autourgos-openaichat 1.0.0

Autourgos LLM wrapper for the OpenAI Chat Completions API

autourgos-openaichat

LLM wrapper for the OpenAI Chat Completions API, part of the Autourgos framework.

Fully self-contained — no autourgos-core dependency required.

Install

pip install autourgos-openaichat

Quick start

from autourgos_openaichat import OpenAIChatModel

llm = OpenAIChatModel(model="gpt-4o", api_key="sk-...")

# Sync
reply = llm.invoke("What is the capital of France?")
print(reply)  # "Paris"

# Async
reply = await llm.ainvoke("Translate 'hello' to Spanish.")

# Streaming
for chunk in llm.stream("Tell me a joke."):
    print(chunk, end="", flush=True)

# Async streaming
async for chunk in llm.astream("Explain AI."):
    print(chunk, end="", flush=True)

# Batch
replies = llm.batch_invoke(["Q1", "Q2", "Q3"])

# Async batch (concurrent)
replies = await llm.abatch_invoke(["Q1", "Q2", "Q3"])

Multi-modal (vision)

llm = OpenAIChatModel(model="gpt-4o")
reply = llm.invoke("What is in this image?", files=["photo.jpg"])

Structured output

from pydantic import BaseModel

class Answer(BaseModel):
    capital: str
    country: str

llm = OpenAIChatModel(model="gpt-4o", response_schema=Answer)
result = llm.invoke("Capital of France?")
# result["response"] contains the JSON string

Native tool-calling

tools = [
    {
        "name": "get_weather",
        "description": "Get current weather for a city",
        "parameters": {
            "type": "object",
            "properties": {"city": {"type": "string"}},
            "required": ["city"],
        },
    }
]

response = llm.invoke_with_tools("What's the weather in Paris?", tools)
if response.has_tool_calls:
    for call in response.tool_calls:
        print(call.name, call.arguments)

Prompt templates

llm = OpenAIChatModel(
    model="gpt-4o",
    prompt_template="Translate the following to {language}: {text}",
)
reply = llm.invoke(prompt_variables={"language": "French", "text": "Hello"})

System instruction

llm = OpenAIChatModel(
    model="gpt-4o",
    system_instruction="You are a concise assistant. Reply in one sentence.",
)

Context manager

with OpenAIChatModel(model="gpt-4o") as llm:
    reply = llm.invoke("Hello")

Constructor options

Parameter	Type	Default	Description
model	str	required	Model name, e.g. "gpt-4o"
api_key	str	env OPENAI_API_KEY	OpenAI API key
base_url	str	env OPENAI_BASE_URL	Custom API base URL
organization	str	None	OpenAI org ID
project	str	None	OpenAI project ID
system_instruction	str	None	System prompt
prompt_template	str	None	Template with {placeholders}
temperature	float	None	Sampling temperature (0–2)
top_p	float	None	Nucleus sampling (0–1)
max_tokens	int	None	Max output tokens
response_schema	Pydantic model / dict	None	Structured JSON output
response_mime_type	str	None	"application/json" for JSON mode
structured_output	bool	False	Return metadata dict instead of string
streaming	bool	False	Stream internally and join
max_retries	int	3	Retry attempts on API errors
timeout	float	60.0	Request timeout in seconds
backoff_factor	float	0.5	Exponential back-off base
input_pricing	float	None	USD per 1M input tokens
output_pricing	float	None	USD per 1M output tokens
circuit_failure_threshold	int	5	Failures before circuit opens
circuit_cooldown_time	float	30.0	Seconds circuit stays open

License

MIT