Open-source text-to-speech library with a 1.7B 1-bit speech model and lightweight ONNX models that run on CPU.
Python library and SDK that builds a hierarchical tree index of documents so an LLM can reason through it for retrieval, without vector databases or chunking.