Phoenix does not provide a native .NET client, but you can integrate your .NET AI application with Phoenix by using OpenTelemetry (OTEL) to manually instrument your code and export traces to Phoenix鈥檚 OTLP HTTP endpoint. You should use the OpenTelemetry .NET SDK, set the exporter endpoint to `http://<phoenix_host> :6006/v1/traces`, and follow the OpenInference semantic conventions for span attributes to ensure Phoenix can recognize your data. Evaluation metrics, however, can only be sent via the Python client at this time; .NET can only send traces, not evals or metrics directly. For embedding visualization, you must upload embeddings as datasets using the Python client, as this is not supported via .NET tracing alone.
For a working .NET example and further details, see the discussion and code snippets in <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1713792115.640359|this Slack thread>. For OpenInference semantic conventions, refer to <https://github.com/Arize-ai/openinference/blob/main/spec/semantic_conventions.md|this specification>.
Would you like a step-by-step example or code snippet for .NET integration?
Sources:
- <https://arize-ai.slack.com/archives/C04R3GXC8HK/p1713792115.640359|Slack thread: .NET integration with Phoenix>
