LangChain 51 深入理解LangChain 表达式语言十四自动修复配置RunnableConfig LangChain Expression Language (LCEL)

最新推荐文章于 2024-09-26 11:55:47 发布

AI架构师易筋

最新推荐文章于 2024-09-26 11:55:47 发布

阅读量1k

点赞数 29

分类专栏： LLM-Large Language Models 文章标签： langchain python chatgpt LLM prompt 语言模型

本文链接：https://blog.csdn.net/zgpeace/article/details/135329621

版权

LLM-Large Language Models 专栏收录该内容

185 篇文章 121 订阅

订阅专栏

LangChain系列文章

在这里插入图片描述

1. 接受可运行的配置RunnableConfig

可运行的lambda表达式可以选择接受一个RunnableConfig，它们可以使用它来传递回调，日志打点和其他配置信息给嵌套的运行。

本例子自动修复解析Json失败，通过OpenAI 的API prompt，自动修复未正常格式的Json。

# 从 'operator' 模块导入 'itemgetter' 函数，用于项查找
from operator import itemgetter

# 从 langchain 和 langchain_core 包中导入各种类和函数
from langchain_core.runnables import RunnablePassthrough
from langchain.prompts import ChatPromptTemplate
from langchain.chat_models import ChatOpenAI
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableParallel
from langchain_core.runnables import RunnableLambda
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnableConfig
import json  # 导入 JSON 库来解析和生成 JSON
from langchain.callbacks import get_openai_callback  # 导入获取 OpenAI 回调的函数

from dotenv import load_dotenv  # 导入从 .env 文件加载环境变量的函数
load_dotenv()  # 调用函数实际加载环境变量

from langchain.globals import set_debug  # 导入在 langchain 中设置调试模式的函数
set_debug(True)  # 启用 langchain 的调试模式

# 定义一个函数来解析 JSON 或修复它（如果它格式不正确）
def parse_or_fix(text: str, config: RunnableConfig):
    # 使用模板提示和 OpenAI 模型定义一系列操作来修复文本
    fixing_chain = (
        ChatPromptTemplate.from_template(
            "Fix the following text:\n\n```text\n{input}\n```\nError: {error}"
            " Don't narrate, just respond with the fixed data."
        )
        | ChatOpenAI()  # 连接到 ChatOpenAI 模型来处理提示
        | StrOutputParser()  # 解析聊天模型的字符串输出
    )
    
    # 尝试解析 JSON 三次，如果有错误，使用链来修复
    for _ in range(3):
        try:
            return json.loads(text)  # 尝试将文本解析为 JSON
        except Exception as e:  # 捕获解析异常
            # 使用格式错误的文本和错误信息调用修复链
            text = fixing_chain.invoke({"input": text, "error": e}, config)
    return "Failed to parse"  # 如果三次解析都失败，返回失败消息

# 使用上下文管理器从 OpenAI 获取回调，并调用 parse_or_fix 函数
with get_openai_callback() as cb:
    # 创建一个 RunnableLambda 对象，其中包含 parse_or_fix 函数
    # 并使用类似 JSON 的字符串和配置字典调用它
    output = RunnableLambda(parse_or_fix).invoke(
        "{foo: bar}", {"tags": ["my-tag"], "callbacks": [cb]}
    )
    print(output)  # 打印调用结果
    print(cb)  # 打印回调对象

运行结果

develop* $ python LCEL/config.py                                                                                             [21:49:30]
[chain/start] [1:chain:parse_or_fix] Entering Chain run with input:
{
  "input": "{foo: bar}"
}
[chain/start] [1:chain:parse_or_fix > 2:chain:RunnableSequence] Entering Chain run with input:
[inputs]
[chain/start] [1:chain:parse_or_fix > 2:chain:RunnableSequence > 3:prompt:ChatPromptTemplate] Entering Prompt run with input:
[inputs]
[chain/end] [1:chain:parse_or_fix > 2:chain:RunnableSequence > 3:prompt:ChatPromptTemplate] [2ms] Exiting Prompt run with output:
{
  "lc": 1,
  "type": "constructor",
  "id": [
    "langchain",
    "prompts",
    "chat",
    "ChatPromptValue"
  ],
  "kwargs": {
    "messages": [
      {
        "lc": 1,
        "type": "constructor",
        "id": [
          "langchain",
          "schema",
          "messages",
          "HumanMessage"
        ],
        "kwargs": {
          "content": "Fix the following text:\n\n```text\n{foo: bar}\n```\nError: Expecting property name enclosed in double quotes: line 1 column 2 (char 1) Don't narrate, just respond with the fixed data.",
          "additional_kwargs": {}
        }
      }
    ]
  }
}
[llm/start] [1:chain:parse_or_fix > 2:chain:RunnableSequence > 4:llm:ChatOpenAI] Entering LLM run with input:
{
  "prompts": [
    "Human: Fix the following text:\n\n```text\n{foo: bar}\n```\nError: Expecting property name enclosed in double quotes: line 1 column 2 (char 1) Don't narrate, just respond with the fixed data."
  ]
}
[llm/end] [1:chain:parse_or_fix > 2:chain:RunnableSequence > 4:llm:ChatOpenAI] [2.00s] Exiting LLM run with output:
{
  "generations": [
    [
      {
        "text": "{\n  \"foo\": \"bar\"\n}",
        "generation_info": {
          "finish_reason": "stop",
          "logprobs": null
        },
        "type": "ChatGeneration",
        "message": {
          "lc": 1,
          "type": "constructor",
          "id": [
            "langchain",
            "schema",
            "messages",
            "AIMessage"
          ],
          "kwargs": {
            "content": "{\n  \"foo\": \"bar\"\n}",
            "additional_kwargs": {}
          }
        }
      }
    ]
  ],
  "llm_output": {
    "token_usage": {
      "completion_tokens": 9,
      "prompt_tokens": 56,
      "total_tokens": 65
    },
    "model_name": "gpt-3.5-turbo",
    "system_fingerprint": null
  },
  "run": null
}
[chain/start] [1:chain:parse_or_fix > 2:chain:RunnableSequence > 5:parser:StrOutputParser] Entering Parser run with input:
[inputs]
[chain/end] [1:chain:parse_or_fix > 2:chain:RunnableSequence > 5:parser:StrOutputParser] [1ms] Exiting Parser run with output:
{
  "output": "{\n  \"foo\": \"bar\"\n}"
}
[chain/end] [1:chain:parse_or_fix > 2:chain:RunnableSequence] [2.01s] Exiting Chain run with output:
{
  "output": "{\n  \"foo\": \"bar\"\n}"
}
[chain/end] [1:chain:parse_or_fix] [2.38s] Exiting Chain run with output:
{
  "foo": "bar"
}
{'foo': 'bar'}
Tokens Used: 65
        Prompt Tokens: 56
        Completion Tokens: 9
Successful Requests: 1