Free tools Windows power users keep installed
One-click scans. No signup required.
This tutorial builds a small FastAPI service that accepts a Python exercise submission, asks a configured AI model for structured tutoring feedback, validates the response, and saves the attempt and topic mastery in SQLite. Here, “adaptive” means the service uses a learner’s previously stored topic score as context and updates a bounded score after each submission—not that it measures learning with a validated assessment.
What the tutor does—and does not do
The feedback loop has four parts: receive a submission, load the learner’s prior mastery for that topic, request feedback from a model, and validate and store the result. Feedback is designed to identify a likely issue, recognize something useful in the attempt, suggest a next hint, and ask a question.
As an Amazon Associate I earn from qualifying purchases.
The example does not run submitted Python, decide whether a learner passes a course, or replace an instructor. Its mastery score is application state, not an established measure of learning or evidence that the system improves educational outcomes.
Prerequisites and setup
The tutorial assumes Python 3.10 or later, an API key, a terminal, an HTTP client such as curl, and basic familiarity with Python functions, JSON, and HTTP requests. Its example uses FastAPI, Uvicorn, the OpenAI SDK, Pydantic, pydantic-settings, and SQLite. Verify that the versions you install are compatible with one another; the tutorial does not establish compatibility for particular package releases or model names.
#1 Best Overall
Install the named packages with:
pip install fastapi uvicorn openai pydantic pydantic-settingsConfigure the API key, model name, and SQLite database path through environment-driven settings. Keep the local .env file and database out of version control. These are project practices, not a guarantee that a deployed service is secure.
Shape the request, feedback, and stored state
Keep three kinds of data distinct: the incoming exercise submission, the model’s feedback, and the application’s persisted progress. The request contains a learner identifier, topic, exercise, and submitted code. Constrain these fields before using them.
Define a response model for the feedback fields and validate the model’s structured output against it. Treat model output as untrusted input: validation should fail the request rather than letting malformed or incomplete data silently update progress. Keep the score transition in application code so the model supplies feedback, while the service controls how stored mastery changes.
Do these 3 things before closing this tab:
1Scan for outdated or missing drivers - takes under a minute2Repair Windows errors before they cause bigger problems3Fix the driver behind crashes, sound loss and screen glitchesImplement the feedback and persistence flow
- Accept the submission. Define a FastAPI endpoint whose request schema validates the learner identifier, topic, exercise, and code.
- Load prior topic mastery. Query SQLite for the learner’s existing score for the submitted topic. Use that value as context for the model’s tutoring response.
- Request structured feedback. Send the exercise, submitted code, and relevant prior mastery to the configured model. Ask for feedback in the fields your response model requires.
- Validate before changing state. Parse the model response and validate it against the declared response model. Do not treat an unvalidated model response as a database update.
- Calculate and bound the new score. Apply the tutorial’s score transition in application code and clamp the aggregate mastery value to its defined range.
- Persist the attempt and mastery. Record the attempt and updated topic score in SQLite using parameterized SQL writes.
- Return the validated result. Send the structured feedback and relevant progress result to the client only after the processing steps succeed.
This separation makes the state transition inspectable: the model proposes teaching feedback, while ordinary application logic validates the shape of that feedback, calculates the bounded score, and writes the resulting state.
What “adaptive” means in this example
The service adapts its prompt context using prior stored mastery for the same topic, then updates that score after feedback. This is a narrow mechanism for varying context and tracking attempts. The source does not establish that the score is calibrated, that its updates correspond to actual learning, or that the approach has been evaluated as an educational intervention. Avoid using it alone for consequential grading or placement.
Try the endpoint with an HTTP client
Once the app is running and its settings are configured, submit a JSON request containing the learner identifier, topic, exercise, and code. For example, the request body should follow the fields defined by your endpoint schema; consult the running API’s generated documentation to confirm the route and exact schema for your implementation. The tutorial’s request identifier is only an input field—it does not establish who is making the request.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Security and deployment boundaries
Do not execute learner code in the API process
The example treats submitted code as data for feedback. Executing arbitrary learner code inside the FastAPI process would create a different and substantially riskier system. If an exercise requires real test results, use a separate isolated runner with strict resource and network restrictions; that runner is outside this example.
Quick wins for a faster PC:
Clear out junk files and repair common Windows errorsFree Scan →Fix the driver behind crashes, sound loss and screen glitchesFind Drivers →Repair Windows errors before they cause bigger problemsFix Now →Authenticate identity separately
A learner identifier supplied in a request body is not authentication. In a real application, derive the learner identity from an authenticated session or token and authorize access to that learner’s records.
Limit sensitive logging
Do not log raw submitted code by default. A submission can contain credentials, personal information, internal configuration, or proprietary material. Design error logging and retention deliberately rather than assuming learner code is harmless.
Keep high-stakes decisions with people
Use human review for high-stakes educational decisions. The example’s model feedback and bounded mastery score are not a substitute for instructor judgment.
When SQLite is enough—and when the design must change
SQLite provides local persistence for this focused example. A separately managed database is a distinct deployment choice, not a performance winner established by this tutorial. Likewise, descriptive model feedback is not equivalent to test results from isolated code execution, and a model-suggested progress update is not equivalent to instructor review. Choose among these designs based on the service’s operational and educational requirements rather than treating them as interchangeable features.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
