A command-line interface for interacting with various Large Language Models with beautiful markdown-formatted responses.
Design Principle: Frictionless interaction. Just type lm hello - no need for subcommands. The CLI defaults to chat mode for the fastest possible workflow.
uv tool install streamlmpip install streamlmbrew install jeffmylife/streamlm/streamlmAfter installation, you can use the lm command. The CLI defaults to chat mode - just type your prompt:
lm explain quantum computing
lm -m gpt-4o "write a Python function"
lm -m claude-3-5-sonnet "analyze this data"
# Explicit 'chat' command also works
lm chat "hello world"StreamLM supports routing requests through different gateways for cost optimization and flexibility:
# Route through Vercel AI Gateway (no markup, low latency)
lm --gateway vercel "explain quantum computing"
# Route through OpenRouter (model discovery, transparent pricing)
lm --gateway openrouter -m gpt-4o "write a function"
# Direct provider access (default, supports reasoning models)
lm --gateway direct "analyze this data"Gateway Benefits:
- Vercel: $5/month free credits, no token markup, <20ms latency
- OpenRouter: Model discovery, pricing transparency, bring-your-own-key
- Direct: Full provider feature support, reasoning models, lowest latency
StreamLM can be configured via config file (~/.streamlm/config.yaml), environment variables, or CLI flags:
# Interactive setup wizard
lm config setup
# Set default gateway
lm config set gateway.default vercel
# Configure gateway API keys
lm config set gateway.vercel.api_key sk-your-ai-gateway-key
# Set default model
lm config set models.default gpt-4o
# View current configuration
lm config get
# Validate configuration and API keys
lm config validate
# List available gateways
lm config list-gatewaysConfiguration Priority (highest to lowest):
- CLI flags (
--gateway,--model) - Environment variables (
STREAMLM_GATEWAY, provider API keys) - Config file (
~/.streamlm/config.yaml) - Defaults (direct gateway, gemini-2.5-flash model)
Define shortcuts for your favorite models in the config:
# ~/.streamlm/config.yaml
models:
aliases:
gpt: "gpt-4o"
claude: "claude-3-5-sonnet"
fast: "gemini/gemini-2.5-flash"
smart: "gpt-4o"Then use them:
lm -m fast "quick question" # Uses gemini-2.5-flash
lm -m smart "complex analysis" # Uses gpt-4oStreamLM includes beautiful built-in markdown formatting, but you can also output raw markdown for piping to other tools:
# Output raw markdown without Rich formatting
lm --md "explain machine learning" > output.md
# Pipe to your favorite markdown formatter (like glow)
lm --md "write a Python tutorial" | glow
# Use with other markdown tools
lm --raw "create documentation" | pandoc -f markdown -t htmlStreamLM provides access to various Large Language Models including:
- OpenAI: GPT-4o, o1, o3-mini, GPT-4o-mini
- Anthropic: Claude-3-7-sonnet, Claude-3-5-sonnet, Claude-3-5-haiku
- Google: Gemini-2.5-flash, Gemini-2.5-pro, Gemini-2.0-flash-thinking
- DeepSeek: DeepSeek-R1, DeepSeek-V3
- xAI: Grok-4, Grok-3-beta, Grok-3-mini-beta
- Local models: Via Ollama (Llama3.3, Qwen2.5, DeepSeek-Coder, etc.)
--model/-m: Choose the LLM model (or use alias from config)--gateway/-g: Route through gateway (direct, vercel, openrouter)--image/-i: Include image files for vision models--context/-c: Add context from a file--max-tokens/-t: Set maximum response length--temperature/-temp: Control response creativity (0.0-1.0)--think: Show reasoning process (reasoning models, direct gateway only)--session/-s: Session ID to continue conversation--session-name: Name for a new session (only when creating)--debug/-d: Enable debug mode--raw/--md: Output raw markdown without Rich formatting
StreamLM supports conversation sessions to maintain context across multiple messages:
# Create a new session or continue an existing one
lm --session my-project "How do I implement authentication?"
lm --session my-project "Can you show me an example?" # Continues with context
# Name your session when creating it
lm --session dev-2025 --session-name "Development Session" "Let's start coding"
# List all sessions
lm sessions --list
# Show session details and conversation history
lm sessions --show my-project
# Export session to JSON
lm sessions --export my-project > session.json
# Clear messages from a session (keeps session metadata)
lm sessions --clear my-project
# Delete a session completely
lm sessions --delete my-projectSession Features:
- Automatic conversation history - context is maintained across messages
- Token usage tracking per session
- Local-first storage using libSQL (SQLite compatible)
- Optional remote sync with Turso (not required)
- Export/import sessions for backup or sharing
- Metadata support for images and context files
lm config setup: Interactive configuration wizardlm config get [key]: Get configuration valuelm config set <key> <value>: Set configuration valuelm config validate: Validate configuration and API keyslm config list-gateways: Show available gateways and their status
- 🎨 Beautiful markdown-formatted responses
- 💬 Conversation sessions with persistent history
- 🌐 Gateway routing (Vercel AI Gateway, OpenRouter, or direct)
- ⚙️ Flexible configuration (config file, env vars, CLI flags)
- 🔑 Model aliases for quick access to favorite models
- 🖼️ Image input support for compatible models
- 📁 Context file support
- 🧠 Reasoning model support (DeepSeek, OpenAI o1, etc.)
- 📊 Token usage tracking per session
- 💾 Local-first database storage (no cloud required)
- 🔧 Extensive model support across providers
- ⚡ Fast and lightweight
- 🛠️ Easy configuration management
MIT License - see LICENSE file for details.
# Clone the repository
git clone https://github.com/jeffmylife/streamlm.git
cd streamlm
# Install with dev dependencies
uv pip install -e ".[dev]"All tests use uv run for consistency:
# Run all tests
uv run pytest tests/ -v
# Run with coverage
uv run pytest tests/ -v --cov=src --cov-report=term-missing
# Run specific test types
uv run pytest tests/test_cli.py -v # Unit tests only
uv run pytest tests/test_integration.py -v # Integration tests only# Make your changes
uv version --bump patch
git add .
git commit -m "feat: your changes"
git push
# Create GitHub release (this triggers everything automatically)
gh release create v0.1.11 --generate-notes