KlassenPilot (Local Prototype)

KlassenPilot is a minimal teacher-level AI copilot prototype for a Gymnasium teacher. It currently supports one class (Klasse 9b) with Chemie + Englisch focus and two workflows:

1. Log a lesson from rough notes.
2. Generate the next lesson plan from accumulated class memory.

Tech stack

• Python
• Streamlit UI
• OpenAI Agents SDK
• Local markdown files as memory
• No database/auth/grading

Project structure

app.py
teacher_agent/
|-- __init__.py
|-- agent.py
|-- backend.py
|-- contracts.py
|-- prompts.py
|-- schemas.py
|-- tools.py
`-- wiki_store.py
teacher_wiki/
|-- AGENTS.md
|-- index.md
|-- log.md
|-- teacher_profile.md
|-- subjects/
|   |-- chemie.md
|   `-- english.md
`-- classes/
    `-- class_9b_2026_27/
        |-- course_state.md
        |-- lesson_graph.md
        |-- student_notes.md
        |-- misconceptions.md
        `-- open_loops.md

Architecture

• app.py contains only Streamlit UI logic: layout, session state, buttons, messages, and rendering.
• teacher_agent/contracts.py defines the fixed UI/backend boundary: request/response types, bootstrap data, and UI-facing memory drafts.
• teacher_agent/backend.py is the facade the UI calls for app bootstrap, lesson structuring, lesson planning, and approved memory saves.
• teacher_agent/schemas.py defines backend/domain schemas: LessonLogInput, WikiUpdateProposal, WikiUpdateBundle, and LessonPlan.
• teacher_agent/wiki_store.py owns all wiki file reads/writes and restricts writes to the approved class-memory files.
• teacher_agent/tools.py exposes safe app actions for loading context, preparing proposals, and applying approved updates.
• teacher_agent/prompts.py contains the system prompt and task prompt builders.
• teacher_agent/agent.py creates and runs the OpenAI Agents SDK agents and validates their outputs into schemas.

Setup

1. Create and activate a virtual environment.
2. Install dependencies:

pip install -r requirements.txt

3. Set your API key:

export OPENAI_API_KEY="your_key_here"

4. Run the app:

streamlit run app.py

How it works

• Enter rough notes in "What happened in today's lesson?".
• Click "Structure lesson notes".
• The agent proposes complete markdown updates for:
  • lesson_graph.md
  • course_state.md
  • misconceptions.md
  • student_notes.md
  • open_loops.md
• Review proposed changes in the UI.
• Click "Approve and save memory" to persist updates.
• Click "Generate next lesson plan" to create a plan from current memory.

Safety & memory rules

• The agent does not write directly to memory files.
• Memory writes only happen after explicit user approval.
• Student notes should only use pseudonymous IDs like S-001, S-002.