> ## Documentation Index
> Fetch the complete documentation index at: https://docs.trulayer.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Python SDK overview

> The trulayer Python package — install, authenticate, instrument.

The Python SDK (`trulayer` on PyPI) is the primary way to send traces to TruLayer from Python code. It supports manual tracing, auto-instrumentation of popular LLM clients, and async/sync usage patterns.

## Install

```bash theme={null}
pip install trulayer
```

Python 3.11+ required.

## Minimal usage

```python theme={null}
import os
import trulayer

trulayer.init(
    api_key=os.environ["TRULAYER_API_KEY"],
    project_name="my-app",
)

with trulayer.trace("do_work") as trace:
    trace.set_input({"prompt": "hello"})
    # ... your code ...
    trace.set_output({"response": "world"})
```

## What's included

| Import                                                        | Purpose                                       |
| ------------------------------------------------------------- | --------------------------------------------- |
| `trulayer.init(...)`                                          | Initialise the global client                  |
| `trulayer.trace(name)`                                        | Context manager for a trace                   |
| `trulayer.atrace(name)`                                       | Async variant                                 |
| `trulayer.current_trace()`                                    | Get the active trace from async-local context |
| `trulayer.instrument_openai(client)`                          | Auto-patch an OpenAI client                   |
| `trulayer.instrument_anthropic(client)`                       | Auto-patch an Anthropic client                |
| `trulayer.instrument_langchain()`                             | Register a LangChain callback handler         |
| `trulayer.TruLayerClient`                                     | Class for explicit multi-instance use         |
| `trulayer.TraceData`, `SpanData`, `EventData`, `FeedbackData` | Pydantic models                               |

Full signatures: [Reference](/sdks/python/reference).

## Where to go next

<CardGroup cols={2}>
  <Card title="Tutorial" icon="graduation-cap" href="/sdks/python/tutorial">
    Step-by-step: manual tracing → auto-instrumentation → async → feedback → production.
  </Card>

  <Card title="Configuration" icon="sliders" href="/sdks/python/configuration">
    Batch size, flush interval, sampling, PII scrubbing, debug mode.
  </Card>

  <Card title="Reference" icon="book" href="/sdks/python/reference">
    Every public export, with signatures and examples.
  </Card>

  <Card title="Concepts" icon="diagram-project" href="/concepts/overview">
    Understand the data model before going deep.
  </Card>
</CardGroup>