v0.11.0#

These are the release notes for gptme version v0.11.0.

Contributors#

Thanks to everyone who contributed to this release:

@AntonOsika, @erikbjare

Changelog#

Changes since v0.10.5:

📦 gptme#

✨ Features (2)#

  • feat: mirror working directory in shell and Python process (#49) (0b9c3228)

  • feat: support paths/URLs in any prompt, refactored entrypoint to call a new public API with core logic (#37) (aaf60e57)

🐛 Fixes (9)#

Click to expand

  • fix: exit with appropriate exitcode if evals pass/fail (3a0e4dca)

  • fix: fixed shell cd test (9932b27a)

  • fix: fixed bugs in eval, upload/download binary files, cd to cwd before every shell command (cefbbe86)

  • fix: fixed shell output printing (no extra newlines) (cf91873c)

  • fix: fixed a spelling error (8c1eadab)

  • fix: import NotRequired from typing_extensions (2718ebac)

  • fix: improved path detection in prompt (3f74635d)

  • fix: add price_input and price_output to model metadata, refactored ModelDict TypedDict into ModelMeta dataclass (a0f1a731)

  • fix: switched to ipython for handling Python execution (#41) (b75182c7)

🔨 Misc (14)#

Click to expand

  • docs: improved entrypoints’ docs for better cli docs (89506507)

  • docs: added demos page to docs (ab2687c9)

  • test: run evals as tests, refactor evals, added python-xdist for parallel testing (14ca2df6)

  • test: switch from gpt-3.5-turbo to gpt-4-1106-preview (“gpt-4-turbo”) in cli tests (75e79bda)

  • refactor: made eval abstractions more general (#48) (bf64f208)

  • refactor: moved init code into init.py (0e1a0f5e)

  • test: minor improvements to eval (e3aa3363)

  • test: further eval improvements (bfc2f14f)

  • test: added eval test that accepts stdin (28e3a3cc)

  • test: fixed bugs and improved output in evals (953614f5)

  • test: continued work on evals (0c070ec2)

  • test: added basic eval code (be678f3a)

  • docs: fixed link to demo (66e9e49a)

  • docs: added more demos (7de6c74a)

Full Changelog: https://github.com/gptme/gptme/compare/v0.10.5…v0.11.0