PROMPT 2.0: A Modular and Scalable Management Platform for Project-Based Teaching

What is PROMPT?

PROMPT (Project-Oriented Modular Platform for Teaching) is a course management tool built to simplify the administration of project-based university courses.

Originally developed for the iPraktikum at the Technical University of Munich, PROMPT is now a modular and extensible platform that adapts to the needs of instructors and scales across different teaching formats.

Key Features

Core Features

• Course Configuration: Build custom course flows using reusable course phases.
• Student Management: Manage participant data and applications.
• Application Phase: Streamlined workflows for student applications.

Dynamically Loaded Course Phases

PROMPT supports custom course phases as independently deployable modules:

• 🗓 Interview Phase – Schedule and manage student interviews.
• 🎤 Presentation Phase – Schedule presentations, collect materials, and provide structured multi-instructor feedback.
• 🧑‍🤝‍🧑 Team Phase – Allocate students to projects and teams.
• 📄 TUM Matching Export – Export data in TUM-compatible format.
• 🧩 Custom Phases – Easily extend PROMPT with your own logic.

---

Guide to PROMPT

Explore the live system here 👉 https://prompt.aet.cit.tum.de/

PROMPT is built for teaching teams who run complex project-based university courses. This guide gives you a first impression of the product, its modules, and typical workflows.

Build and Configure Courses

PROMPT lets you define courses by combining reusable “phases”:

• Add application, interview, matching, and many more phases
• Define prerequisites and outcomes for each phase

---

Manage Students and Teams

PROMPT provides powerful tools for handling student participation throughout your course:

• Track incoming applications and monitor their progress
• Schedule and conduct interviews with applicants
• Form teams and assign projects using manual or semi-automated workflows
• Evaluate student performance on a per-phase basis, enabling you to decide who passes or fails at each stage of the course independently

---

Extend PROMPT with Custom Phases

Each course phase can be developed and plugged in independently—ideal for institutions with evolving or specialized workflows.

---

Development Guide

This section is for developers and contributors looking to run PROMPT locally or contribute to its development.

Project Structure

Clients

• Built with React, TypeScript, and Module Federation (rspack)
• Micro-frontends for each course phase
• Shared design system using shadcn/ui

Server

• Developed in Go
• Core service + modular service architecture for each course phase
• PostgreSQL for data storage
• Keycloak for authentication

---

Getting Started

Prerequisites

Tool	Version	Installation
Go	1.26+	go.dev/doc/install
Node.js	22.10.0 - 22.x	nodejs.org
Docker	Latest	docker.com
golang-migrate	Latest	github.com/golang-migrate/migrate
sqlc	Latest	docs.sqlc.dev

1. Enable Corepack (for Yarn 4)

# macOS with Homebrew (Homebrew strips corepack from Node)
brew install corepack

# Other platforms (included with Node 16.9+)
npm install -g corepack

# Enable corepack
corepack enable

2. Clone & Configure Environment

git clone https://github.com/prompt-edu/prompt.git
cd prompt

# Create environment files from templates
cp .env.template .env
cp .env.dev.template .env.dev  # Local development overrides

The .env file contains Docker/production configuration. The .env.dev file contains local development overrides (localhost instead of Docker hostnames) and is loaded by make server.

3. Start Database, Keycloak and File Storage

# every service database and Keycloak (make db-down stops them again)
make db
# object storage
docker compose up -d seaweedfs-volume seaweedfs-s3

4. Configure Keycloak (first time only)

1. Navigate to http://localhost:8081

2. Login with username admin and password admin

3. In the top-left dropdown, select Create Realm and upload keycloakConfig.json

4. In the new Prompt realm, create a user:

   • Go to Users → Add user
   • Fill in all required fields (these are needed for token claims):
     • Username
     • Email (required)
     • First name (required)
     • Last name (required)
   • Click Join Groups and select PROMPT-Admins
   • Go to Credentials tab and set a password
   • Go to Attributes tab and add:
     • university_login → ab12cde
     • matriculation_number → 01234567

5. Configure client mappers (to include user attributes in tokens):

   • Go to Clients → prompt-server → Client scopes tab

   • Click prompt-server-dedicated

   • Click Add mapper → By configuration → User Attribute

   • Create mappers for each attribute:

     Name	User Attribute	Token Claim Name	Claim JSON Type
     university_login	university_login	university_login	String
     matriculation_number	matriculation_number	matriculation_number	String

   • Ensure Add to ID token and Add to access token are enabled

6. Generate a client secret:

   • Go to Clients → prompt-server → Credentials
   • Click Regenerate and copy the secret
   • Paste the secret into your and files as

5. Start the Server

The .env file is not automatically loaded by Go. Use one of these methods:

1. Use Make (recommended, works in any shell):

   make server
   

2. Manual export (bash/zsh only):

   set -a && source .env.dev && set +a
   cd servers/core && go run main.go
   

6. Start the Clients

make clients
# or manually:
cd clients && yarn install && yarn run dev

This launches all micro-frontends simultaneously using Lerna. The app runs at http://localhost:3000.

To start a specific micro-frontend only, navigate to its subdirectory and run yarn run dev there.

7. Seed the Databases (optional)

make seed

This loads one fully populated demo course - every phase type, with participants, applications, interviews, teams, assessments and certificates - into all eight service databases. Run it after the servers have started at least once, since they create the schemas. See the seeding guide.

---

Keycloak Passkey Testing

Option 1 (Preconfigured user with passkey):

Log in as a user with credentials username: student-passkey | password: student-passkey. You will be asked to register your passkey. With next login you can use your registered passkey instead of password-based authentication.

Option 2 (Configure passkey manually):

1. Enable Registration for a User

To test the flow, you must force a specific user to register their biometric data:

Access Keycloak Admin: Go to http://localhost:8081 (Login: admin / admin).

Select Realm: Ensure the Prompt realm is selected in the top-left corner.

Configure User:

Navigate to "Users" → Click on the user (e.g., student).

In the "Details" tab, find "Required user actions" field.

Select "Webauthn Register Passwordless".

Click "Save".

2. Register the Passkey

Open http://localhost:3000 and click "Login".

Sign in using standard credentials (e.g., student / student).

Keycloak will prompt you to register a passkey.

Your browser/OS will prompt for user verification, such as Touch ID, Face ID, Windows Hello, a PIN, or a security key. Follow the prompts to complete registration.

3. Verify Passwordless Login

Log out or open a new Incognito/Private window.

Go to http://localhost:3000 → Login.

Click "Try another way".

Select "Passkey" → Sign in with Passkey.

Use your new user verification method. You should be logged in instantly without entering a password.

---

Useful Commands

# Generate sqlc code (after changing SQL queries)
cd servers/<service> && sqlc generate

# Run Go tests
cd servers/<service> && go test ./...

# Lint clients
cd clients && yarn biome check "core"

Shared UI components are not kept in this repository. They live in prompt-edu/prompt-lib and are consumed here as @tumaet/prompt-ui-components, so new shadcn/ui primitives are added there and released as a new package version.

Git Hooks (pre-commit)

We use the cross-platform pre-commit framework for git hooks. Install pre-commit itself once (pipx install pre-commit, pip install --user pre-commit, or brew install pre-commit), then set up the hooks:

make install-hooks            # or: pre-commit install

On every commit the hooks run against the staged files:

• hygiene — trailing whitespace, end-of-file newline, YAML/JSON validity, merge-conflict markers, large files;
• gofmt — fails if a staged Go file is not formatted (needs Go on PATH);
• biome — lints/format-checks staged client files (same as make lint);
• swagger docs — when Go service sources are staged, regenerates and re-stages the committed swagger specs. Requires swag on PATH (go install github.com/swaggo/swag/cmd/swag@latest). Because this hook regenerates files, a commit with stale docs is aborted once with the docs updated and staged — re-run git commit to include them.

Check the whole repo at any time with pre-commit run --all-files. To bypass the hooks for a single commit, use git commit --no-verify.

If you previously installed the old bash hook via core.hooksPath, clear the override once with git config --local --unset core.hooksPath.

Ports Reference

Service	Port
Core Client	3000
Infrastructure Setup Client	3012
Keycloak	8081
Core Server	8080
Interview Server	8087
Team Allocation Server	8083
Self Team Allocation Server	8084
Assessment Server	8085
Example Server	8086
Certificate Server	8088
Presentation Server	8089
Infrastructure Setup Server	8091
PostgreSQL (core)	5432
PostgreSQL (infrastructure setup)	5441
SeaweedFS Master	9333
SeaweedFS Volume	8888
SeaweedFS S3 Gateway	8333

---

💬 Community & Support

We are building a community around PROMPT! Whether you are an instructor using the platform or a developer looking to contribute, we'd love to have you.

• Discord: Join our Discord Server for real-time discussion and support.
• GitHub Issues: Report bugs or request features on our Issue Tracker.
• Documentation: Visit our Documentation Portal for detailed guides.

🤝 Contributing

We welcome contributions of all kinds! Please see our Contributing Guidelines to get started.

---

Configuration

PROMPT can be customized for your course by composing it from different modular phases. These configurations are handled at the course level and dynamically loaded on demand.