Anthropic Integration

Automatic tracing for Anthropic Claude API calls.

Quick Start

Prereq: set TRACIUM_API_KEY (see Installation).

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
import tracium

# Enable auto-instrumentation
tracium.trace()

# Import clients after enabling tracing
import anthropic

# Use Anthropic normally - all calls are traced
client = anthropic.Anthropic()

message = client.messages.create(
    model="claude-3-opus-20240229",
    max_tokens=1024,
    messages=[
        {"role": "user", "content": "Hello, Claude!"}
    ]
)

What Gets Captured

Input messages - The messages array sent to Claude
System prompt - If provided
Model - claude-3-opus, claude-3-sonnet, etc.
Output - The response content
Token usage - Input and output tokens
Latency - API call duration
Stop reason - Why generation stopped

Streaming

1
2
3
4
5
6
7
8
9
10
11
12
13
14
import tracium

tracium.trace()
import anthropic
client = anthropic.Anthropic()

# Streaming is fully supported
with client.messages.stream(
    model="claude-3-sonnet-20240229",
    max_tokens=1024,
    messages=[{"role": "user", "content": "Write a story."}]
) as stream:
    for text in stream.text_stream:
        print(text, end="", flush=True)

Async Support

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
import tracium
import asyncio

tracium.trace()
import anthropic
client = anthropic.AsyncAnthropic()

async def chat(prompt: str) -> str:
    message = await client.messages.create(
        model="claude-3-sonnet-20240229",
        max_tokens=1024,
        messages=[{"role": "user", "content": prompt}]
    )
    return message.content[0].text

result = asyncio.run(chat("Explain quantum computing."))

Tool Use

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
import tracium

tracium.trace()
import anthropic
client = anthropic.Anthropic()

# Tool definitions
tools = [
    {
        "name": "get_weather",
        "description": "Get the current weather",
        "input_schema": {
            "type": "object",
            "properties": {
                "location": {
                    "type": "string",
                    "description": "City name"
                }
            },
            "required": ["location"]
        }
    }
]

# Tool use is captured in the trace
message = client.messages.create(
    model="claude-3-sonnet-20240229",
    max_tokens=1024,
    tools=tools,
    messages=[
        {"role": "user", "content": "What's the weather in Paris?"}
    ]
)

Multi-Turn Conversations

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
import tracium

client = tracium.init()
import anthropic
anthropic_client = anthropic.Anthropic()

with client.agent_trace(agent_name="claude-chat") as trace:
    messages = []
    
    # First turn
    messages.append({"role": "user", "content": "Hi, I'm learning Python."})
    response = anthropic_client.messages.create(
        model="claude-3-sonnet-20240229",
        max_tokens=1024,
        messages=messages
    )
    messages.append({"role": "assistant", "content": response.content[0].text})
    
    # Second turn - all calls are linked in the same trace
    messages.append({"role": "user", "content": "Can you show me a loop example?"})
    response = anthropic_client.messages.create(
        model="claude-3-sonnet-20240229",
        max_tokens=1024,
        messages=messages
    )

← OpenAI Google Gemini →