使用 Homebrew, Nix 安装 oterm

项目链接

• 主页
• 上游文档
• 软件包管理器页面

命令和别名

• oterm

项目历史

GitHub records `ggozad/oterm` as created on 2023-10-10, during the wave of local-LLM tooling around Ollama. Early package descriptions and PyPI metadata framed it as a text-based terminal client for Ollama with persistent chat sessions, SQLite storage, model selection, system prompts, and parameter customization.

The 2026 release line changed the project's scope. Release 0.15.0, published on 2026-04-27, is described as the multi-provider release: chats moved through pydantic-ai, MCP integration was rewritten, the UI was refreshed, streaming improved, and old Ollama-only assumptions such as an internal pull-model command were removed. The README describes oterm as a terminal client for Ollama, OpenAI, Anthropic, and any pydantic-ai-supported provider.

采用历史

oterm's adoption follows the developer interest in local and terminal-first LLM workflows. It is listed in the Ollama repository's terminal/CLI ecosystem list, and Homebrew/Nix packaging made it installable in the same package-manager channels as other developer CLIs.

The tool's multi-provider turn broadened its appeal from an Ollama companion to a terminal chat front end that can sit over local models, hosted APIs, and OpenAI-compatible servers such as vLLM, LM Studio, llama.cpp, OpenRouter, and LiteLLM.

使用方式

Users install and run `oterm` to get a terminal UI with persistent chat sessions, provider/model selection, streamed Markdown output, configurable system prompts and parameters, and optional MCP tool configuration. The documented config locations in the av.db input show it stores per-user configuration under the platform's application data directory, with environment-variable overrides for secrets and data directory selection.

For package nerds, the interesting usage pattern is that oterm packages a full TUI experience as a CLI entry point: the Python package is not merely a library, but an end-user application distributed through PyPI-style tooling and system package managers.

为什么软件包爱好者会关心

oterm is a good example of the fast-moving LLM CLI package category: a small terminal app can quickly become a package-manager target because users want one-command installation, upgrades, and config discovery. Its 0.15.0 migration also shows how model-provider abstraction moved from app-specific integrations to library-mediated provider layers such as pydantic-ai.

时间线

• 2023-10-10: GitHub records creation of `ggozad/oterm`.
• 2024: PyPI release 0.6.0 described oterm as a text-based terminal client for Ollama with persistent SQLite sessions and model customization.
• 2026-04-27: release 0.15.0 changed oterm into a multi-provider pydantic-ai-based terminal client and rewrote MCP configuration.
• 2026-06-30: release 0.20.0 updated pydantic dependencies and improved Modelfile-based prompt/parameter prepopulation.

Related projects

• Ollama is the original local-model runtime oterm was built around.
• pydantic-ai became the provider abstraction layer for oterm's multi-provider architecture.
• MCP-compatible configuration connects oterm to the same tool-server ecosystem used by desktop AI clients and editors.

来源

• https://api.github.com/repos/ggozad/oterm
• https://formulae.brew.sh/formula/oterm
• https://ggozad.github.io/oterm/
• https://github.com/ggozad/oterm
• https://github.com/ggozad/oterm/releases/tag/0.15.0
• https://github.com/ggozad/oterm/releases/tag/0.20.0
• https://github.com/ollama/ollama
• https://pypi.org/project/oterm/0.6.0/
• https://raw.githubusercontent.com/ggozad/oterm/main/README.md