pi-ai

Python port of pi-mono: AI coding agent with unified LLM API.

Installation

pip install pi-ai

CLI Usage

# Basic usage (defaults to gpt-4o)
pi "Explain this codebase"

# Test without API key using dummy model
pi -m dummy "hello world"

# Specify a model
pi -m claude-sonnet-4 "Refactor this function"

# List available models
pi --list-models

# Use a session
pi -s my-session "Continue our discussion"

# Interactive
pi --mode interactive -m dummy 

Available Models:

• dummy - For testing (no API key required)
• gpt-4o, gpt-4o-mini - OpenAI
• claude-sonnet-4, claude-3-5-sonnet - Anthropic
• gemini-2.0-flash - Google

Environment Variables:

• OPENAI_API_KEY - OpenAI API key
• ANTHROPIC_API_KEY - Anthropic API key
• GOOGLE_API_KEY - Google API key

Modules

pi.ai - LLM API Abstraction

Unified API for multiple LLM providers with streaming support.

from pi.ai import Model, stream, Context

# Create a model
model = Model(
    id="gpt-4o",
    name="GPT-4o",
    api="openai-completions",
    provider="openai",
    baseUrl="https://api.openai.com/v1",
)

# Stream responses
context = Context(messages=[{"role": "user", "content": "Hello"}])
async for event in stream(model, context):
    if event.type == "text_delta":
        print(event.delta, end="")

Supported Providers:

• OpenAI (Completions API)
• Anthropic (Messages API)
• Google (Generative AI)
• Amazon Bedrock

pi.agent - Agent Runtime

Agent loop implementing the ReAct (Reason-Act-Observe) pattern.

from pi.agent import Agent, tool
from pi.ai.models import resolve_model

# Define a tool
@tool
def read(file_path: str) -> str:
    """Read a file."""
    with open(file_path) as f:
        return f.read()

# Create and run agent
model = resolve_model("gpt-4o")
agent = Agent(model=model, tools=[read])

async for event in agent.prompt("Read the README.md file"):
    print(event)

Features:

• @tool decorator with automatic JSON Schema generation
• Tool registry for managing tools
• Steering messages (interrupt mid-run)
• Follow-up messages (continue after stop)
• Event-based subscription for UI updates
• Abort handling

Development

# Install dev dependencies
uv sync

# Run tests
uv run pytest

# Type check
uv run mypy src/

# Lint
uv run ruff check src/

License

MIT