PS-Prep

Before the college Project School selection test, programming materials, question sets, solutions, compiler tools, and chatbot conversations were scattered. Practising a single problem meant moving between several tools.

PS-Prep combines that workflow into a focused preparation platform: browse a problem, write Python, run it, inspect the output, reveal an explanation, and ask for contextual guidance. Its question bank contains more than 200 questions across over ten topics.

Students browse questions by topic and difficulty, then write Python in a Monaco-powered browser editor. Piston executes the code and returns standard output or errors. A solution and detailed explanation can be revealed when needed.

Questions and topics are stored dynamically in Supabase. React, Vite, React Router, and Tailwind provide the responsive application interface.

The initial version made programs using input() inconvenient to practise. Student feedback made that gap clear: running code was useful, but students also needed a straightforward way to supply input.

A dedicated Command Line Input field now collects standard input before the student presses Run. The supplied text goes to Piston with the code, allowing input-based problems without building a full interactive terminal.

The tutor receives the current problem’s title and description, the student’s code, and their question. Its instructions favour concise hints and explanations before handing over a complete solution.

The server-side Edge Function provides the boundary to Gemini. Keeping the current exercise in the request avoids asking the student to copy their problem and code into a separate generic chat.

The README reports that the platform was shared through college WhatsApp groups and used by roughly 100+ students during exam preparation. This is the project author’s account of adoption, rather than an independently measured analytics result.

The most concrete outcome was the feedback loop: students encountered an execution limitation, reported it, and the platform gained a dedicated stdin workflow. The repository does not publish accuracy, latency, or test-pass benchmarks.

Execution is Python-only and uses input supplied before a run. There is no hidden-test online judge, and AI guidance still needs a student’s judgement.

The README proposes automated test cases, more language support, progress tracking, stronger automated testing, structured hint levels, richer question metadata, and student contributions as future improvements. These are next steps, not shipped features.