This consists of two scripts: getThem and summarize. The getThem script downloads the results of an anonymous open-ended survey in Canvas. The summarize uses Ollama and an LLM to generate a summary of the student responses for the instructor.
This uses the python canvasapi library and your Canvas API key. At the top of the file, enter your Canvas URL and Canvas API key into the API_URL and API_KEY variables. It's setup to use a dictionary of section number -to- Canvas Course ID numbers, so also add that data in the sections list and dictCourseID dictionary as indicated in the file. It is currently set to prompt the user for a week number, and uses that to determine the name of the Canvas quiz that is being used as the survey. Update that to point instead to the name of your survey quiz. It downloads the result (as a csv file that Canvas generates) and writes it to a csv file named by the section number.
This uses Ollama and the python ollama library.
In Ollama, pull the LLM model of your choice. This script is setup to use the
gemma3:27b model. Update the aiModel variable to match the LLM you have pulled.
This is currently setup for surveys of 3 questions. The q1Col, q2Col, and q3Col variables near the top of the file indicate which column in the csv (downloaded from Canvas using getThem) correspond to the student responses for the questions. It generates a markdown file with the results.
- How summarize works: summarize builds a list of the csv files in the directory. For each of those files, it writes a header with the question students were asked. Following that, it builds a querystring to send to the ollama chat giving details of the question and all the student responses and asking for a summary, with some details about how to format it's response. When the response is generated, it gets printed below the question, and summarize moves to the next.