A python CLI for the generation of written lesson content based on video and textual context. Leverages Gemini 1.5 Pro and multiple inference layers to provide a selection of lessons to choose from.
git clone https://github.com/Equious/nevermore.git
cd nevermore
Important
You'll need to enter the path to your own GOOGLE_APPLICATION_CREDENTIALS
json in line 14
of Nevermore.py
Install Nevermore:
pip install --editable .
Important
Line 82
is configured currently to support .mov
video types. This will be configurable via CLI in future, but will need to be adjusted to your particular case.
Once installed you can run:
$ nevermore --root-directory "PATH/TO/YOUR/COURSE/DIRECTORY"
Your course directory must be in the following structure:
├── Course
│ ├── Section 1
│ │ ├── Lesson 1
│ │ │ └── lesson.mov
│ │ └── Lesson 2
│ │ └── lesson.mov
│ └── repo-context
│ ├── additional_context.md
│ └── additional_context.sol
Additional context to the video lesson can be added to the repo-context
folder (in future this should pull from a provided GitHub repo).
Note
This Repo contains a random selection of testing context files, remove as needed.