A minimal command-line Agent built with smolagents. Supports any LLM provider with an OpenAI-compatible API.

# Create and activate virtual environment
python3 -m venv venv
source venv/bin/activate  # Windows: venv\Scripts\activate

# Install dependencies
pip install -r requirements.txt

# Copy and edit environment variables
cp .env.example .env
# Edit .env with your LLM provider details

MiniAgent uses OpenAI-compatible APIs by default. Configure your provider in .env:

OpenAI:

LLM_API_KEY=sk-your-openai-key
LLM_BASE_URL=https://api.openai.com/v1
LLM_MODEL=gpt-4o-mini

DeepSeek:

LLM_API_KEY=sk-your-deepseek-key
LLM_BASE_URL=https://api.deepseek.com/v1
LLM_MODEL=deepseek-chat

Other providers (Moonshot, Zhipu, local Ollama, etc.): just set the corresponding BASE_URL and MODEL.

source venv/bin/activate
python agent.py

• read_file(path) — Read file contents
• write_file(path, content) — Write content to file (system paths blocked)
• run_command(command) — Execute shell commands (dangerous commands require confirmation)
• web_search(query) — Search the web via DuckDuckGo

1. Step 1: Basic chat (understand Agent loop)
2. Step 2: File I/O tools (understand Function Calling)
3. Step 3: Shell command tool (understand safety boundaries)
4. Step 4: Web search tool (understand tool chaining)

MiniAgent/
├── agent.py              # Main CLI entry
├── config.py             # Configuration & API keys
├── tools/
│   ├── file_tool.py      # File read/write with path guard
│   ├── shell_tool.py     # Command execution with safety guard
│   └── search_tool.py    # Web search
├── .env                  # API keys (gitignored)
├── .env.example          # Environment template
└── requirements.txt      # Dependencies