VertexAI | Sample Single Agent App

instrumentation-genai/opentelemetry-instrumentation-vertexai/examples/single-agent/.env.example

@@ -0,0 +1,5 @@
+GOOGLE_CLOUD_PROJECT=your-project-id
+GOOGLE_CLOUD_LOCATION=us-central1
+OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:4317
+OTEL_SERVICE_NAME=vertexai-single-agent
+OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT=SPAN_AND_EVENT

instrumentation-genai/opentelemetry-instrumentation-vertexai/examples/single-agent/main.py

@@ -0,0 +1,218 @@
+# pylint: skip-file
+"""VertexAI Single-Agent Example.
+
+Demonstrates a ReAct-style agentic loop using the native VertexAI SDK
+(GenerativeModel + tool calling) with automatic OpenTelemetry instrumentation
+via VertexAIInstrumentor.
+
+Run modes:
+1. Default (no CLI args): queries "What is the weather in San Francisco?" and exits.
+2. CLI mode: python main.py --city "Paris"
+
+Required environment variables:
+    GOOGLE_CLOUD_PROJECT   - GCP project ID
+    GOOGLE_CLOUD_LOCATION  - GCP region (default: us-central1)
+    OTEL_EXPORTER_OTLP_ENDPOINT - OTLP collector endpoint (default: http://localhost:4317)
+
+Optional:
+    GOOGLE_APPLICATION_CREDENTIALS - path to service-account JSON for ADC
+"""
+
+import argparse
+import os
+
+import requests
+import vertexai
+from dotenv import load_dotenv
+from vertexai.generative_models import (
+    FunctionDeclaration,
+    GenerativeModel,
+    Part,
+    Tool,
+)
+
+# NOTE: OpenTelemetry Python Logs and Events APIs are in beta
+from opentelemetry import _logs, trace
+from opentelemetry.exporter.otlp.proto.grpc._log_exporter import (
+    OTLPLogExporter,
+)
+from opentelemetry.exporter.otlp.proto.grpc.trace_exporter import (
+    OTLPSpanExporter,
+)
+from opentelemetry.instrumentation.vertexai import VertexAIInstrumentor
+from opentelemetry.sdk._logs import LoggerProvider
+from opentelemetry.sdk._logs.export import BatchLogRecordProcessor
+from opentelemetry.sdk.trace import TracerProvider
+from opentelemetry.sdk.trace.export import BatchSpanProcessor
+
+load_dotenv()
+
+# --- OTEL environment defaults ---
+os.environ.setdefault("OTEL_SERVICE_NAME", "vertexai-single-agent")
+os.environ.setdefault(
+    "OTEL_SEMCONV_STABILITY_OPT_IN", "gen_ai_latest_experimental"
+)
+os.environ.setdefault(
+    "OTEL_INSTRUMENTATION_GENAI_CAPTURE_MESSAGE_CONTENT", "SPAN_AND_EVENT"
+)
+
+# --- Configure tracing ---
+trace.set_tracer_provider(TracerProvider())
+trace.get_tracer_provider().add_span_processor(
+    BatchSpanProcessor(OTLPSpanExporter())
+)
+
+# --- Configure logging / events ---
+_logs.set_logger_provider(LoggerProvider())
+_logs.get_logger_provider().add_log_record_processor(
+    BatchLogRecordProcessor(OTLPLogExporter())
+)
+
+# --- Instrument VertexAI ---
+VertexAIInstrumentor().instrument()
+
+
+# ---------------------------------------------------------------------------
+# Tool implementation — uses the free Open-Meteo API (no key required)
+# ---------------------------------------------------------------------------
+
+
+def get_weather(location: str) -> dict:
+    """Get current weather for a city using the Open-Meteo geocoding and weather APIs."""
+    geocode_url = "https://geocoding-api.open-meteo.com/v1/search"
+    geo_resp = requests.get(
+        geocode_url,
+        params={
+            "name": location,
+            "count": 1,
+            "language": "en",
+            "format": "json",
+        },
+        timeout=10,
+    )
+    geo_resp.raise_for_status()
+    geo_data = geo_resp.json()
+
+    results = geo_data.get("results")
+    if not results:
+        return {"error": f"Location '{location}' not found"}
+
+    lat = results[0]["latitude"]
+    lon = results[0]["longitude"]
+    name = results[0].get("name", location)
+
+    weather_url = "https://api.open-meteo.com/v1/forecast"
+    weather_resp = requests.get(
+        weather_url,
+        params={
+            "latitude": lat,
+            "longitude": lon,
+            "current_weather": True,
+            "temperature_unit": "celsius",
+        },
+        timeout=10,
+    )
+    weather_resp.raise_for_status()
+    weather_data = weather_resp.json()
+
+    current = weather_data.get("current_weather", {})
+    return {
+        "location": name,
+        "temperature_celsius": current.get("temperature"),
+        "wind_speed_kmh": current.get("windspeed"),
+        "weather_code": current.get("weathercode"),
+    }
+
+
+# Map tool name → callable for dispatch
+_TOOLS = {"get_weather": get_weather}
+
+# FunctionDeclaration for the model
+_get_weather_decl = FunctionDeclaration(
+    name="get_weather",
+    description="Get the current weather for a given city or location.",
+    parameters={
+        "type": "object",
+        "properties": {
+            "location": {
+                "type": "string",
+                "description": "The city or location to get weather for, e.g. 'San Francisco'",
+            }
+        },
+        "required": ["location"],
+    },
+)
+
+_weather_tool = Tool(function_declarations=[_get_weather_decl])
+
+
+# ---------------------------------------------------------------------------
+# Agent loop
+# ---------------------------------------------------------------------------
+
+
+def run_agent(query: str) -> str:
+    """Run a single-turn agentic query against Gemini with tool calling."""
+    project = os.environ.get("GOOGLE_CLOUD_PROJECT")
+    location = os.environ.get("GOOGLE_CLOUD_LOCATION", "us-central1")
+
+    vertexai.init(project=project, location=location)
+
+    model = GenerativeModel(
+        "gemini-2.5-flash-lite",
+        tools=[_weather_tool],
+        system_instruction=(
+            "You are a helpful weather assistant. "
+            "Use the get_weather tool to fetch real weather data when asked about weather. "
+            "After presenting the data, add a short commentary about the conditions."
+        ),
+    )
+
+    chat = model.start_chat()
+    response = chat.send_message(query)
+
+    # ReAct loop: keep dispatching tool calls until the model produces a text reply
+    while response.candidates and response.candidates[0].function_calls:
+        fc = response.candidates[0].function_calls[0]
+        tool_fn = _TOOLS.get(fc.name)
+        if tool_fn is None:
+            tool_result = {"error": f"Unknown tool: {fc.name}"}
+        else:
+            try:
+                tool_result = tool_fn(**{k: v for k, v in fc.args.items()})
+            except Exception as exc:  # noqa: BLE001
+                tool_result = {"error": str(exc)}
+
+        response = chat.send_message(
+            Part.from_function_response(
+                name=fc.name, response={"output": tool_result}
+            )
+        )
+
+    return response.text
+
+
+# ---------------------------------------------------------------------------
+# Entry point
+# ---------------------------------------------------------------------------
+
+
+def main():
+    parser = argparse.ArgumentParser(
+        description="VertexAI single-agent weather demo with OpenTelemetry instrumentation"
+    )
+    parser.add_argument(
+        "--city",
+        default="San Francisco",
+        help="City to query weather for (default: San Francisco)",
+    )
+    args = parser.parse_args()
+
+    query = f"What is the weather in {args.city}?"
+    print(f"Query: {query}\n")
+    result = run_agent(query)
+    print(result)
+
+
+if __name__ == "__main__":
+    main()

instrumentation-genai/opentelemetry-instrumentation-vertexai/examples/single-agent/requirements.txt

@@ -0,0 +1,8 @@
+google-cloud-aiplatform>=1.64
+
+opentelemetry-sdk
+opentelemetry-exporter-otlp-proto-grpc
+opentelemetry-instrumentation-vertexai
+
+python-dotenv
+requests