Multi-Provider Agent

One waxell.init() call, and all your LLM providers are automatically traced -- zero manual instrumentation code needed.

Environment variables

This example requires OPENAI_API_KEY, ANTHROPIC_API_KEY, GROQ_API_KEY, WAXELL_API_KEY, and WAXELL_API_URL.

import asyncio
import waxell_observe as waxell
from waxell_observe import generate_session_id

# waxell.init() auto-instruments all three providers -- zero manual code needed
waxell.init()

from openai import OpenAI
import anthropic
from groq import Groq

openai_client = OpenAI()
anthropic_client = anthropic.Anthropic()
groq_client = Groq()


@waxell.observe(agent_name="multi-provider")
async def multi_provider_analysis(topic: str) -> dict:
    session = generate_session_id()
    waxell.metadata("providers_used", ["openai", "anthropic", "groq"])
    waxell.tag("task", "multi-perspective-analysis")

    # OpenAI -- technical perspective
    waxell.tag("provider", "openai")
    openai_resp = openai_client.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": f"Technical perspective on: {topic}"}],
    )
    waxell.score("response_quality", 0.9, comment="openai technical")

    # Anthropic -- creative perspective
    waxell.tag("provider", "anthropic")
    anthropic_resp = anthropic_client.messages.create(
        model="claude-sonnet-4-20250514",
        max_tokens=300,
        messages=[{"role": "user", "content": f"Creative perspective on: {topic}"}],
    )
    waxell.score("response_quality", 0.85, comment="anthropic creative")

    # Groq -- fast inference summary
    waxell.tag("provider", "groq")
    groq_resp = groq_client.chat.completions.create(
        model="llama-3.3-70b-versatile",
        messages=[{"role": "user", "content": f"Concise summary of: {topic}"}],
    )
    waxell.score("response_quality", 0.88, comment="groq summary")

    return {
        "technical": openai_resp.choices[0].message.content,
        "creative": anthropic_resp.content[0].text,
        "summary": groq_resp.choices[0].message.content,
    }


asyncio.run(multi_provider_analysis("The future of AI agents"))

What this demonstrates

• Zero-effort auto-instrumentation -- waxell.init() patches all three providers before import; every LLM call is captured automatically.
• @observe decorator -- creates a named trace with automatic lifecycle management.
• Top-level convenience functions -- waxell.tag(), waxell.score(), and waxell.metadata() enrich the trace without needing a context object.
• Groq integration -- Groq uses the OpenAI-compatible SDK, so auto-instrumentation works out of the box.
• Unified observability -- one dashboard view across all providers with per-call model attribution.
• Session correlation -- generate_session_id() ties related runs together for analysis.

Run it

export OPENAI_API_KEY="sk-..."
export ANTHROPIC_API_KEY="sk-ant-..."
export GROQ_API_KEY="gsk_..."
export WAXELL_API_KEY="your-waxell-api-key"
export WAXELL_API_URL="https://api.waxell.ai"

python multi_provider.py