AI terminal assistant powered by OpenRouter — minimal, fast, developer UX.

---

Source Code: https://github.com/ndugram/hitmos

---

Hitmos is a modern AI terminal assistant for developers, powered by OpenRouter. It brings a Claude Code–styled UX to your terminal — live markdown streaming, interactive model switching, project context injection, and Tab completion, all with zero latency overhead.

Key features:

• Streaming — responses stream token-by-token with live Markdown rendering via rich.
• Multi-model — switch between DeepSeek, Gemini, Claude, GPT-4o, Llama, and more with an interactive arrow-key picker.
• Project context — automatically injects your project files into the system prompt so the model can answer project-specific questions.
• Tab completion — /model <Tab> lists all available models; / completes all commands.
• Minimal UX — Claude Code–inspired welcome panel, ◆ status lines, clean > prompt.
• Fast — built on fasthttp-client with full async SSE streaming.
• Configurable — API key via env (OPEN_TOKEN, OPENROUTER_API_KEY) or ~/.hitmos/config.toml.

Requirements

Python 3.13+

Hitmos depends on:

• fasthttp-client — async HTTP transport with SSE streaming.
• rich — live Markdown rendering and styled console output.
• typer — CLI interface.
• orjson — fast JSON parsing.
• prompt-toolkit — async input with Tab completion (bundled with questionary).
• questionary — interactive model picker.
• pydantic-settings — config management.

Installation

$ pip install hitmos

---> 100%

Quickstart

Save your API key

$ hitmos login

Or set an environment variable:

$ export OPEN_TOKEN=sk-or-...

Start chatting

$ hitmos

You will see:

╭─────────────────────────────────────────────────────╮
│                                                     │
│  ✻  Welcome to Hitmos                               │
│                                                     │
│  /help for commands                                 │
│                                                     │
╰─────────────────────────────────────────────────────╯

 ◆ ~/my-project
 ◆ deepseek/deepseek-chat
 ◆ context 12 KB

>

Ask anything about your project:

> what does this project do?
> add error handling to client.py
> explain the streaming logic

Commands

Command	Description
/help	Show available commands
/model	Interactive model picker (↑↓ + Enter)
/model <name>	Switch model directly
/clear	Clear conversation history
/reset	Reset context
/exit	Exit Hitmos

Switch model interactively

Type /model and press Enter to open the arrow-key picker:

? Select model  (↑↓ navigate, Enter confirm, Ctrl+C cancel)
❯ deepseek/deepseek-chat                   DeepSeek Chat · fast, cheap, great for code
  deepseek/deepseek-r1                     DeepSeek R1 · reasoning model
  google/gemini-2.5-flash                  Gemini 2.5 Flash · fast multimodal
  google/gemini-2.5-pro                    Gemini 2.5 Pro · powerful multimodal
  anthropic/claude-3.5-haiku               Claude 3.5 Haiku · fast Anthropic
  anthropic/claude-sonnet-4-5              Claude Sonnet 4.5 · balanced Anthropic
  openai/gpt-4o-mini                       GPT-4o Mini · OpenAI fast
  openai/o4-mini                           o4-mini · OpenAI reasoning
  meta-llama/llama-3.3-70b-instruct        Llama 3.3 70B · open source
  mistralai/mistral-small-3.2-24b-instruct Mistral Small 3.2 · lightweight

Or use Tab completion: type /model deep then press Tab.

Configuration

Hitmos resolves the API key in this order:

1. OPEN_TOKEN environment variable
2. OPENROUTER_API_KEY environment variable
3. ~/.hitmos/config.toml

Config file format:

token = "sk-or-..."
model = "deepseek/deepseek-chat"

Contributing

Contributions are welcome! Please open an issue or pull request on GitHub.

License

This project is licensed under the terms of the MIT license.