The primary purpose of this PR is to add the source code of the new Marie that uses in-context learning and a React frontend.
However, please note that because of the history of how the development came about (initial experiments with Marie re-implementation, then Zaha for some CARES demos, then back to Marie all the way for the CDE paper), the code also contains some implementations for Zaha, which are beyond the scope of this PR. Although the parts that concern Zaha are documented for the purpose of completeness, they are not guaranteed to work and will be removed or refactored in a subsequent PR.
The primary purpose of this PR is to add the source code of the new Marie that uses in-context learning and a React frontend.
However, please note that because of the history of how the development came about (initial experiments with Marie re-implementation, then Zaha for some CARES demos, then back to Marie all the way for the CDE paper), the code also contains some implementations for Zaha, which are beyond the scope of this PR. Although the parts that concern Zaha are documented for the purpose of completeness, they are not guaranteed to work and will be removed or refactored in a subsequent PR.