verdictlens 0.2.0

Open-source observability for AI agents — hierarchical tracing, auto-instrumentation, blame analysis.

verdictlens (Python SDK)

Python client for VerdictLens — hierarchical tracing for AI agent workloads with auto-instrumentation.

See the repository root for the full platform quickstart.

Install

pip install verdictlens

Quickstart

from verdictlens import configure, trace, wrap_openai
from openai import OpenAI

configure(base_url="http://localhost:8000")
client = wrap_openai(OpenAI())

@trace(name="my_pipeline", span_type="chain")
def my_pipeline(query: str) -> str:
    context = retrieve(query)
    return generate(context, query)

@trace(name="retrieve", span_type="retrieval")
def retrieve(query: str) -> list:
    return vector_db.search(query)

@trace(name="generate", span_type="agent")
def generate(context: list, query: str) -> str:
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": f"{context}\n{query}"}]
    )
    return response.choices[0].message.content

This produces a hierarchical span tree:

my_pipeline (chain)
├── retrieve (retrieval)
└── generate (agent)
    └── openai.chat.completions.create(gpt-4o-mini) (llm)  ← auto-traced

Key Features

• Nested @trace — child spans auto-attach to parent via contextvars
• wrap_openai(client) — auto-trace OpenAI chat completions with zero code
• wrap_anthropic(client) — auto-trace Anthropic messages with zero code
• wrap_google(client) — auto-trace Google Gemini generate_content with zero code
• asyncio.gather safe — parallel branches maintain correct parent context
• Non-blocking — async transport with disk-backed offline queue