| 1 | # SPDX-License-Identifier: Apache-2.0 |
| 2 | # SPDX-FileCopyrightText: Copyright contributors to the vLLM project |
| 3 | |
| 4 | import requests |
| 5 | from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import OTLPSpanExporter |
| 6 | from opentelemetry.sdk.trace import TracerProvider |
| 7 | from opentelemetry.sdk.trace.export import BatchSpanProcessor, ConsoleSpanExporter |
| 8 | from opentelemetry.trace import SpanKind, set_tracer_provider |
| 9 | from opentelemetry.trace.propagation.tracecontext import TraceContextTextMapPropagator |
| 10 | |
| 11 | trace_provider = TracerProvider() |
| 12 | set_tracer_provider(trace_provider) |
| 13 | |
| 14 | trace_provider.add_span_processor(BatchSpanProcessor(OTLPSpanExporter())) |
| 15 | trace_provider.add_span_processor(BatchSpanProcessor(ConsoleSpanExporter())) |
| 16 | |
| 17 | tracer = trace_provider.get_tracer("dummy-client") |
| 18 | |
| 19 | url = "http://localhost:8000/v1/completions" |
| 20 | with tracer.start_as_current_span("client-span", kind=SpanKind.CLIENT) as span: |
| 21 | prompt = "San Francisco is a" |
| 22 | span.set_attribute("prompt", prompt) |
| 23 | headers = {} |
| 24 | TraceContextTextMapPropagator().inject(headers) |
| 25 | payload = { |
| 26 | "model": "facebook/opt-125m", |
| 27 | "prompt": prompt, |
| 28 | "max_tokens": 10, |
| 29 | "n": 3, |
| 30 | "use_beam_search": "true", |
| 31 | "temperature": 0.0, |
| 32 | # "stream": True, |
| 33 | } |
| 34 | response = requests.post(url, headers=headers, json=payload) |