---
title: "Python"
description: "Using llms-sdk from Python."
---

> Documentation Index
> Fetch the complete documentation index at: https://llms-sdk.cc/llms.txt
> Use this file to discover all available pages before exploring further.

# Python

Python bindings for the `llms-sdk` Rust crate, built with PyO3. Requires Python 3.10+.

## Installation

```bash
pip install llms-sdk-py
# or, with uv
uv add llms-sdk-py
```

## Quick start

```python
import asyncio
import llms_sdk_py as llm

async def main():
request = llm.LLMRequest(
    model="gpt-5.4-mini",
    api_key="sk-...",
    messages=[
        llm.Message("user", [llm.TextPart("Hello, world!")])
    ],
    stream=False,
)

client = llm.LLM()
response = await client.respond(request)
print(response.message.content[0].text)

asyncio.run(main())
```

## Streaming

Set `stream=True` and iterate over the async iterator:

```python
import asyncio
import llms_sdk_py as llm

async def main():
request = llm.LLMRequest(
    model="gpt-5.4-mini",
    api_key="sk-...",
    messages=[llm.Message("user", [llm.TextPart("Count to 5")])],
    stream=True,
)

client = llm.LLM()
async for part in client.stream_response(request):
    if str(part.type) == "text":
        print(part.text.text_delta, end="", flush=True)
    elif str(part.type) == "end":
        print("\nDone!")

asyncio.run(main())
```

## Multimodal input

### Image

```python
import llms_sdk_py as llm

image = llm.ImagePart("path/to/photo.jpg")      # or a URL string
message = llm.Message("user", [
llm.TextPart("Describe this image."),
image,
])
```

### Audio (OpenAI only)

```python
import llms_sdk_py as llm

audio = llm.AudioPart("path/to/audio.wav")       # or raw bytes
message = llm.Message("user", [
llm.TextPart("Transcribe this audio."),
audio,
])
```

### Document (Anthropic only)

```python
import llms_sdk_py as llm

doc = llm.DocumentPart("path/to/file.pdf")       # or raw bytes
message = llm.Message("user", [
llm.TextPart("Summarize this document."),
doc,
])
```

## Tool use

```python
import llms_sdk_py as llm

tool = llm.Tool(
name="get_weather",
description="Return weather for a city.",
parameters_dict={
    "type": "object",
    "properties": {
        "city": {"type": "string"}
    },
    "required": ["city"]
},
)

request = llm.LLMRequest(
model="gpt-5.4-mini",
api_key="sk-...",
messages=[llm.Message("user", [llm.TextPart("What's the weather in Paris?")])],
stream=False,
tools=[tool],
tool_choice="auto",
)
```

## Structured output

```python
import llms_sdk_py as llm

output_format = llm.OutputFormat(
name="capital",
description="Country capital",
schema_dict={
    "type": "object",
    "properties": {
        "country": {"type": "string"},
        "capital": {"type": "string"}
    },
    "required": ["country", "capital"]
},
)

request = llm.LLMRequest(
model="gpt-5.4-mini",
api_key="sk-...",
messages=[llm.Message("user", [llm.TextPart("France")])],
stream=False,
output_format=output_format,
)
```

## Retry policy

```python
import llms_sdk_py as llm

# intervals are in milliseconds
policy = llm.RetryPolicy(max_retries=5, min_retry_interval=1000)
client = llm.LLM(retry_policy=policy)
```

Source: https://llms-sdk.cc/python/index.mdx