langextract-azureopenai 0.1.4

LangExtract provider plugin for Azure OpenAI

LangExtract Azure OpenAI Provider

A provider plugin for LangExtract that integrates the Azure OpenAI Chat Completions API for robust, structured information extraction.

Features

• Native Azure OpenAI: Uses the official openai Python SDK with Azure endpoints.
• Safe parameter handling: Whitelist filtering; unsupported params raise clear errors.
• Concurrent batching: Parallel inference for multi-prompt workloads.
• Schema-aware: Optional structured output mode (JSON mode) from LangExtract examples.
• Modern packaging: pyproject.toml with Hatch; works well with uv.

Installation

Using UV (Recommended)

# Install UV if you haven't already
curl -LsSf https://astral.sh/uv/install.sh | sh

# Install the package
uv add langextract-azureopenai

Using pip

pip install langextract-azureopenai

From Source

git clone <repository-url>
cd langextract-azureopenai
uv sync

Quick Start

1. Set up Azure OpenAI credentials

export AZURE_OPENAI_API_KEY="your-api-key"
export AZURE_OPENAI_ENDPOINT="https://your-endpoint.openai.azure.com/"
export AZURE_OPENAI_API_VERSION="2024-12-01-preview"  # or your current API version

2. Basic Usage

import os
import langextract as lx

# Explicit configuration is recommended
config = lx.factory.ModelConfig(
    model_id="azureopenai-gpt-4.1",  # your Azure deployment name
    provider="AzureOpenAILanguageModel",
    provider_kwargs={
        "api_key": os.getenv("AZURE_OPENAI_API_KEY"),
        "azure_endpoint": os.getenv("AZURE_OPENAI_ENDPOINT"),
        "api_version": os.getenv("AZURE_OPENAI_API_VERSION"),
    },
)

prompt = "Extract people, organizations, and locations from the text."
examples = [
    lx.data.ExampleData(
        text="John Smith works at Microsoft in Seattle.",
        extractions=[
            lx.data.Extraction(
                extraction_class="person",
                extraction_text="John Smith",
                attributes={"role": "employee"},
            ),
            lx.data.Extraction(
                extraction_class="organization",
                extraction_text="Microsoft",
                attributes={"type": "company"},
            ),
            lx.data.Extraction(
                extraction_class="location",
                extraction_text="Seattle",
                attributes={"type": "city"},
            ),
        ],
    )
]

result = lx.extract(
    text_or_documents="Sarah Johnson is the CEO of TechCorp in San Francisco.",
    prompt_description=prompt,
    examples=examples,
    config=config,
)

for e in result.extractions:
    print(e.extraction_class, "->", e.extraction_text, e.attributes)

3. Advanced Parameters

result = lx.extract(
    text_or_documents=input_text,
    prompt_description=prompt,
    examples=examples,
    config=config,                 # reuse the explicit configuration
    # Azure OpenAI generation params
    temperature=0.3,
    top_p=0.9,
    frequency_penalty=0.1,
    presence_penalty=0.1,
    seed=42,
    user="user-123",
    logprobs=True,
    top_logprobs=2,
)

Supported Model IDs

The provider handles model IDs with the pattern ^azureopenai:

• azureopenai-gpt-4 → Uses deployment: gpt-4
• azureopenai-gpt-35-turbo → Uses deployment: gpt-35-turbo
• azureopenai-your-custom-deployment → Uses deployment: your-custom-deployment

You can also specify the deployment name explicitly:

result = lx.extract(
    # ... other parameters
    model_id="azureopenai-any-name",
    provider_kwargs={"deployment_name": "your-actual-deployment-name"}
)

Environment Variables

Variable	Description	Required
AZURE_OPENAI_API_KEY	Azure OpenAI API key	✅ Yes
AZURE_OPENAI_ENDPOINT	Azure OpenAI endpoint URL	✅ Yes
AZURE_OPENAI_API_VERSION	Azure OpenAI API version (e.g., 2024-12-01-preview)	✅ Yes

Parameters

• Supported: temperature, top_p, frequency_penalty, presence_penalty, stop, logprobs, top_logprobs, seed, user, logit_bias, and advanced response_format.
• Unsupported (raises InferenceConfigError): stream, tools, tool_choice, parallel_tool_calls.

Notes:

• When schema constraints are enabled via examples, the provider sets response_format={"type": "json_object"} to encourage valid JSON output. Strict JSON Schema mode is not enabled at this time.

Development

Prerequisites

# Install UV
curl -LsSf https://astral.sh/uv/install.sh | sh

# Verify installation  
uv --version

Setup

# Clone repository
git clone <repository-url>
cd langextract-azureopenai

# Install dependencies
uv sync

# Install development dependencies
uv sync --group dev

Testing

# Run unit tests
uv run pytest tests/ -v

# Run parameter filtering tests (no credentials required)
uv run python tests/test_parameter_filtering.py

# Run full Azure API tests (requires credentials)
export AZURE_OPENAI_API_VERSION="2024-12-01-preview"
uv run python tests/test_azure_parameters.py

# Run with coverage
uv run pytest tests/ --cov=langextract_azureopenai --cov-report=html

Code Quality

# Format code
uv run black .
uv run isort .

# Lint code  
uv run ruff check .

# Type checking
uv run mypy langextract_azureopenai

Building and Publishing

# Build package
uv build

# Check build
ls dist/

# Publish to PyPI (requires API token)
uv publish --token your-pypi-token

Version Management

# Bump version
python scripts/bump_version.py patch   # 0.1.0 -> 0.1.1
python scripts/bump_version.py minor   # 0.1.0 -> 0.2.0  
python scripts/bump_version.py major   # 0.1.0 -> 1.0.0

Developer Scripts

For a quick overview of the helper scripts (testing, releasing, and versioning), see:

• scripts/README.md — summaries and usage for bump_version.py, run_tests.py, check_build.py, and release.py.

Examples

See the examples/ directory for complete usage examples:

• examples/example_usage.py - Basic entity extraction
• tests/test_azure_parameters.py - Parameter compatibility testing
• tests/test_parameter_filtering.py - Security validation

Testing

The package includes a comprehensive test suite:

• Unit Tests: Core functionality without API calls
• Integration Tests: Real Azure OpenAI API testing
• Security Tests: Parameter filtering validation
• Performance Tests: Batch processing and concurrency

Run specific test categories:

# Unit tests only
uv run pytest tests/ -m "unit"

# Integration tests (requires credentials)
uv run pytest tests/ -m "integration"

Contributing

1. Fork the repository
2. Create a feature branch: git checkout -b feature-name
3. Make changes and add tests
4. Run the test suite: uv run pytest
5. Format code: uv run black . && uv run isort .
6. Submit a pull request

Troubleshooting

Common Issues

ImportError after installation:

# Ensure package is properly installed
uv sync
uv run python -c "import langextract_azureopenai; print('✅ Import successful')"

Authentication errors:

# Verify credentials are set
echo $AZURE_OPENAI_API_KEY
echo $AZURE_OPENAI_ENDPOINT
echo $AZURE_OPENAI_API_VERSION

# Test credentials
uv run python tests/test_provider_basic.py

Parameter errors:

# Check parameter compatibility
uv run python tests/test_azure_parameters.py

License

This project is licensed under the Apache License 2.0 - see the LICENSE file for details.

Links

• LangExtract Documentation
• Azure OpenAI Documentation
• UV Documentation

---

Happy Extracting with Azure OpenAI! 🚀