docscan

Document analyzer for teachers

This tool is a document analyzer for teachers to create material, analyze files, grade files, etc. using LLMs

Currently, the application is only able to analyze text based files such as .txt and .pdf. Images are not supported yet.

A demo of the application is here

Getting Started

Prerequisites Either

• Docker Desktop
• LLM either running locally with an application such as Ollama or with OpenAPI Key
  • Currently, this application only supports local LLMS through runners such as Ollama

Or

• Node >= 18
• Python >= 3.9
• LLM either running locally with an application such as Ollama or with OpenAPI Key
  • Currently, this application only supports local LLMS through runners such as Ollama
• PostgreSQL

Docker

The preferred setup since there are many components.

1. Navigate to the root directory and run docker compose up -d

   • docker compose down to stop the containers
   • docker compose down --volumes to stop the containers and remove the volumes
   • docker compose watch to watch for changes in the code and rebuild the containers
   • This creates a backend, frontend, and PostgreSQL container onto a shared docker network that they use to communicate

2. Head to localhost:5174 to view the frontend

3. Head to localhost:8000/docs to view the API documentation

4. Start your Ollama server by running ollama serve

   • Your ollama server should be running on the same network as the docker containers. I.e. you need to bind the server to your computer's local IP/0.0.0.0 or add an ollama container on the same network bridge.
   • Download and fetch deepseek-r1:8b. This is currently the model hardcoded, configurable environments variables to control this will be added in the future.

If you want to run the frontend and backend separately:

1. Navigate to the frontend or backend directory and run the docker commands shown above.
   • Each directory has their own Dockerfile and docker-compose.yml file.

There is currently no production level steps or builds yet. If so, a Dockerfile in root would be used to compile the frontend, add it to an app folder along with the backend server, and then serve the frontend with a web server such as Nginx/Apache or serve from the backend server itself.

Models and Breakdown

Users will need to be able to upload files. Should probably have a classroom/room model that keeps track of all students that are part of a classroom (this will hopefully take care of understanding what files can be accessed by who). File need to also have owners themselves, but as long as they are part of classrooms, a teacher of that classroom can access the file.

Should keep track of assignments. Given an assignment, they can have multiple files, and thus comments File id: uid() filepath: absolute path to file filename: name of file owner: uid()

User id: uid() role: teacher, student, admin