Gptty

**Full Changelog**: https://github.com/signebedi/gptty/compare/0.2.6...0.2.7
**Project Roadmap**: https://github.com/signebedi/gptty/issues/1

Introduces the `UniversalCompletion` class to allow users to abstractly query the OpenAI API without needing to worry excessively about model types. (39) Admittedly, the difference in the structure of queries between `Completion` and `ChatCompletion` models has not been fully abstracted away, but will be addressed in a future release. (68)

This release also added some pdoc-generated documentation, though the quality may be slightly lacking. (66)

**Full Changelog**: https://github.com/signebedi/gptty/compare/0.2.5...0.2.6
**Project Roadmap**: https://github.com/signebedi/gptty/issues/1

Fixes a bug where API calls do not work when an organization ID has not been set. (63)

**Full Changelog**: https://github.com/signebedi/gptty/compare/0.2.4...0.2.5
**Project Roadmap**: https://github.com/signebedi/gptty/issues/1

Adds support for additional context, see 37, and a test suite for this new functionality, see 54. Added checks for internet connectivity, see 48. Added support for JSON output, see 46, and for quiet output, see 47. Added general look and feel improvements, see 50 and 51.

**Full Changelog**: https://github.com/signebedi/gptty/compare/0.2.3...0.2.4
**Project Roadmap**: https://github.com/signebedi/gptty/issues/1

Added verbose / debug mode, see 28. Use `tiktoken` library to assess token counts, see 29. Wrote unittests, see 40, and implemented with Github Actions, see 41.

**Full Changelog**: https://github.com/signebedi/gptty/compare/0.2.2...0.2.3
**Project Roadmap**: https://github.com/signebedi/gptty/issues/1

Added support for ChatCompletions, see 31 and 36. Also added support for validating the model passed by the user, see 32. Also added wider response formatting options, see 24.

**Full Changelog**: https://github.com/signebedi/gptty/compare/0.2.1...0.2.2
**Project Roadmap**: https://github.com/signebedi/gptty/issues/1

Added keyword tokenization options, so users can now choose whether to reduce OpenAI API usage (through keyword tokenization) or (the very likely slightly better results of) the standard approach where the entire context is passed, see 25. We also weight context so that more recent context gets prioritized, see 26.