Anthropic OpenTelemetry Integration

Overview

This integration provides support for using OpenTelemetry with the Anthropic framework. It enables tracing and monitoring of applications built with Anthropic.

Installation

1. Install traceAI Anthropic

pip install traceAI-anthropic

Set Environment Variables

Set up your environment variables to authenticate with FutureAGI

import os

os.environ["FI_API_KEY"] = FI_API_KEY
os.environ["FI_SECRET_KEY"] = FI_SECRET_KEY
os.environ["ANTHROPIC_API_KEY"] = ANTHROPIC_API_KEY

Quickstart

Register Tracer Provider

Set up the trace provider to establish the observability pipeline. The trace provider:

from fi_instrumentation import register
from fi_instrumentation.fi_types import ProjectType

trace_provider = register(
    project_type=ProjectType.OBSERVE,
    project_name="anthropic_app"
)

Configure Anthropic Instrumentation

Instrument the Anthropic client to enable telemetry collection. This step ensures that all interactions with the Anthropic SDK are tracked and monitored.

from traceai_anthropic import AnthropicInstrumentor

AnthropicInstrumentor().instrument(tracer_provider=trace_provider)

Anthropic SDK native tracing

Anthropic SDK 1.14.0 and later can trace Messages calls themselves. To send those spans to Future AGI without this instrumentor, pass the provider to the client:

client = anthropic.Anthropic(open_telemetry={"tracer_provider": trace_provider})

The SDK's spans include model, timing and token usage, and include prompts and responses only with "traces": {"content_mode": "content"}. TraceConfig, FI_HIDE_* settings and the using_session / using_attributes context (session, user, metadata, tags) don't apply to them.

When AnthropicInstrumentor is installed and the SDK's spans would reach the same tracer provider (because you pass it to the client, or because it is the global provider), the instrumentor records messages.create (sync and async) and sync messages.count_tokens calls itself and turns the SDK's span off for each of those calls, so each call is recorded once with the Future AGI context and privacy controls. Your client keeps its own open_telemetry option. To skip this per-call step, create the client with open_telemetry=False. messages.stream() and async messages.count_tokens are not wrapped by the instrumentor, so they keep the SDK's span.

Create Anthropic Components

Set up your Anthropic client with built-in observability.

client = anthropic.Anthropic()

response = client.messages.create(
    model="claude-3-5-sonnet-20240620",
    max_tokens=1024,
    tools=[
        {
            "name": "get_weather",
            "description": "Get the current weather in a given location",
            "input_schema": {
                "type": "object",
                "properties": {
                    "location": {
                        "type": "string",
                        "description": "The city and state, e.g. San Francisco, CA",
                    },
                    "unit": {
                        "type": "string",
                        "enum": ["celsius", "fahrenheit"],
                        "description": "The unit of temperature, either 'celsius' or 'fahrenheit'",
                    },
                },
                "required": ["location"],
            },
        },
        {
            "name": "get_time",
            "description": "Get the current time in a given time zone",
            "input_schema": {
                "type": "object",
                "properties": {
                    "timezone": {
                        "type": "string",
                        "description": "The IANA time zone name, e.g. America/Los_Angeles",
                    }
                },
                "required": ["timezone"],
            },
        },
    ],
    messages=[
        {
            "role": "user",
            "content": "What is the weather like right now in New York?"
            " Also what time is it there? Use necessary tools simultaneously.",
        }
    ],
)
print(response)