**Changes**
- Created a common abstraction for prompts and token helpers. For token helper functions, they now require a tokenizer as input, even for OpenAI. This has the benefit of saving compute time if the token helpers are called frequently since the tokenizer only needs to initialized once.
- Move ollama to `local_llm` and make `llm` a requirement of `local_llm`
- common `chat_completion` returns both prompt and completion tokens
- Add latency as a common output of all `chat_completion` functions
- Automatically set HF telemetry and warnings to off wherever it is imported
- Update dependencies including to numpy 2.0.0
**Full Changelog**: https://github.com/DaveCoDev/not-again-ai/compare/v0.9.0...v0.10.0
Available on [PyPI](https://pypi.org/project/not-again-ai/0.10.0/).