Logwick Python SDK

The official Python SDK for Logwick — audit logging for AI agents.

Installation

pip install logwick

Quick start

import logwick

logwick.init(api_key="sk-lw-your-key")

# Fire and forget — never blocks your code
logwick.fire({
    "agent": "gpt-4o",
    "action": "email_draft",
    "status": "success",
    "input": user_prompt,
    "output": result,
    "tokens": 312,
    "user": user_email,
})

OpenAI wrapper

Automatically logs input, output, tokens, and latency:

from logwick import LogwickClient

lw = LogwickClient(api_key="sk-lw-your-key")

result = lw.openai(
    lambda: client.chat.completions.create(model="gpt-4o", messages=messages),
    {"action": "email_draft", "user": user_email}
)
# result is the normal OpenAI response — nothing changes in your code

Anthropic / Claude wrapper

result = lw.anthropic(
    lambda: anthropic.messages.create(
        model="claude-3-5-sonnet-20241022",
        messages=messages,
        max_tokens=1024
    ),
    {"action": "document_review", "user": user_email}
)

Google Gemini wrapper

result = lw.gemini(
    lambda: model.generate_content(prompt),
    {"action": "data_analysis", "user": user_email}
)

LangChain integration

One handler logs every LLM call in your chain automatically:

handler = lw.langchain_handler(user="ops@acme.com")
chain = LLMChain(llm=llm, prompt=prompt, callbacks=[handler])
# Every call in the chain is now logged automatically

Client options

from logwick import LogwickClient

lw = LogwickClient(
    api_key="sk-lw-your-key",   # required
    silent=False,                # print warnings (default: True = silent)
    tags=["production"],         # default tags added to every log
)

Using the global client

import logwick

logwick.init(api_key="sk-lw-your-key", tags=["production"])

# Use anywhere in your codebase without passing the client around
logwick.fire({"agent": "gpt-4o", "action": "summarize", "status": "success", ...})

Get your API key

Sign up free at logwick.io — 5,000 logs/month free, no credit card required.