refactor: change from standar backend api to python library

This commit is contained in:
AlanSilvaaa
2026-05-28 15:02:42 -04:00
parent 8d3a18ec39
commit da59c4e64d
23 changed files with 331 additions and 833 deletions

View File

@@ -1,38 +1,30 @@
# AGENTS.md
This project is a FastAPI backend for generating structured math problems. The intended design should scale to grades 1 through 6 and multiple problem types per grade.
This project is a Python library for generating structured math problems. The intended design should scale to grades 1 through 6 and multiple problem types per grade.
## Run And Test
- Start the API with `uv run uvicorn main:app --reload`.
- Open API docs at `http://127.0.0.1:8000/docs`.
- Run tests with `uv run python -m unittest`.
- Compile changed Python files with `uv run python -m py_compile <files>`.
- Run the usage example with `uv run python test.py`.
## Endpoint example
## Usage example
- `POST /math/grade_1/join_pictures_with_quantity`
- The caller sends image metadata and optional generation settings.
- The endpoint returns a JSON-ready math problem where students count images and match groups to numbers.
- `from math_problems_structure.grade_1 import join_pictures_with_quantity`
- The caller passes image metadata and optional generation settings.
- The function returns a JSON-ready math problem where students count images and match groups to numbers.
## Responsibilities
- `main.py` is only the uvicorn entry point. Keep it thin.
- `app/main.py` defines `create_app()` and includes the top-level API router.
- `app/routers/math.py` owns the `/math` router and includes grade routers.
- `app/routers/grade_1.py` registers all grade 1 HTTP endpoints.
- `app/__init__.py` exposes the public library functions.
- `app/problems/grade_1/` contains reusable generation logic for grade 1 problems.
- `app/schemas/grade_1/` contains Pydantic request/response models for grade 1 problems.
- `tests/` contains endpoint tests using `fastapi.testclient.TestClient`.
- `tests/` contains direct function tests.
## Design Rules
- Keep HTTP code in `app/routers/`.
- Keep generation logic in `app/problems/`.
- Keep all Pydantic models in `app/schemas/`.
- Do not put generation logic in routers.
- Do not put FastAPI router code in problem generator modules.
- Prefer one router file per grade, not one router file per problem.
- Prefer one schema file per problem type.
- Prefer one generator file per problem type.
@@ -42,22 +34,18 @@ For a new grade 1 problem named `example_problem`:
1. Add schemas in `app/schemas/grade_1/example_problem.py`.
2. Add generator logic in `app/problems/grade_1/example_problem.py`.
3. Register the endpoint in `app/routers/grade_1.py`.
4. Add endpoint tests in `tests/`.
3. Export the function from `app/problems/grade_1/__init__.py` and `app/__init__.py` if it is public.
4. Add direct function tests in `tests/`.
5. Run `uv run python -m unittest`.
For a new grade, follow the same pattern:
1. Add `app/schemas/grade_2/` and `app/problems/grade_2/`.
2. Add `app/routers/grade_2.py`.
3. Include the new grade router from `app/routers/math.py`.
4. Add tests.
Every time you add a new problem, also add an endpoint curl example on the ENDPOINTS-EXAMPLE.md file.
2. Export public functions from the relevant package `__init__.py` files.
3. Add tests.
## Naming Conventions
- Endpoint paths use the existing style: `/math/grade_1/<problem_name>`.
- Python package names should use valid identifiers like `grade_1`.
- Problem files should use snake case, for example `join_pictures_with_quantity.py`.
- Request schema names should end in `Request`.