Azure 认知服务工具包集成

使用 LangChain Python 与 Azure 认知服务工具包集成。

此工具包允许您与 Azure 认知服务 API 交互，以便在 LangChain 内启用多模态功能。

目前，工具包包含：

• AzureCogsImageAnalysisTool — 从图像中提取标题、对象、标签和文本
• AzureCogsFormRecognizerTool — 从文档中提取文本、表格和键值对
• AzureCogsSpeech2TextTool — 将语音转录为文本
• AzureCogsText2SpeechTool — 将文本合成为语音
• AzureCogsTextAnalyticsHealthTool — 提取医疗实体

⚠️ 图像分析工具在 macOS 上不可用，因为 azure-ai-vision 目前仅支持 Windows/Linux。

---

安装

pip install -qU langchain-community
pip install -qU azure-ai-formrecognizer
pip install -qU azure-cognitiveservices-speech
pip install -qU azure-ai-textanalytics

# Windows / Linux only
pip install -qU azure-ai-vision

import os

os.environ["OPENAI_API_KEY"] = "sk-..."
os.environ["AZURE_COGS_KEY"] = ""
os.environ["AZURE_COGS_ENDPOINT"] = ""
os.environ["AZURE_COGS_REGION"] = ""

## Create the toolkit
from langchain_community.agent_toolkits import AzureCognitiveServicesToolkit

toolkit = AzureCognitiveServicesToolkit()
toolkit.get_tools()

['Azure Cognitive Services Image Analysis',
 'Azure Cognitive Services Form Recognizer',
 'Azure Cognitive Services Speech2Text',
 'Azure Cognitive Services Text2Speech']

## Use within an Agent
from langchain.agents import create_agent
from langchain_openai import OpenAI

llm = OpenAI(temperature=0)

agent = create_agent(
    model=llm,
    tools=toolkit.get_tools(),
)

agent.invoke(
    "What can I make with these ingredients? "
    "https://images.openai.com/blob/9ad5a2ab-041f-475f-ad6a-b51899c50182/ingredients.png"
)

text

Entering new AgentExecutor chain…

Action:

{
  "action": "Azure Cognitive Services Image Analysis",
  "action_input": "https://images.openai.com/blob/9ad5a2ab-041f-475f-ad6a-b51899c50182/ingredients.png"
}

Observation: Caption: a group of eggs and flour in bowls Objects: Egg, Egg, Food Tags: dairy, ingredient, indoor, thickening agent, food, mixing bowl, powder, flour, egg, bowl Thought: I can use the objects and tags to suggest recipes Action:

{
  "action": "Final Answer",
  "action_input": "You can make pancakes, omelettes, or quiches with these ingredients!"
}

Finished chain. text ‘You can make pancakes, omelettes, or quiches with these ingredients!’

audio_file = agent.invoke("Tell me a joke and read it out loud.")

audio_file

text

Entering new AgentExecutor chain… Action:

{
  "action": "Azure Cognitive Services Text2Speech",
  "action_input": "Why did the chicken cross the playground? To get to the other slide!"
}

Observation: /tmp/tmpa3uu_j6b.wav Thought: I have the audio file of the joke Action:

{
  "action": "Final Answer",
  "action_input": "/tmp/tmpa3uu_j6b.wav"
}

Finished chain. text ‘/tmp/tmpa3uu_j6b.wav’

from IPython import display

display.Audio(audio_file)

agent.invoke(
    """The patient is a 54-year-old gentleman with a history of progressive angina over the past several months.
The patient had a cardiac catheterization in July of this year revealing total occlusion of the RCA and 50% left main disease ,
with a strong family history of coronary artery disease with a brother dying at the age of 52 from a myocardial infarction and
another brother who is status post coronary artery bypass grafting. The patient had a stress echocardiogram done on July , 2001 ,
which showed no wall motion abnormalities , but this was a difficult study due to body habitus. The patient went for six minutes with
minimal ST depressions in the anterior lateral leads , thought due to fatigue and wrist pain , his anginal equivalent. Due to the patient's
increased symptoms and family history and history left main disease with total occasional of his RCA was referred for revascularization with open heart surgery.

List all the diagnoses.
"""
)

text

Entering new AgentExecutor chain… Action:

{
  "action": "azure_cognitive_services_text_analyics_health",
  "action_input": "The patient is a 54-year-old gentleman with a history of progressive angina over the past several months. The patient had a cardiac catheterization in July of this year revealing total occlusion of the RCA and 50% left main disease, with a strong family history of coronary artery disease with a brother dying at the age of 52 from a myocardial infarction and another brother who is status post coronary artery bypass grafting. The patient had a stress echocardiogram done on July, 2001, which showed no wall motion abnormalities, but this was a difficult study due to body habitus. The patient went for six minutes with minimal ST depressions in the anterior lateral leads, thought due to fatigue and wrist pain, his anginal equivalent. Due to the patient's increased symptoms and family history and history left main disease with total occasional of his RCA was referred for revascularization with open heart surgery."
}

Observation: The text conatins the following healthcare entities: 54-year-old is a healthcare entity of type Age, gentleman is a healthcare entity of type Gender, progressive angina is a healthcare entity of type Diagnosis, past several months is a healthcare entity of type Time, cardiac catheterization is a healthcare entity of type ExaminationName, July of this year is a healthcare entity of type Time, total is a healthcare entity of type ConditionQualifier, occlusion is a healthcare entity of type SymptomOrSign, RCA is a healthcare entity of type BodyStructure, 50 is a healthcare entity of type MeasurementValue, % is a healthcare entity of type MeasurementUnit, left main is a healthcare entity of type BodyStructure, disease is a healthcare entity of type Diagnosis, family is a healthcare entity of type FamilyRelation, coronary artery disease is a healthcare entity of type Diagnosis, brother is a healthcare entity of type FamilyRelation, dying is a healthcare entity of type Diagnosis, 52 is a healthcare entity of type Age, myocardial infarction is a healthcare entity of type Diagnosis, brother is a healthcare entity of type FamilyRelation, coronary artery bypass grafting is a healthcare entity of type TreatmentName, stress echocardiogram is a healthcare entity of type ExaminationName, July, 2001 is a healthcare entity of type Time, wall motion abnormalities is a healthcare entity of type SymptomOrSign, body habitus is a healthcare entity of type SymptomOrSign, six minutes is a healthcare entity of type Time, minimal is a healthcare entity of type ConditionQualifier, ST depressions in the anterior lateral leads is a healthcare entity of type SymptomOrSign, fatigue is a healthcare entity of type SymptomOrSign, wrist pain is a healthcare entity of type SymptomOrSign, anginal equivalent is a healthcare entity of type SymptomOrSign, increased is a healthcare entity of type Course, symptoms is a healthcare entity of type SymptomOrSign, family is a healthcare entity of type FamilyRelation, left is a healthcare entity of type Direction, main is a healthcare entity of type BodyStructure, disease is a healthcare entity of type Diagnosis, occasional is a healthcare entity of type Course, RCA is a healthcare entity of type BodyStructure, revascularization is a healthcare entity of type TreatmentName, open heart surgery is a healthcare entity of type TreatmentName Thought: I know what to respond Action:

{
  "action": "Final Answer",
  "action_input": "The text contains the following diagnoses: progressive angina, coronary artery disease, myocardial infarction, and coronary artery bypass grafting."
}

Finished chain. text ‘The text contains the following diagnoses: progressive angina, coronary artery disease, myocardial infarction, and coronary artery bypass grafting.’

---

Edit this page on GitHub or file an issue.

Connect these docs to Claude, VSCode, and more via MCP for real-time answers.