omega 0.5.0

New

• Keyboard shortcuts. shift+tab cycles the approval mode (asks before risky actions → auto approve → read only) and ctrl+p (or alt+p) switches to the run's next model, from the next turn, without stopping the run. Switching to read only tells the model, which then offers changes on a card instead of running them. A model you pick by hand stays for the rest of the run.
• Slash commands with autocomplete. Type / in the input box for a menu that lists and explains the commands as you type: /help, /mode, /model, /expand, /exit and /quit. Tab completes, Enter runs, and after /mode or /model the menu offers the modes or the run's models. Without an argument, /mode and /model open a card.
• Approve all is visible. Answering a on an approval card now switches to auto approve, shown in the status bar and reversible with shift+tab.

Website

• Logos of the supported providers (Anthropic, OpenAI, OpenRouter, OpenCode Zen, Ollama, LM Studio), a section on keyboard shortcuts and commands, and a new FAQ entry.

Install: brew install lcajigasm/omega/omega

omega 0.4.0

New

• sudo inside the interface. When Mole, softwareupdate or any command needs administrator access, omega asks for the password on a card, checks it with sudo and keeps it in memory for that run only. It is never logged or sent to the AI. Esc skips what needs admin access; --sudo asks before the run.
• System logs review. A new system-logs step reads crash reports, launch agents that keep failing, sleep blockers, processes flooding the log and log sizes, and the agent offers fixes.
• Install offers. When a cleanup needs a tool you do not have (mas, cargo-update, topgrade, Mole, Docker Desktop), omega offers to install it on a card.
• Docker not running. omega offers to start Docker Desktop before measuring or cleaning it.

Fixes

• Commands that call sudo no longer write over the interface or hang on a hidden prompt: commands now run detached from the terminal, and timeouts stop the whole process group.
• Website FAQ layout.

Install: brew install lcajigasm/omega/omega

omega 0.3.0

New

• Selection cards. Recommendations arrive on a card: tick one or several actions and omega carries out exactly those. analyze and tune stay read only until you pick something.
• Sensible model choice. Automatic by default: a fast model (Claude Haiku 5.5, GPT 5.4 mini) for routine work that switches to a stronger one on the first failure. omega model shows every model with its price. Conversation caching on the Anthropic API.
• Local models. Ollama, LM Studio and any OpenAI-compatible endpoint, with no API key.
• omega tune. Reviews the configuration of your installed tools against their current docs (config hints, web search and page reading) and proposes improvements with their source. Guided, or --auto for low-risk ones.
• omega docker. Shows what Docker uses and cleans only what you tick. Docker cleanups always need your confirmation, even with --yes.
• New TUI in the style of Claude Code: tool call blocks, rendered Markdown, animated activity line, status bar with tokens, cost and time, ctrl+o to expand output.

Fixes

• Many read-only commands (git config --list, env, dry runs, --help) are no longer denied in read-only mode; denials are shown and logged.
• Token and cost accounting includes prompt cache reads and writes.
• Cards no longer close on their own in some terminals.

If you set up omega with an earlier version, your config may pin a large model: run omega model auto.

Install: brew install lcajigasm/omega/omega

omega 0.2.1

• omega setup and omega auth set validate the API key before saving it: they reject URLs and text with spaces, strip paste artefacts, warn on unexpected prefixes and check the key with the provider.
• omega auth status flags stored values that are not API keys.

omega 0.2.0

• Interface in English, Spanish, French, German and Portuguese (OMEGA_LANG, language setting or system locale). The agent replies in the same language.
• Documentation in English.
• Agent prompt and tool definitions in English.

Install: brew install lcajigasm/omega/omega

omega 0.1.0

First release.

• AI-driven macOS maintenance CLI/TUI (updates like topgrade, cleans like Mole).
• Providers: Anthropic, OpenAI, OpenRouter and OpenCode Zen.
• Risk-based approval policy, read-only mode and blocked commands.
• Native steps plus topgrade/Mole as backends, memory between runs.