Easy SGR

A simplified interface for building SGR (Schema-Guided Reasoning) agents in a LangChain-like style.

Features

• 🎯 Simple API in the spirit of LangChain
• 🛠️ @tool decorator — turn functions into tools
• 🤖 ChatOpenAI — LLM wrapper (OpenAI, OpenRouter, etc.)
• ⚡ Sync and async — invoke() (no await) and ainvoke() (with await)
• 📐 Structured output — Pydantic schema via output_schema
• 🔧 Flexible agent configuration

Installation

pip install -e .

Or from PyPI (after publishing):

pip install easy-sgr

Quick start

1. Tools

The @tool decorator turns a function into an SGR tool:

from easy_sgr import tool

@tool
def add_numbers(a: int, b: int) -> int:
    """Adds two numbers together.

    Args:
        a: First number
        b: Second number
    """
    return a + b

@tool
async def get_weather(city: str) -> str:
    """Get weather for a city."""
    return f"Weather in {city}: Sunny"

2. LLM

from easy_sgr import ChatOpenAI

llm = ChatOpenAI(
    model="gpt-4o-mini",
    temperature=0,
    api_key="your-api-key",
    base_url="https://api.openai.com/v1",
)

3. Agent and run

from easy_sgr import create_agent

tools = [add_numbers, get_weather]
messages = [
    {"role": "system", "content": "You are a helpful assistant."},
]

agent = create_agent(llm=llm, tools=tools, messages=messages)

# Sync (no await)
result = agent.invoke({"input": "What is 10 + 5?"})
print(result["output"])

# Async (inside async code)
result = await agent.ainvoke({"input": "What is 10 + 5?"})
print(result["output"])

Full example

import os
from dotenv import load_dotenv
from easy_sgr import tool, ChatOpenAI, create_agent

load_dotenv()

@tool
def add_numbers(a: int, b: int) -> int:
    """Adds two numbers together."""
    return a + b

llm = ChatOpenAI(
    model=os.getenv("OPENAI_MODEL", "gpt-4o-mini"),
    temperature=0,
    api_key=os.getenv("OPENAI_API_KEY"),
    base_url=os.getenv("OPENAI_BASE_URL"),
)

agent = create_agent(
    llm=llm,
    tools=[add_numbers],
    messages=[{"role": "system", "content": "You are a helpful assistant."}],
)

result = agent.invoke({"input": "What is 10 + 5?"})
print(result["output"])

Structured output (output_schema)

Pass a Pydantic model as output_schema — the agent will return an instance of it:

from pydantic import BaseModel, Field
from easy_sgr import create_agent, tool, ChatOpenAI

class MathResult(BaseModel):
    sum_result: int = Field(description="Result of addition")
    product_result: int = Field(description="Result of multiplication")
    summary: str = Field(description="Brief summary")

@tool
def add_numbers(a: int, b: int) -> int:
    """Adds two numbers."""
    return a + b

@tool
def multiply_numbers(a: int, b: int) -> int:
    """Multiplies two numbers."""
    return a * b

agent = create_agent(
    llm=llm,
    tools=[add_numbers, multiply_numbers],
    messages=[{"role": "system", "content": "..."}],
    output_schema=MathResult,
)

result = agent.invoke({"input": "Calculate 10 + 5 and 10 * 5, then summarize."})
output = result["output"]  # MathResult
print(output.sum_result, output.product_result, output.summary)

Examples

See the examples/ folder:

• example.py — basic usage with invoke
• example_structured.py — structured output via output_schema