v0.5

What's changed

The big change here is the removal of LiteLLM use internally. LiteLLM server can still be used to achieve the same functionality as previously. This change makes working with local LLMs such as Llama easier. See README for up-to-date documentation.

Full Changelog: v0.4.7...v0.5

v0.4.7

A lot of warnings (1 from CWhy directly, more from LiteLLM) are now silenced and will not appear.

Full Changelog: v0.4.6...v0.4.7

v0.4.6

What's Changed

Some bug fixes, notably when the LLM tries to call multiple functions at once in conversation mode. Recent Gpt-4/GPT-4-Turbo models seem more willing to do this.

Full Changelog: v0.4.5...v0.4.6

Compatibility fix for Python 3.8

Full Changelog: v0.4.4...v0.4.5

v0.4.4

What's Changed

Bug fix release.

Also introducing new experimental, work-in-progress command, diff-converse.

Full Changelog: v0.4.3...v0.4.4

Preliminary Bedrock support

What's Changed

• Moving to litellm; added Bedrock. by @emeryberger in #56

Full Changelog: v0.4.2...v0.4.3

v0.4.2

What's Changed

Windows support.

Full Changelog: v0.4.1...v0.4.2

v0.4.1

What's Changed

• Performance improvements when CWHY_DISABLE is on.
• Name wrapper with a .py by default.

Full Changelog: v0.4...v0.4.1

v0.4

Warning
Wrapper mode is now default. Watch out for the new API. This is a breaking change for the CLI.

What's Changed

• Add new conversation mode: beginning a back-and-forth GPT mode by @nicovank in #48
• Interface cleanup: wrapper mode default by @nicovank in #49

Full Changelog: v0.3.2...v0.4

Prints cost estimate for each query

What's Changed

• Factor out some functions into llm-utils by @nicovank in #37
• Add cost estimation on text completion by @nicovank in #38

Full Changelog: v0.3.1...v0.3.2