Parse reference solutions and evaluation criteria

[Yahiewi]

This PR implements decoding of encrypted reference solutions and parsing grading criteria from cell metadata to guide the AI tutor's feedback.

[github-actions]

👈 Launch a Binder on branch Yahiewi/jupyter-ai-tutor/metadata

[brichet]

Thanks for working on this @Yahiewi, I left a few comments.

[brichet]

Is this necessary ? We can probably keep the solution in plain text.
Otherwise we should really encrypt it using a passphrase, but this involve more complexity for now.

[Yahiewi]

This was something that was mentioned during the last meeting and can be found in the codimd document. I think the idea isn't necessarily to have encryption as ROT13 can be easily decoded, but to at least make it so inspecting the metadata doesn't directly give you the answer, while keeping things simple (not using any actual secure encryption).

[brichet]

OK let's keep it for now then.

[Yahiewi]

Thanks @brichet for the reviews!
I updated the PR to address all your comments. For the ROT13 encryption thing, let me know if you think we should keep it as is now or if you prefer something else.

[brichet]

Thanks @Yahiewi, it works well.

Can you remove the useless description parameter in other function, now that the full message content is build in the frontend ?

jupyter-ai-tutor/src/api.ts

Line 12 in 7a56d45

description?: string,

jupyter-ai-tutor/jupyter_ai_tutor/handlers.py

Line 19 in 7a56d45

description = body.get("description", "")

[Yahiewi]

Thanks @Yahiewi, it works well.

Can you remove the useless description parameter in other function, now that the full message content is build in the frontend ?

jupyter-ai-tutor/src/api.ts

Line 12 in 7a56d45

description?: string,

jupyter-ai-tutor/jupyter_ai_tutor/handlers.py

Line 19 in 7a56d45

description = body.get("description", "")

Done. Thanks for the review, this looks good to go now.

[brichet]

Thanks, LGTM.

Failing test is not related, jupyter-server/jupyter_releaser#659.