• Python 3.6 or higher
• Pip installed
• OpenAI account with access to the Assistant API
• A valid API key

First, install dependencies:

pip3 install -r requirements.txt

Then, change the default configuration in config.toml to your own.
It's okay to leave ASSISTANT_ID and THREAD_ID blank, but OPENAI_API_KEY and LINUX_PATH are necessary.

Use the below command to run the main script in auto mode

python3 gpt_4.py <path to file> <start line> <end line>

or

./gpt_4.py <path to file> <start line> <end line>

This will automatically generate the KUnit code for the file you, test it and self-debug.
Please use relative path to Linux kernel root instead of absolute path.
For example, if you want to generate KUnit code for <linux_path>/drivers/nvme/host/trace.c with function located between line 21 and 35, your command would be:

python3 gpt_4.py drivers/nvme/host/trace.c 21 35

There's a variable max_debug_time in line 377 of gpt_4.py that controls the maximum times for self-debugging. It's set to 5 by default, you can change it to any number you want.

Use the below command to run the main script by manually choosing options

python3 gpt_4.py

or

./gpt_4.py

Then you will see a menu containing some options:

This function script will ask you for the path of the file you want to generate the KUnit code for.
Then it will compile the file and run the tests.
After that, if errors occurred, you can then send the errors to GPT and return back the fixed code.

This function is also included in the test generating mode. If you run it individually, you will need to enter a path to a text file that contains errors.
This function will send the errors to GPT and return back the fixed code.

This function will remove the test file from Makefile, Kconfig and .kunitconfig, but not the test file itself.
There's a double check confirmation to make sure you want to remove the test file.

To display all the conversation messages belongs to current thread, you can run the following command:

python3 display_messages.py

This script will display all the messages from the conversation with GPT.
Please be aware that you will have to configure the threads correctly.
This script will display the messages on the screen as well as generating a text file message_log.txt with the messages.

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