respan-instrumentation-vertexai

Respan instrumentation plugin for the Google Vertex AI Python SDK.

The package patches vertexai.generative_models.GenerativeModel and ChatSession generation methods and emits canonical Respan chat spans through the active OTEL pipeline. It captures sync calls, async calls, streamed responses, prompt and completion content, token usage, tool declarations, and model function calls.

Installation

pip install respan-ai respan-instrumentation-vertexai google-cloud-aiplatform

Usage

import vertexai
from respan import Respan
from respan_instrumentation_vertexai import VertexAIInstrumentor
from vertexai.generative_models import GenerativeModel

respan = Respan(instrumentations=[VertexAIInstrumentor()])

vertexai.init(project="your-gcp-project", location="us-central1")

model = GenerativeModel("gemini-2.0-flash")
response = model.generate_content("Say hello in three languages.")
print(response.text)
respan.flush()