调用Azure Document Intelligence的begin_analyze_document接口时出现“Connection string is either blank or malformed”错误求助

Let's break down why you're seeing this error and how to fix it—even though your code uses an endpoint + API key instead of a connection string, there are a few common culprits here:

Possible Causes & Fixes #

1. Incomplete or Malformed API Key #

The key in your code is truncated (ANZ...CzH), which is likely the core issue. API keys are full, unbroken strings—if you copied only part of it, or included extra spaces/newlines by accident, the client can't authenticate properly.

• Fix:
  • Head back to your Azure portal, navigate to your Document Intelligence resource, and go to the Keys and Endpoint page.
  • Copy the full Key1 or Key2 value exactly (no leading/trailing spaces, no ellipses).
  • Replace the truncated key in your code with the full string:

    key = "your-full-untruncated-api-key-here"
    

2. Invalid Endpoint Format #

Double-check your endpoint for typos or formatting issues:

• A valid Document Intelligence endpoint should look like https://<your-resource-name>.cognitiveservices.azure.com/ (note the trailing slash—while some SDK versions tolerate its absence, it's safest to include it).
• Make sure di-myService matches the exact resource name in your Azure portal (typos here will break the connection immediately).

3. SDK Version or Import Mismatch #

If you're migrating from the old Form Recognizer SDK to the new Document Intelligence SDK, mismatched imports or outdated packages can cause unexpected errors:

• Update the SDK: Run this in your terminal to get the latest stable version:

  pip install azure-ai-documentintelligence --upgrade
  

• Verify Imports: Ensure you're using the correct classes for the new SDK:

  from azure.ai.documentintelligence import DocumentIntelligenceClient
  from azure.core.credentials import AzureKeyCredential
  from azure.ai.documentintelligence.models import AnalyzeDocumentRequest, AnalyzeResult
  

  Old Form Recognizer imports (like azure.ai.formrecognizer) won't work with the new DocumentIntelligenceClient, so double-check your code for leftover old references.

4. Accidental Connection String Configuration #

Sometimes, an environment variable or hidden config can override your explicit endpoint/key setup:

• Check if you have an environment variable named AZURE_DOCUMENT_INTELLIGENCE_CONNECTION_STRING set—if so, the client might be trying to use that instead of your coded endpoint/key.
• If you want to switch to using a connection string, grab the full valid string from the Keys and Endpoint page in Azure (format: Endpoint=https://xxx.cognitiveservices.azure.com/;ApiKey=xxx), and initialize the client like this:

  from azure.ai.documentintelligence import DocumentIntelligenceClient
  
  connection_string = "your-full-connection-string-here"
  client = DocumentIntelligenceClient.from_connection_string(connection_string)
  

Quick Validation Test #

First, confirm your client initializes correctly before running the full analysis:

from azure.ai.documentintelligence import DocumentIntelligenceClient
from azure.core.credentials import AzureKeyCredential

endpoint = "https://di-myService.cognitiveservices.azure.com/"
key = "your-full-api-key"

try:
    client = DocumentIntelligenceClient(endpoint=endpoint, credential=AzureKeyCredential(key))
    print("Client initialized successfully!")
except Exception as e:
    print(f"Initialization failed: {str(e)}")

If this works, proceed with your original begin_analyze_document code—if not, the error message here will give you more specific clues to debug.

内容的提问来源于stack exchange，提问作者RT.