v0.27.0

v0.27.0#

These are the release notes for gptme version v0.27.0.

Contributors#

Thanks to everyone who contributed to this release:

@0xbrayo, @erikbjare, @jrmi, @Miyou

Changelog#

Changes since v0.26.0:

📦 gptme#

✨ Features (19)#

  • feat: added GPTME_CONTEXT_TREE which includes tree --gitignore . output in initial prompt (#461) (593ee900)

  • feat(style): print <think>/<thinking> in dimmed text (dfab1e05)

  • feat: added basic files support to server API (#440) (24b9dc95)

  • feat: added script to auto-rename past conversations (#439) (9a9f0114)

  • feat: added terminal title status output (#438) (37c877f3)

  • feat: much improved RAG, added LLM post-processing of results (#435) (f4b6f02e)

  • feat: use gnome-screenshot in wayland (#436) (6d1d47a4)

  • feat: add status emojis to system messages (#429) (69c5f011)

  • feat: added support for nvidia provider (7f387d33)

  • feat: auto-enable pre-commit checks when config exists (#430) (5cd9b1e8)

  • feat: enhance vision tool with better image processing (6a982987)

  • feat: add option to override base prompt (#426) (78e92091)

  • feat(computer): add macOS support for computer tool (#324) (7a08e341)

  • feat: added support for GPTME_BREAK_ON_TOOLUSE flag to not stop generation when tooluse occurs in stream (e7021b9c)

  • feat: auto-run precommit checks when tooluse exhausted (#415) (254d1b61)

  • feat: added support for deepseek-reasoner (#410) (ac3b8258)

  • feat: added /model command to list/switch models, support mixing providers (#409) (a2bc4e00)

  • feat: add env var feature flag for llm prompt suggestions (978b1c4b)

  • feat: added gptme-wut (#402) (91bdce67)

🐛 Fixes (65)#

Click to expand

  • fix: fixed reduce_context script (2f8246c6)

  • fix: added reduce_context.py script (f0edadde)

  • fix: added gh-pr-view-with-pr-comments.sh script (3e0dd23d)

  • fix: add gemini-2.0-flash-thinking-exp-01-21 to models (#465) (489f73e6)

  • fix: fix /export command (#460) (0ce55149)

  • fix: fix missing dependency warning (#455) (ced1b9ac)

  • fix: fixed incorrect reference to GPTME_FRESH_CONTEXT (now just GPTME_FRESH) (ce626ed1)

  • fix: fix circular import (417613c1)

  • fix: correctly handle too long path names (ec1118c6)

  • fix: dont expand paths in user commands for tools, like /shell (#453) (96304695)

  • fix: fixed a bunch of get_default_model uses, return None instead of exception if unset (a3626c82)

  • fix: make model fully optional in prompts.py, no fallback to possibly unset default (29a0cb19)

  • fix: correct content block processing in Anthropic LLM module (b4d2d962)

  • fix(anthropic): use model_meta.supports_reasoning to determine wether to use thinking (91af56c1)

  • fix: add supports_reasoning to model_meta, disabling <thinking> prompting for such models (b15fb6a5)

  • fix(anthropic): set max_tokens from model metadata (a89ca157)

  • fix: refactor only-once warning of unknown model metadata (5d5f0e95)

  • fix: disabled tool tooluse format for Sonnet 3.7 with thinking, for now (4354f3e7)

  • fix: only log missing model metadata warning once (ffb83789)

  • fix: added model metadata for sonnet 3.7 (1df392ac)

  • fix: add support for sonnet 3.7 (cb83725c)

  • fix: add model parameter to step function (dont rely on global) (ebc076bb)

  • fix: refactored get_project_dir helper function (d231311e)

  • fix: prompt to prefer absolute paths (ba11552a)

  • fix: fixes after RAG improvement PR (486d9858)

  • fix: refactor workspace initialization (6936a4e1)

  • fix: fix support for explicitly disabling pre-commit checks (603ecaae)

  • fix: added basic gptme-util chats search command, fixes to gptme-util chats overall (#434) (b9a64578)

  • fix: adjust tts speed to 1.0 by default after tts_server.py updated to kokoro 1.0 which seems faster by default (6e3af447)

  • fix: upgrade tts_server.py to use kokoro 1.0, fix audio output device on linux (#432) (007a0750)

  • fix: improved status emojis in formatted messages (964b5280)

  • fix: automatically clone Kokoro-82M repo in tts_server.py script (a6109c71)

  • fix: fixed support for o3 (3a810109)

  • fix: log reasoning_content when available (i.e. Deepseek R1) (e13af3f3)

  • fix: added support for GPTME_PATCH_RECOVERY where file is returned in error for non-matching patches (96ceec24)

  • fix: minor improvements to ipython function description formatting (7d2fec38)

  • fix: auto-step in server (wip), dont use logmanager lock in API (c948e3b1)

  • fix: improve formatting for large patch warnings (4ad126e3)

  • fix: handle None workspace in get_project_config to prevent TypeError (d0fee8a6)

  • fix: minor logging improvements (48782fcd)

  • fix(screenshot): add prompting for screenshot permissions on macos (3bc328ea)

  • fix(vision): rescale large images passed to view_image (1a1d969c)

  • fix(tts): fix logging output in tts server (060b5d09)

  • fix(tts): added GPTME_VOICE_FINISH flag to wait for speech to finish before exiting (95042d40)

  • fix(tts): combine short sentences into larger chunks before sending to tts server (d6a941ea)

  • fix: remove exclusive mode (#423) (0c99cb8f)

  • fix(tts): make tts server requests non-blocking, improve clean_for_speech (#422) (44c3dfe9)

  • fix: improve pre-commit output message when files automatically fixed by hook (9b4e9117)

  • fix: added support for knowledge cutoff in model metadata (d166b4cb)

  • fix(eval): use seperate status emoji for timeouts (00fdd187)

  • fix: locate espeak library more intelligently on macOS (583a6b43)

  • fix: strip leading/trailing silence from tts output (#420) (96c54c41)

  • fix: fix call_id regex format for deepseek (ec30e9b0)

  • fix: fixed tts_server.py on macOS (#418) (477a81e2)

  • fix: enable tool format for deepseek provider (7fda8bdc)

  • fix: fixed summarization for openrouter and deepseek (d8a9bec6)

  • fix: added TODOs for better openrouter support (e56491f1)

  • fix: fixed output of codeblocks with unescaped rich [style] tags (07ed85fd)

  • fix: detect tooluses with common path characters when cleaning for speech (#412) (9124640b)

  • fix: improve default models+toolformats to run in evals (autodetect from available keys) (598f4096)

  • fix: fixed broken openrouter support due to missing entry in model metadata (0cd60a08)

  • fix: added gptme-eval-docker.sh helper script (9e62c3d1)

  • fix: broken mixed tool formats execution while using tool format (#407) (bcf45cd5)

  • fix: improved logging output for gptme-rag calls (incl time taken) (33db67ec)

  • fix: cleaned up tts tool, detect if tts server isnt running (#404) (d27da5e0)

🔨 Misc (42)#

Click to expand

  • chore: bump version to 0.27.0 (96a5b54e)

  • docs: replaced github.com/ErikBjare/gptme links with github.com/gptme/gptme (6ac6f6e7)

  • docs: reference both usage and examples in getting-started guide (1bd6a607)

  • docs: fixed link (bb1c3f3f)

  • docs: link to bob from README (2349c046)

  • refactor: make computer tool share logic with screenshot tool (#442) (412bace5)

  • docs: mention Claude Code in alternatives (05648fb4)

  • refactor: refactored include_paths in chat.py by moving in and helpers into gptme.util.context (f008b6a9)

  • docs: update the evals page to mention recommended model (sonnet) and available evals (#451) (92965cd1)

  • refactor: extract get_default_model_summary (161451e1)

  • docs: fixed broken document_prompt_function (004ced9e)

  • refactor: separate all get_default_model logic from get_model (5e3aa80a)

  • tests: uncomment test and mark to skip instead (40dc44c1)

  • tests: disable broken search_ddg test (4a241f4c)

  • refactor: refactored retry_(generator_)on_overloaded (almost-duplicate decorator) (e64006cc)

  • docs: add OpenHands to alternatives (1edf80d9)

  • docs: split sections from ‘dev guide’ into new ‘about’ toctree in index (8845a1a6)

  • chore: updated gitignore (2bdad1a7)

  • docs: improved lead in README (d07e799c)

  • docs: improved tools page (dae30218)

  • docs: updated page with alternatives/comparison (5d53aa42)

  • test: fixed conversation search test (8dfc3a06)

  • refactor: move default rag post-process prompt from config.py into tools/rag.py (0b9ea550)

  • docs: fixes to config docs (2b11a3c7)

  • test: add requires_api test mark to run tests without API keys (such as for untrusted PRs) (#433) (19895aa8)

  • docs: improve documentation and code style (8c536f17)

  • docs: improve TTS documentation and update README (00198a35)

  • docs: added complexity metrics with radon to arewetiny docs, as makefile target (#428) (0bd43f3e)

  • docs: fix RST formatting in config.rst (2059dddf)

  • docs: document potential DeepSeek/Gemini reasoning content support (ab780b99)

  • style: add newline between patch success message and warnings (8698758b)

  • style: improve readability of base_prompt assignment (79df878a)

  • refactor: improve imports and error handling in cli.py (de764dac)

  • docs: document environment variables and feature flags (693b3be0)

  • chore: added config.toml to gitignore (456d9661)

  • chore: added config.toml to gitignore (87d23cfa)

  • tests: fixed tests for splitting/chunking sentences in tts (af0d3588)

  • test: generate unique run name for retries (6ff8b94b)

  • test: fixed test expecting ‘Cost’ which isn’t known for some models (db212bd7)

  • test: fixed missing argument in test (0dbcf09e)

  • test: disabled test_search_google since its failing, improved error logging for failed searches with playwright (51f54e9a)

  • chore: updated contributor cache (719da49f)

(excluded 9 less relevant commits)

Full Changelog: https://github.com/gptme/gptme/compare/v0.26.0…v0.27.0