vantra 0.1.3

Observability SDK for AI agents — traces, costs, and alerts in 3 lines of code.

Vantra Python SDK

Observability for AI agents — traces, costs, and alerts in 3 lines of code.

Install

pip install vantra

Quickstart

import vantra

vantra.init(
    api_key="van_live_...",
    project="my-agent"
)

@vantra.trace
def run_agent(message: str):
    return agent.run(message)  # your existing code — untouched

Every call to run_agent appears in your Vantra dashboard with timing, token usage, cost, and status.

OpenAI auto-patch

import vantra
import openai

vantra.init(api_key="van_live_...", project="my-agent")
client = openai.OpenAI()

@vantra.trace
def ask(question: str) -> str:
    # captured automatically — tokens, cost, latency
    response = client.chat.completions.create(
        model="gpt-4o",
        messages=[{"role": "user", "content": question}]
    )
    return response.choices[0].message.content

Anthropic auto-patch

import vantra
import anthropic

vantra.init(api_key="van_live_...", project="my-agent")
client = anthropic.Anthropic()

@vantra.trace
def ask(question: str) -> str:
    response = client.messages.create(
        model="claude-opus-4-5",
        max_tokens=1024,
        messages=[{"role": "user", "content": question}]
    )
    return response.content[0].text

Manual spans

@vantra.trace
def run_agent(message: str):
    with vantra.span("search_kb", kind="tool") as span:
        results = search(message)
        span.set_output({"results": results, "count": len(results)})

    with vantra.span("generate", kind="llm"):
        return llm.chat(message, context=results)

Options

Parameter	Type	Default	Description
api_key	str	required	Your van_live_... key from Settings
project	str	required	Project name shown in dashboard
patch_openai	bool	True	Auto-instrument OpenAI calls
patch_anthropic	bool	True	Auto-instrument Anthropic calls
enabled	bool	True	Set False to disable in tests

Links

• vantra.dev
• Docs
• Dashboard
• GitHub