AI Developer Challenge: Creative Content Generation Pipeline by Ariya BirjandiAI Developer Challenge: Creative Content Generation Pipeline by Ariya Birjandi

AI Developer Challenge: Creative Content Generation Pipeline

Ariya Birjandi

AI Engineer

Redis

SQLite

Electronics

🚀 The AI Developer Challenge

Make Something Insanely Great

Welcome. This isn’t just a coding task. This is a mission. A calling for the bold and curious—those who dare to think differently. If you're ready to build something magical, something powerful, something insanely great—read on.

🌟 The Vision

Imagine this: A user types a simple idea —

“Make me a glowing dragon standing on a cliff at sunset.”

And your app...

Understands the request using a local LLM.

Generates stunning visuals from text.

Transforms that image into an interactive 3D model.

Remembers it. Forever.

You're not building an app. You're building a creative partner.

🎯 The Mission

Create an intelligent, end-to-end pipeline powered by Openfabric and a locally hosted LLM:

Step 1: Understand the User

Use a local LLM like DeepSeek or Llama to:

Interpret prompts

Expand them creatively

Drive meaningful, artistic input into the generation process

Step 2: Bring Ideas to Life

Chain two Openfabric apps together:

Text to Image App ID:

f0997a01-d6d3-a5fe-53d8-561300318557

View on Openfabric

Image to 3D App ID:

69543f29-4d41-4afc-7f29-3d51591f11eb

View on Openfabric

Use their manifest and schema dynamically to structure requests.

Step 3: Remember Everything

Build memory like it matters.

🧠 Short-Term: Session context during a single interaction

💾 Long-Term: Persistence across sessions using SQLite, Redis, or flat files Let the AI recall things like:

“Generate a new robot like the one I created last Thursday — but this time, with wings.”

🛠 The Pipeline

User Prompt ↓ Local LLM (DeepSeek or LLaMA) ↓ Text-to-Image App (Openfabric) ↓ Image Output ↓ Image-to-3D App (Openfabric) ↓ 3D Model Output

Simple. Elegant. Powerful.

📦 Deliverables

What we expect:

✅ Fully working Python project

✅ README.md with clear instructions

✅ Prompt → Image → 3D working example

✅ Logs or screenshots

✅ Memory functionality (clearly explained)

🧠 What We’re Really Testing

Your grasp of the Openfabric SDK (Stub, Remote, schema, manifest)

Your creativity in prompt-to-image generation

Your engineering intuition with LLMs

Your ability to manage context and memory

Your attention to quality — code, comments, and clarity

🚀 Bonus Points

🎨 Visual GUI with Streamlit or Gradio

🔍 FAISS/ChromaDB for memory similarity

🗂 Local browser to explore generated 3D assets

🎤 Voice-to-text interaction

✨ Example Experience

Prompt:

“Design a cyberpunk city skyline at night.”

→ LLM expands into vivid, textured visual descriptions → Text-to-Image App renders a cityscape → Image-to-3D app converts it into depth-aware 3D → The system remembers the request for remixing later

That’s not automation. That’s imagination at scale.

💡 Where to start

You’ll find the project structure set, the entrypoint is in main.py file.

############################################################
# Execution callback function
############################################################
def execute(model: AppModel) -> None:
    """
    Main execution entry point for handling a model pass.

    Args:
        model (AppModel): The model object containing request and response structures.
    """

    # Retrieve input
    request: InputClass = model.request

    # Retrieve user config
    user_config: ConfigClass = configurations.get('super-user', None)
    logging.info(f"{configurations}")

    # Initialize the Stub with app IDs
    app_ids = user_config.app_ids if user_config else []
    stub = Stub(app_ids)

    # ------------------------------
    # TODO : add your magic here
    # ------------------------------
                                
                                
                                
    # Prepare response
    response: OutputClass = model.response
    response.message = f"Echo: {request.prompt}"

Given schema, stub implementation and all the details you should be able to figure out how eventing works but as an extra hint (if needed) here is an example of calling and app get the value and save it as an image:

    # Call the Text to Image app
    object = stub.call('c25dcd829d134ea98f5ae4dd311d13bc.node3.openfabric.network', {'prompt': 'Hello World!'}, 'super-user')
    image = object.get('result')
    # save to file
    with open('output.png', 'wb') as f:
        f.write(image)

How to start

The application can be executed in two different ways:

locally by running the start.sh

on in a docker container using Dockerfile

If all is fine you should be able to access the application on http://localhost:8888/swagger-ui/#/App/post_execution and see the following screen:

Ground Rules

Step up with any arsenal (read: libraries or packages) you believe in, but remember:

👎 External services like chatGPT are off-limits. Stand on your own.

👎 Plagiarism is for the weak. Forge your own path.

👎 A broken app equals failure. Non-negotiable.

This Is It

We're not just evaluating a project; we're judging your potential to revolutionize our landscape. A half-baked app won’t cut it.

We're zeroing in on:

👍 Exceptional documentation.

👍 Code that speaks volumes.

👍 Inventiveness that dazzles.

👍 A problem-solving beast.

👍 Unwavering adherence to the brief

Like this project

Posted Jun 24, 2025

Created an AI-driven pipeline for text-to-3D model generation using Openfabric and local LLMs.

Likes

Views