arbis-llmwrap 0.3.5

Decorator to wrap LLM calls for production use with flexible prompt binding.

arbis-llmwrap

Lightweight wrappers for LLM calls in Python:

• function wrapper: wrap_llm_call
• line wrapper: wrap_llm_line

This page intentionally documents usage only. Internal implementation details are not published.

Table of Contents

• Install
• Quick Start
• Function Wrapper (wrap_llm_call)
• Line Wrapper (wrap_llm_line)
• Common Config
• License

Install

pip install arbis-llmwrap

Quick Start

from llmwrap import wrap_llm_call

@wrap_llm_call(
    company_name="Example Co",
    project_name="Support",
    agent_name="quickstart_agent",
    secret_key="vt_live_xxxx",
    prompt_arg="prompt",
    max_tries=1,
)
def ask(prompt: str) -> str:
    return client.responses.create(model="gpt-4.1-mini", input=prompt).output_text

Function Wrapper (wrap_llm_call)

from llmwrap import wrap_llm_call

def has_tool_calls(raw) -> bool:
    try:
        return bool(raw.choices[0].message.tool_calls)
    except Exception:
        return False

@wrap_llm_call(
    company_name="Example Co",
    project_name="Agent Runtime",
    agent_name="function_wrapper_example",
    secret_key="vt_live_xxxx",
    prompt_arg="messages",
    passthrough_when=has_tool_calls,
    max_tries=1,
)
def run_turn(messages: list[dict]):
    return client.chat.completions.create(
        model="gpt-4.1-mini",
        messages=messages,
        tools=[...],
        tool_choice="auto",
        temperature=0,
    )

Line Wrapper (wrap_llm_line)

from llmwrap import wrap_llm_line

payload = {
    "model": "gpt-4.1-mini",
    "messages": [{"role": "user", "content": "Give one sentence with token TOKEN-42."}],
}

completion = wrap_llm_line(
    llm_call=lambda p: client.chat.completions.create(
        model=p["model"],
        messages=p["messages"],
        temperature=0,
    ),
    prompt=payload,
    prompt_json_pointer="/messages/0/content",
    company_name="Example Co",
    project_name="Runtime",
    agent_name="line_wrapper_example",
    secret_key="vt_live_xxxx",
    max_tries=1,
)

Common Config

• company_name, project_name, agent_name, secret_key
• max_tries (>= 1)
• optional: response_extractor, prompt_json_pointer, passthrough_when, response_answer_json_pointer, return_merger

License

MIT. See LICENSE.