poludmik / unittest_generation_server_demo

[DEMO] Flask server generating unittests for a given chunk of code

[DEMO] Flask server generating unittests for a given chunk of code.

• Gets a chunk of code as a string.
• Returns generated unittests as a string.

Install dependencies:

pip install openai langchain requests Flask langchain_openai transformers accelerate

The /generate_unittest endpoint accepts a json body with code_chunk field and returns a string in response.content.

The app uses OPENAI_API_KEY that needs to be set as an environment variable before running python app.py:

export OPENAI_API_KEY='<YOUR_KEY>'

You can also specify the OpenAI LLM to be used with model_name field, or a codellama model for local inference. Codellama-7B needs around 20GB of ram to run. Haven't tested codellama yet.