structured output. You can install the llama-index starter bundle or start from llama-index-core and add only the integration packages you need, and there are hundreds of them covering hosted LLM providers, local runtimes such as Ollama, embedding models, and vector stores. It needs Python 3.10 or newer and no GPU unless you also run models locally. The code is MIT licensed and free for commercial use. The company behind it now puts most of its effort into LlamaParse, a commercial document parsing cloud, so the framework best suits developers who want a broad, mature toolkit and are happy to follow its own release notes.