v0.4.1

Changes since v0.4.0

• llm_utils.py — shared LLM utilities extracted from entrypoint.py (PR #36)
• Token usage tracking — token_usage.json written to output directory after tutor runs
• Concise success feedback — shorter responses when all tests pass

Grader template update

The grader Dockerfile (python-pytest-template) needs two changes to use this version:

1. Bump --branch v0.3.13 → --branch v0.4.1
2. Add mv /app/temp/prompt_pipeline/ /app/ai_tutor/prompt_pipeline/

See kangwonlee/python-pytest-template#10.

v0.2.1

• Add workflow_dispatch trigger to the Github Actions workflow.
• Remove GITHUB_OUTPUT file writing.
• Remove raising an additional exception when n_failed is non zero.

v0.2.0

• Write to $GITHUB_STEP_SUMMARY

v0.1.10

• Will include available stderr values when generating comments. Expected to be helpful sometimes.

[v0.1.9] - 2025-02-22

Changed

• for edu-base docker, write to GITHUB_OUTPUT only if the env var exists. If the script is running within the docker, there would be no GITHUB_OUTPUT enviroment variable.

[v0.1.8] - 2025-02-20

Added

• Manual publishing

Removed

• no pip cache to save docker size
• support for linux/386,linux/arm/v7,linux/arm/v6 to save docker size
  Expected Github Actions runners : AMD64 or ARM64

v0.1.7

[v0.1.7] - 2025-02-18

Added

• Added the ability to specify the Gemini model when using the AI Tutor GitHub Action. A new model input has been added to the action, allowing users to select different Gemini models. The default model remains gemini-2.0-flash for backward compatibility.
• Added Norwegian support
• Added Docker Image Build & Push to the CI/CD pipeline to save build time per each user commit.
  Please use the following line

    uses: docker://ghcr.io/github-id/action-name:tag

Removed

• Removed major version number yaml seemingly not working in the action.

v0.1.6.1

test docker push

exclude common instructions

Added

• MECE principle in comment generation
• if failed, assert the error message
• fail-expected argument to fail the test if the expected fail count not correct
• Add a feature to exclude common content of README.md assignment instruction.
  • Common content is marked by the starting marker From here is common to all assignments. and the ending m arker Until here is common to all assignments. in the README.md file, surrounded by double backtick (ascii 96) characters.
• Add start and end markers to mutable code block in the prompt for Gemini.

Changed

• Update license to BSD 3-Clause + Do Not Harm.
• Change the default value of fail-expected to false.
• Improve prompt for Gemini.
  • Add header and footer to the prompt.
  • Modify instruction for failed tests to "Please generate comments mutually exclusive and collectively exhaust ive for the following failed test cases.".
  • Add start and end markers to mutable code block.

Removed

• 'Currently Korean Only' from README.md.

Locale folder

Add Swedish support

Italian Support

Add Italian Support

International Support

Default : English

'Chinese'
'French'
'German'
'Japanese'
'Korean'
'Spanish'
'Thai'