Azure AI Language Question Answering

1.1.0 verified Mon Apr 27 auth: no python

Microsoft Azure Question Answering client library for Python, enabling extraction of answers from custom knowledge bases or FAQ data. Version 1.1.0 supports both custom and prebuilt question answering, with synchronous and asynchronous clients. Released monthly as part of Azure SDK.

pip install azure-ai-language-questionanswering

Common errors

error AttributeError: 'QuestionAnsweringClient' object has no attribute 'get_answers' ↓

cause Using an older version of the library (<1.0.0) where method was `query_knowledgebase`

fix

Upgrade to version 1.1.0 and use get_answers method.

error azure.core.exceptions.ResourceNotFoundError: (NotFound) The specified project does not exist. ↓

cause `project_name` or `deployment_name` missing or incorrect

fix

Ensure both parameters are provided and match exactly the names in Azure AI Language resource.

error TypeError: get_answers() got an unexpected keyword argument 'top' ↓

cause Using `top` parameter which was renamed to `top_answers` in version 1.1.0

fix

Replace top with top_answers in the call.

Warnings

breaking In version 1.1.0, the `get_answers` method signature changed: `top` parameter renamed to `top_answers` and `confidence_threshold` renamed to `confidence_score_threshold`. ↓

fix Update calls to use new parameter names.

deprecated The `text_documents` parameter in `get_answers_from_text` is deprecated in favor of `text_documents` as part of `AnswersFromTextOptions`. ↓

fix Use `AnswersFromTextOptions` model to pass text documents.

gotcha Both `project_name` and `deployment_name` are required for custom QnA (non-text queries). Missing them results in a ResourceNotFound error. ↓

fix Ensure both parameters are provided when querying a project.

gotcha When using `get_answers_from_text`, the `text_documents` parameter expects a list of `TextDocument` objects, not raw strings. ↓

fix Import `TextDocument` from models and wrap strings: `from azure.ai.language.questionanswering.models import TextDocument`.

Imports

QuestionAnsweringClient

wrong

from azure.ai.language.questionanswering.models import QuestionAnsweringClient

correct

from azure.ai.language.questionanswering import QuestionAnsweringClient

Client class is in the top-level module, not models

AnswersFromTextOptions

wrong

from azure.ai.language.questionanswering import AnswersFromTextOptions

correct

from azure.ai.language.questionanswering.models import AnswersFromTextOptions

Model classes are under .models

AnswersOptions

from azure.ai.language.questionanswering.models import AnswersOptions

Quickstart

Authenticate using endpoint and key, then query a knowledge base.

import os
from azure.ai.language.questionanswering import QuestionAnsweringClient
from azure.core.credentials import AzureKeyCredential

endpoint = os.environ.get('AZURE_QA_ENDPOINT', '')
credential = AzureKeyCredential(os.environ.get('AZURE_QA_KEY', ''))
client = QuestionAnsweringClient(endpoint, credential)

response = client.get_answers(
    question="What is a good pet?",
    project_name="SampleProject",
    deployment_name="production"
)
if response.answers:
    print(response.answers[0].answer)