---
title: AI Teaching Assistant Prototype (Student-Facing Version)
emoji: 🤖
colorFrom: yellow
colorTo: purple
sdk: gradio
sdk_version: 5.42.0
app_file: app.py
pinned: false
short_description: |
  Multimodal & model-agnostic AITA for ESCP students 
---
# ESCP AI Teaching Assistant Prototype

This Space hosts a **student-facing AI Teaching Assistant prototype** developed for ESCP Business School.

The assistant is designed to support students across **four complementary roles**, each with its own memory, behavior, and constraints.

---

## Assistant Modes

### 1. Course Detailer  
Provides **operational and logistical information only**, such as:
- course structure and schedule
- grading scheme and deliverables
- platforms and tools
- policies (attendance, late work, AI usage)

### 2. Reviewer  
Helps students:
- summarize slides
- generate exam-style questions
- review course material for assessment preparation

### 3. Socratic Reviewer  
Supports learning through:
- guided questions
- conceptual clarification
- reflective prompts  
(No direct rewriting unless explicitly requested.)

### 4. Stakeholder  
Simulates a **realistic business stakeholder** reacting to a case study,
remaining fully in character and aligned with the case context.

---

## Knowledge Sources

The assistant uses two structured folders:

- **`CourseContent/`**
  - Syllabi
  - Slides
  - Course-related documents

- **`CaseStudyContent/`**
  - Case descriptions
  - Stakeholder context
  - Internal documents

If these folders contain only placeholder files, the assistant automatically switches to **Demo Mode**.

---

## Demo Mode

When no official material is loaded:
- The assistant explicitly states that it is operating in **Demo Mode**
- A consistent **demo course canon** is used across *all* assistant modes
- Invented details remain internally coherent and aligned

Once demo context is introduced, it is treated as the course “reality” for the remainder of the session.

---

## File Uploads

Students may upload:
- PDF, DOCX, PPTX
- CSV, XLSX, JSON
- Images (PNG/JPG, OCR-enabled if available)

Uploaded files are:
- used only for the current interaction
- truncated for speed and safety
- not persisted beyond the session

---

## Data & Privacy Notes

- This prototype is intended for **educational demonstration purposes**
- Users should **not upload personal or sensitive data**
- Conversations are logged only for development/debugging
- API keys are managed via environment secrets and never stored in code

---

## Technical Stack

- **Gradio** for UI
- **Hugging Face Inference API** for LLM access
- Modular per-mode memory and instruction control
- Bounded file ingestion for performance and safety

---

## Status

This Space is a **prototype** intended for:
- instructional experimentation
- pedagogical design
- controlled student-facing pilots

Not yet intended for production deployment without institutional validation.