In order to reduce the number of times, we gotta cache the code with the title, with scene_number, and code to be followed for one single key.
While retrieving, we check what's similar and add another node to check what's similar, then let the scene_rendering node create the code by modifying what's left
{'key':"scene_1 import manim \n ..... scene_2 import manim \n...... "} cons: fewer keys, but it's hard to split by scenes, too clustered
or
{'LinearRegressionScene02': "import manim \n ...."} cons: more keys to compare with, kinda expensive
In order to reduce the number of times, we gotta cache the code with the title, with
scene_number, and code to be followed for one single key.While retrieving, we check what's similar and add another node to check what's similar, then let the scene_rendering node create the code by modifying what's left
{'key':"scene_1 import manim \n ..... scene_2 import manim \n...... "}cons: fewer keys, but it's hard to split by scenes, too clusteredor
{'LinearRegressionScene02': "import manim \n ...."}cons: more keys to compare with, kinda expensive