Prosperus Python SDK

Instrument your LLM applications with traces, spans, evaluations, and metrics.

Install with:

pip install prosperus
# or
uv add prosperus

The distribution name and the Python import package are both prosperus.

Quick Start

from prosperus import Prosperus, llm, workflow

ph = Prosperus(
    api_key="ph-...",
    app_name="my-chatbot",
    endpoint="http://localhost:4100",  # use your local server during development
)
ph.enable()

@workflow
def handle_request(user_msg: str) -> str:
    return call_model(user_msg)

@llm(model_name="gpt-5.4", model_provider="openai")
def call_model(prompt: str) -> str:
    reply = f"Echo: {prompt}"
    Prosperus.annotate(
        input_data=[{"role": "user", "content": prompt}],
        output_data=[{"role": "assistant", "content": reply}],
        metrics={"input_tokens": 5, "output_tokens": 7},
    )
    return reply

print(handle_request("Hello"))
ph.shutdown()

Development

cd packages/sdk-python
uv sync
uv run pytest
uv run ruff check src/
uv run mypy src/

Configuration

What You Can Import

The top-level package re-exports the most common SDK entry points:

from prosperus import Prosperus, llm, workflow, tool, SpanKind, SpanContext