v0.28.0

v0.28.0#

These are the release notes for gptme version v0.28.0.

Contributors#

Thanks to everyone who contributed to this release:

@0xbrayo, @erikbjare, @Miyou, @RickardCarlsson4

Changelog#

Changes since v0.27.0:

Summary#

This is a major feature release packed with exciting improvements! 🎉

🚀 Major New Capabilities:

  • GPT-5 Support - Works with OpenAI’s latest model

  • Claude 4 Support - Works with Anthropic’s latest model

  • Bell and Tool Sounds 🔊 - Pleasant notification sounds for user input requested and different operations (enable via /setup or GPTME_DING/GPTME_TOOL_SOUNDS env vars)

  • Perplexity Search 🔍 - Enhanced web search capabilities through the browser tool

  • MCP Integration 🔌 - Support for Model Context Protocol servers, expanding tool ecosystem

🤖 Agent & Automation Enhancements:

  • Setup Command ⚙️ - Easy configuration with /setup

  • Auto-commit - Automatically tell agent to git commit, or do it manually with /commit

  • Agent Workspace Auto-detection - Smarter handling of agent environments

  • Choice Tool - Interactive decision-making capabilities

🛠 Developer Experience:

  • Shell Timeouts ⏱️ - Configurable command timeouts (set GPTME_SHELL_TIMEOUT)

  • Morph Tool ✨ - Alternative fast patching tool for precise code edits

  • Computer Tool Improvements 💻 - Better key sequence chaining for GUI automation

  • Enhanced TTS 🗣️ - Improved text-to-speech with Chatterbox support

🌐 Server & API:

  • Redesigned Server API - Complete API overhaul for better web UI integration

  • Tasks API - New task management capabilities

  • OpenAPI Specification - Comprehensive API documentation

🔧 Quality of Life:

  • 90+ Bug Fixes - Significantly improved stability and reliability

  • Better Configuration - Enhanced config system with user-, project-, and chat-specific settings

  • Improved Error Handling - More user-friendly error messages and recovery

📦 gptme#

✨ Features (39)#

  • feat: enhance telemetry with Prometheus metrics and comprehensive instrumentation (#592) (a73f60b0)

  • feat: improve eval system and fix tool format content extraction (#618) (099c80b4)

  • feat: add GPT-5 support (4ff6d5c9)

  • feat: add timeout support to shell tool (06d85f94)

  • feat: add last/all option to /replay command (ec4bec77)

  • feat: added /api/v2/models endpoint (#612) (b08f64e0)

  • feat: add tool sounds for different operations (#597) (ce6f50a2)

  • feat: add Perplexity search support to browser tool (#566) (c91c9d18)

  • feat: implemented choice tool (#598) (07b602ee)

  • feat: add /setup command for user and project configuration (#596) (2379e665)

  • feat: add ding sound notification with bell audio (#588) (6096474b)

  • feat: add OpenTelemetry integration for performance monitoring (#587) (41ab4283)

  • feat: add support for custom agent names in CLI/TUI interface (658fd3d6)

  • feat: auto-detect agent workspaces (420d3072)

  • feat: added support for –agent-path in CLI, include agent context in addition to workspace context (c40477e0)

  • feat: implement auto-commit (#441) (cf0ed1c5)

  • feat: support the openrouter /v1/models api to list all models with metadata (#575) (88df6238)

  • feat: added alternate patch tool using morph fast apply v2 via openro… (#574) (23258e0b)

  • feat: add support for a GPTME_REASONING flag to explicitly enable/disable reasoning (#568) (df0e0ca7)

  • feat: add automatic inclusion of default project files (#565) (ac05da57)

  • feat: add chatterbox tts support (#541) (84151f97)

  • feat: openapi spec for server (#563) (62bd55d6)

  • feat: add tasks API for new tasks UI in gptme-webui (#562) (6807d484)

  • feat: added ‘worked for <time>’ output behind feature flag (due to issues) (ce78dda1)

  • feat: added read_logs function to browser tool (playwright backend) (#553) (9291e000)

  • feat: add gptme-util models list/info commands with fish completion (82c551f7)

  • feat: add support for setting chat name in chat config (#549) (5c30f5df)

  • feat: added gptme-util llm generate command, rework gptme-util context (#535) (87832050)

  • feat: added basic workspace file browsing api (#530) (0ab8abf9)

  • feat: set enable_suspend (Ctrl+Z) and enable_open_in_editor (Ctrl-X Ctrl-E) for prompt (fixes #525) (#526) (ef33c190)

  • feat: add api_conversation_delete (#519) (4fb3561b)

  • feat: support ChatConfig in server and CLI (#504) (0266e350)

  • feat: added support for context_cmd project config (#511) (8805aa28)

  • feat: added ChatConfig, refactored config, support overriding values (#500) (fc6497c9)

  • feat: added support for MCP server tools (#486) (c43f10a6)

  • feat: redesigned server API (#469) (d8f4fc4a)

  • feat: enhance computer tool with key sequence chaining for both macOS and Linux (#479) (9f05e048)

  • feat: support entering newlines with Ctrl+J (#464) (9204f1ab)

  • feat: support LLM_PROXY_URL and LLM_PROXY_API_KEY (fixes #470) (#471) (1f2bd182)

🐛 Fixes (90)#

Click to expand

  • fix: use correct max_tokens for Anthropic models (b78bd34e)

  • fix: unify thinking budget configuration in Anthropic LLM (f96f610f)

  • fix: updated metadata for gpt-5-mini and gpt-5-nano (72527978)

  • fix: updated metadata for opus 4/4.1 and sonnet 4 (0a5bd991)

  • fix: add permissions to GitHub Actions eval workflow (92b6818d)

  • fix: add system dependencies for PyInstaller builds on Ubuntu (5a91a44d)

  • fix: resolve _struct module error in PyInstaller build (dc4743db)

  • fix: support quoted heredoc delimiters with spaces and consolidate shell tests (9d2ff5e1)

  • fix: switch to morph-v3-fast (morph-v2 deprecated) (051cdfcb)

  • fix: fix /fork incorrectly recursively copying workspace directory (2db544f5)

  • fix: fixed bad syntax in dockerfile (52aacddd)

  • fix: add tree to gptme Dockerfile (5fcd296e)

  • fix: set git config for gptme container so it can make commits (9f4cdcc7)

  • fix: pass provider prefix when using LLM proxy (833a236d)

  • fix: remove outdated print message (06dd8b53)

  • fix: minor improvement to instructions for RAG tool (61853abf)

  • fix: fixed running nested gptme sessions with shell tool waiting for piped stdin by redirecting stdin to /dev/null (#604) (97110820)

  • fix: moved sound media (31f54127)

  • fix: validate CLI tool arguments and provide clear error messages (#600) (e7df4f5c)

  • fix: respect saved conversation config when resuming (#595) (1d78ca30)

  • fix: minor formatting fix in shell tool response to removing excessive newline (b03ec16b)

  • fix: resolve OpenAPI validation errors with undefined dict schema (dcd05f9e)

  • fix: fix failing eval tests by correctly looking up agent name (6210bba6)

  • fix: complete tuple return type refactoring for precommit checks (8c98f140)

  • fix: fix rich console log formatting (be666183)

  • fix: workaround for incorrect pwd in new chats (daae7754)

  • fix: disambiguate agent path and workspace (e4698c09)

  • fix: use shlex.split() for proper shell command parsing in agent creation (04931116)

  • fix: bad rebase (1e92afa5)

  • fix: separate agent path and workspace (f21077f0)

  • fix: strip thinking when checking markdown for codeblocks (#579) (b1c22f1f)

  • fix: morph should fail to apply if file changed since patch generated (edits before confirmation) (#578) (2faa01f8)

  • fix: improved mcp tool confirmation preview (#577) (28c17214)

  • fix: improvements to morph tool, better checks for valid results and output diff in response (c899e789)

  • fix: update default models and add more gemini models (4835fdbf)

  • fix: improve support for nested markdown codeblocks in markdown tool format (#571) (d5b82d71)

  • fix: add AGENTS.md to default context files (2a2a3ee0)

  • fix: fix default serve command for gptme-server using click-default-group (ca5578a9)

  • fix: improve id naming convention for new task conversations (0652c7e6)

  • fix: move workspace prompt to get_prompt (#521) (f0ab6faf)

  • fix: support heredocs with quoted delimiters in shell tool (#558) (df09d8a1)

  • fix: made assistant prompt name and color settable via env vars (315cec9d)

  • fix: improve shell command interrupt handling with partial output capture (#556) (d5e89b6d)

  • fix: streamline gptme-util subcommands to all use list instead of ls (d6ef0b5e)

  • fix: improve how tildes are handled in workspace paths (store/expose with tildes, process without) (bde855f2)

  • fix: resolve workspace path before saving config (63c00411)

  • fix: include conversation workspace metadata in api (#551) (ae7013c6)

  • fix: fixed server tests (aaf80fdf)

  • fix: improvements to server (#547) (077aa878)

  • fix: fix config loading from cli params and env (#545) (0d5a46d0)

  • fix: delay tts availability checks to tool init (#543) (40ca7475)

  • fix: check if pre-commit executable exists before enabling checks (239c1a83)

  • fix: better support for reasoning via openrouter (#538) (f8f1de95)

  • fix: added support for TEMPERATURE and TOP_P environment variables (9c944e62)

  • fix: added shell example to use single quotes for complex content (e5f5d0f2)

  • fix: fixed –workspace=@log behavior after config refactor (#537) (b1824077)

  • fix: disable pre-commit checks running before first user message, misc fixes (87781b6b)

  • fix: improve browser init logging, made ToolSpec.available optionally a Callable with new is_available property (01a8616e)

  • fix: change misleading config warning to debug level (23799d34)

  • fix: output cwd changes in shell responses, error on extra ======= for patch (#534) (7c4fbbb5)

  • fix: use quadruple backticks in markdown codeblocks to avoid conflicts (#533) (2abf6025)

  • fix: add metadata for Claude 4 models (c54ef4a2)

  • fix: improved text file detection in workspace API (350d8532)

  • fix: dont fail pre-commit check if config file not found (76b55482)

  • fix: updated with more Gemini models (e9c4cf3d)

  • fix: reverted accidentally commited changes (7d9d0d92)

  • fix: added check for placeholder lines in save tool (and tests) (#522) (b8d32b6b)

  • fix: added convert_convo.py script (344a1539)

  • fix: better support for interrupting pre-commit checks (d50af6af)

  • fix: fix streaming files support in server (#518) (543e2ee3)

  • fix: fix use of workspace paths in server by syncing workspace set in chat config (new) with workspace symlink (old) (#517) (984d6412)

  • fix: use absolute paths for files in view_image messages, improve markdown markup for paths (f2f57c6e)

  • fix(server): don’t store chat config obj in session, load mcp server from user/project config (#515) (a6504e0e)

  • fix: change mcp tool format to mcp_tool.mcp_func (instead of mcp_tool_mcp_func) (385901a3)

  • fix: fixed prompt not showing when prompt text overflows (#510) (e7657223)

  • fix: remove default BASE ARG from Dockerfile.server to remove the need for skaffold workarounds (#512) (a0b507ef)

  • fix: misc fixes by mike (#501) (adfd31f8)

  • fix: fixes auto-stepping in server (#499) (3b70dd8c)

  • fix: improved the gh-pr-view script and rewrote in Python (#483) (ba032393)

  • fix: change default gptme-server port to 5700 (#491) (33c2ba7f)

  • fix: add screenshot example (d9be0727)

  • fix: fix weird needed {!r} in typecheck (https://github.com/gptme/gptme/pull/486#issuecomment-2763345203) (f5c58db0)

  • fix: dont run tree outside git repositories (likely to timeout) (2fb0210d)

  • fix: support credentials in CORS (958bcb89)

  • fix(server): support non-streaming LLM requests (#481) (085b726b)

  • fix: refactored tooluse detection in server to save a call (64888eb9)

  • fix(server): fixes interruptions during generation (adad1144)

  • fix: fixed ‘f-string expression part cannot include a backslash’ (cbc69653)

  • fix: fix missing MODELS entry for azure (fixes #472) (67cad0ac)

  • fix(tts): correctly strip <think> tags and leading hashes for ‘# Headers’ (7fecf09f)

🔨 Misc (57)#

Click to expand

  • chore: bump version to 0.28.0 (41d31f10)

  • docs: updated README usage, CLI help strings, and set gpt-5 as default OpenAI model (b5a1abd7)

  • docs: clarify max_tokens calculation rationale for thinking feature (3395bf23)

  • docs: improved prometheus config and docs (037e51d4)

  • docs: improve setup instructions for agents (include uv) (4ac4a45f)

  • docs: add bench of startup time using hyperfine (a0c3ed39)

  • refactor: refactor to eliminate duplicate lines (#608) (309ae85f)

  • docs: mention lessons system (1e8d62f3)

  • docs: fixed ref and improved note about agents in web ui (85680cf5)

  • docs: added mermaid diagrams to agents docs (63f2345e)

  • refactor: change how extra_headers and extra_data is set in openai client (860f1e89)

  • docs: link to llms.txt and llms-full.txt in docs index (e5b4fb97)

  • docs: improve github bot doc (d65c1ee4)

  • docs: minor formatting fix (2e9cf77b)

  • docs: add telemetry setup guide and simplify telemetry init (87cb53d4)

  • docs: more updates to docs to recommend Claude Sonnet 4 (9cf99057)

  • docs: update docs to recommend Claude Sonnet 4 (2740eca5)

  • refactor: split api_v2.py into focused modules (c2b67525)

  • refactor: break down complex functions into smaller focused units (2207190c)

  • refactor: reduce code duplication with CommandContext (c884d8d1)

  • refactor: convert command handling to decorator-based registry (04a6af0d)

  • refactor: simplify chat function signatures and improve autocommit prompting (a37ddf7c)

  • refactor: simplify chat loop and fix interruption handling (#582) (a012815e)

  • feat: add create agent endpoint in v2 api (c2759a5a)

  • refactor: improve error handling and message formatting in morph/patch tools (#581) (5b19708b)

  • chore: improve gitignore and dockerignore (ab9627a0)

  • docs: updated docs (#564) (f0842e84)

  • docs: added sphinx_llms_txt extension to generate llms.txt and llms-full.txt (#561) (8181e276)

  • tests: fixed test (a78c0561)

  • docs: fix duplicate autosectionlabel warnings by limiting maxdepth (5ca5e83b)

  • docs: added back missing file from last PR (d5507b7e)

  • docs: mention Ctrl+X Ctrl+E support in –help output (ae3d2857)

  • docs: improve alternatives, config and usage documentation (4d310b83)

  • docs: changed MCP example to use mcp-server-sqlite (d7d9470a)

  • docs: moved environment variables section to bottom of config doc, add link to it from lead (f83568a4)

  • docs: improved docs for config, including context_cmd description and Chat Config section (46a4628c)

  • refactor: thread-local globals for _config and _loaded_tools (#506) (ad888d38)

  • docs: improved docs, added docstrings and minor improvements to config code (#505) (c0c5c739)

  • refactor: improve config and tools architecture (#503) (dc7be5fa)

  • docs: add a copiable markdown for powerd by gptme badge (#490) (0b3a3967)

  • docs: mention mcp in custom_tool doc (b24c0f95)

  • tests: mark tests failing in ci with xfail (ed0e8ada)

  • docs: strip ansi color codes from jscpd output (5396094e)

  • docs: improved MPC docs (150a35b5)

  • tests: improved mcp tests, replaced weather mcp server test with memory server (c7f7014b)

  • docs: fix crazy wrong MCP explanation in docs (ff52d7ba)

  • docs: fixed title underline too short (15c3f403)

  • docs: moved MCP guide into documentation (078e667f)

  • tests: add xfail to flaky tests (6ddf8a3b)

  • refactor: reorder Message attributes (22f590ab)

  • chore: add dev/ to .gitignore for dev branch worktrees (c19893b3)

  • docs: add jscpd output to make metrics and arewetiny doc (50bcae47)

  • docs: added sitemap.xml (#480) (5d4df786)

  • docs: improved timeline further (2665a9e3)

  • docs: fixed rst list formatting, added check_rst_formatting.py script and added to pre-commit hooks (e88bd567)

  • docs: updated timeline (34bddb2b)

  • docs: improved tts usage docs (598e1b29)

(excluded 19 less relevant commits)

Full Changelog: https://github.com/gptme/gptme/compare/v0.27.0…v0.28.0