Tropos

A system that gives feedback on student assignments using pretrained models.

How does it work?

Installing

In order to install this tool you must run:

pip install -r requirements.txt

Running

Locally

Run this command in the root directory of the repository.

python .

In Google Collab

Open the notebook.ipynb file and run all of the cells.

Data Dictionary

test_feedback_console(prompt_type, model, requirements_path, example_dir, target_dir, output_dir, output_mode, max_examples)

Parameter	Definition	Accepted Values / Notes
prompt_type	Prompting strategy used to guide feedback generation.	"FewShot" or "FewShot-Llama", "OneShot", "ZeroShot"
model	Model name used to generate feedback.	"gpt-4o", "gpt-4.1", "deepseek-chat", "gemini-1.5-pro-latest", "meta-llama/llama-4-scout-17b-16e-instruct", "claude-3-opus-20240229"
requirements_path	File path to the assignment requirements document.	.docx file
example_dir	Directory containing example student submissions for the model.	Folder path
target_dir	Directory containing target (unmarked) student submissions to generate feedback for.	Folder path
output_dir	Directory to save generated feedback documents.	Folder path (created if doesn't exist)
output_mode	How to display feedback in console.	"pretty" = prints readably, "raw"= prints raw for debugging, "none" = no print
max_examples	Limit the number of examples loaded (for RateLimitErrors).	Integer or None