v0.12.0

v0.12.0#

These are the release notes for gptme version v0.12.0.

Contributors#

Thanks to everyone who contributed to this release:

@erikbjare

Changelog#

Changes since v0.11.1:

📦 gptme#

✨ Features (3)#

  • feat: added ToolSpec for all tools, added basic XML-callstyle for tools (40905b0e)

  • feat(wip): started work on subagent (78aa1734)

  • feat: improvements to browsing, including link-following (15bd0acb)

🐛 Fixes (25)#

Click to expand

  • fix: removed order-sensitivity for test (c435b30b)

  • fix: added support for gpt-4o and gpt-4o-mini (9d81b319)

  • fix: fixed issues with applying patches (120233c0)

  • fix: fix loading of base tools when browser tool not available (74e4b189)

  • fix: fixed typing (4c955ffa)

  • fix: disabled test for search with duckduckgo (1411c8d3)

  • fix: simplified is_supported_codeblock (cf30c44f)

  • fix: added debug logging for test (0f3c0654)

  • fix: leftover change leading to test failure (a4042d27)

  • fix: disable stripping dates and common prefixes from stdout/stderr in shell tool (3803f33b)

  • fix: fixed escaping when serializing messages to toml (50e81598)

  • fix: fixed model metadata (added gpt-4-turbo) (9d74bf5d)

  • fix(nit): fixed comment and unused call (f9441b7c)

  • fix: construct prompt from ToolSpec (0c76b975)

  • fix: include a summary of function-tools registered in the Python REPL (0eb5d5a8)

  • fix: renamed register_function_conditional to register_function_if (e256fc96)

  • fix: let other tools register functions in the Python tool REPL (f4688aef)

  • fix: fixed imports in test_eval.py (83a747c2)

  • fix: strip ANSI escape sequences in Python output (a41a3145)

  • fix: fixed bug where Python output is duplicated in result message (a8cc4ef4)

  • fix: fixed bug where tool prompts wouldn’t show on installs without browser extras (92d0f75c)

  • fix: use browser tools by writing Python code, added stripping of data:image’s (bd32bc7a)

  • fix: reraise exception if over tries (b3de8f90)

  • fix: cleaned up code that is now run as test (99d3d3d0)

  • fix: restart shell on broken pipe, handle composite expressions (#70) (8612ca2b)

🔨 Misc (6)#

  • test: fixed _shorten_stdout tests (edb82ded)

  • nit: renamed func loop -> step (cafede58)

  • refactor: refactored eval/main.py into seperate files (3eb49f22)

  • test: increase system prompt token allowance in tests (f1655f63)

  • test: marked evals, disable running by default (14097057)

  • docs: fixed link (d3238e36)

(excluded 2 less relevant commits)

Full Changelog: https://github.com/gptme/gptme/compare/v0.11.1…v0.12.0