Metadata-Version: 2.5
Name: trayce
Version: 0.1.0
Summary: Python SDK for Trayce, observability and evals for LLM agents
Project-URL: Homepage, https://github.com/MannyNQ/Trayce
Project-URL: Repository, https://github.com/MannyNQ/Trayce
Project-URL: Issues, https://github.com/MannyNQ/Trayce/issues
License-Expression: MIT
License-File: LICENSE
Keywords: agents,evals,llm,observability,opentelemetry,tracing
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.12
Classifier: Topic :: Software Development :: Testing
Classifier: Topic :: System :: Monitoring
Requires-Python: >=3.12
Requires-Dist: pydantic>=2.7
Provides-Extra: eval
Requires-Dist: jsonschema>=4.20; extra == 'eval'
Requires-Dist: pyyaml>=6.0; extra == 'eval'
Requires-Dist: regex>=2024.11.6; extra == 'eval'
Provides-Extra: otel
Requires-Dist: opentelemetry-sdk>=1.30; extra == 'otel'
Description-Content-Type: text/markdown

# trayce

Python SDK for [Trayce](https://github.com/MannyNQ/Trayce), open-source observability and evals for LLM agents. It traces agent runs into a Trayce server and provides the `trayce` CLI (`trayce eval`, `trayce replay`).

```python
from trayce import instrument_anthropic, trace

instrument_anthropic()  # each Anthropic model call becomes an LLM span


@trace(name="support_agent", kind="agent")
def run(question: str) -> str: ...
```

The SDK reads `TRAYCE_API_KEY` and `TRAYCE_ENDPOINT`. Extras: `trayce[eval]` for `trayce eval`, `trayce[otel]` for the OpenTelemetry bridge. The `llm_judge` scorer currently uses Anthropic models only.

To run a Trayce server on this machine (needs Docker):

```bash
trayce up     # prints the web UI address, a setup token, and the settings for your agent
trayce down   # stops it; --delete-data also deletes its data
```

See the [repository README](https://github.com/MannyNQ/Trayce#readme) for more.
