LangPrompt

LangPrompt is a Python library for creating chat prompts with template support.

Some code come from https://github.com/cfahlgren1/observers，https://github.com/Mirascope/mirascope

Installation

pip install langprompt

Usage

from pydantic import BaseModel
from langprompt import TextOutputParser, Prompt
from langprompt.llms.openai import OpenAI
from langprompt.cache import SQLiteCache
from langprompt.store import DuckDBStore


class Input(BaseModel):
    text: str
    language: str = "Chinese"

prompt = Prompt[Input]("""
<|system|>
You are a professional translator. Please accurately translate the text while maintaining its original meaning and style.
<|end|>

<|user|>
Translate the following text into {{language}}: {{text}}
<|end|>
""")

if __name__ == "__main__":

    parser = TextOutputParser()
    provider = OpenAI(model="gpt-4o-mini", cache=SQLiteCache(), store=DuckDBStore(), query_per_second=0.2)
    inputs = [
        Input(text="Hello, how are you?", language="Chinese"),
        Input(text="Hello, how are you?", language="English"),
        Input(text="Hello, how are you?", language="Chinese"),
    ]

    messages = [prompt.parse(input) for input in inputs]
    responses = provider.batch(messages, batch_size=2, enable_retry=True)

    # 处理结果
    for i, response in enumerate(responses):
        print(f"\n--- Result {i+1} ---")
        print(f"Original: {inputs[i].text}")
        result = parser.parse(response)
        print(f"Translated: {result}")

执行结果：

--- Result 1 ---
Original: Hello, how are you?
Translated: 你好，你好吗？
Cache Key: None

--- Result 2 ---
Original: Hello, how are you?
Translated: Hello, how are you?
Cache Key: None

--- Result 3 ---
Original: Hello, how are you?
Translated: 你好，你好吗？
Cache Key: 5693a0a16dff61866ca18feef969719cc78979a5f44cd412fef1add7bdcaaa42

Todo

1. OpenAI Tools 完全支持
2. OpenAI JSON Mode 支持
3. support data viewer
4. add to langeval/ragas/distilabel distilabel support qianfan、llm cache