README.md

The models folder contains:

Data

Passive Pydantic shapes for request/response payloads. No behaviour. Used for LLM logic e.g., app/services/chat.py and API requests e.g., app/api/chat.py

Name	Purpose
chat	Schemas for chat messages, chat state, and chat-endpoint requests.
feedback	Schemas for feedback responses and feedback-endpoint requests.

Environments (env configs)

Configuration that defines the context a chat or feedback turn runs in: who the user is, what lesson they're doing.

Name	Purpose
lesson	Lesson config (UI presentation, instructions per lesson type, feedback focus).
session	User session config (name, language, proficiency level).

NB. Environments is a made up name; no idea what the convention is!

LLMs

Wrappers that define how a model should behave for a given task: build prompts, hold per-call parameters. They do not call the LLM. That lives in app/services.

Name	Purpose
chat_model	Build the system prompt and chat-call parameters from session and lesson config.
feedback_model	Building the immediate and general feedback prompts from session, lesson, and conversation.