Oklathon 2025: API Challenge Board

Welcome to the Oklathon 2025 “Climb the Challenge Board” track! These short, 1-hour challenges are designed to get you started with the NRP AI API.

Before you begin:

  1. Get your NRP_API_KEY from an event organizer.
  2. Go to colab.research.google.com and click “New notebook”.
  3. In your new notebook, you must add your API key as a ‘secret’ to keep it secure. In Colab, click the key icon (🔑) on the left sidebar, click ‘Add a new secret’, name it NRP_API_KEY, paste your key in the ‘Value’ field, and ensure the ‘Notebook access’ toggle is on.
# ==============================================================================
# Oklathon 2025 - NRP AI API Challenge Boilerplate
#
# Instructions for Participants:
# 1.  Copy this entire code block into a new Google Colab notebook.
# 2.  Get the NRP_API_KEY from an event organizer.
# 3.  In Colab, click the "🔑" (Secrets) icon on the left panel,
#     create a new secret named `NRP_API_KEY`, and paste your key there.
#     Make sure the "Notebook access" toggle is on.
# ==============================================================================

import os
import requests # We need this for listing the models
from openai import OpenAI, APIError

# --- Configuration ---
# This safely gets your API key from Colab's Secret Manager.
try:
    from google.colab import userdata
    NRP_API_KEY = userdata.get('NRP_API_KEY')
except (ImportError, ModuleNotFoundError):
    # Fallback for running locally if you set an environment variable
    NRP_API_KEY = os.environ.get("NRP_API_KEY")
    if not NRP_API_KEY:
      raise RuntimeError("API Key not found. Please set the NRP_API_KEY environment variable or run in Google Colab with secrets.")
except userdata.SecretNotFoundError:
    raise RuntimeError("API Key not found. Please add it to Colab Secrets as 'NRP_API_KEY'.")

# Define the base URL and initialize the OpenAI client
BASE_URL = "https://llm.nrp-nautilus.io/v1"
client = OpenAI(
    api_key=NRP_API_KEY,
    base_url=BASE_URL
)

# --- Challenge Code Goes Below This Line ---

Task 0: See Your Available Models (IMPORTANT: RUN THIS FIRST!)

Goal: Ask the API which models your key has access to. You’ll need one of these model names for all the other tasks.

Why this is important: Not all API keys have access to all models. This task lets you see your personalized list so you know what you can use.

Spoiler warning! (It's OK to peek at this one!) 
        

```python
# ==============================================================================
# Task 0: See Your Available Models (IMPORTANT: RUN THIS FIRST!)
# Goal: Ask the API which models your key has access to.
# ==============================================================================
print("--- Task 0: See Your Available Models ---")
try:
    # The 'openai' library has a method for this, let's use it!
    models_list = client.models.list()
    print("Your API key has access to the following models:")
    for model in models_list.data:
        print(f"- {model.id}")
except APIError as e:
    print(f"An API error occurred while trying to list models: {e}")
except Exception as e:
    print(f"An unexpected error occurred: {e}")
```

General Purpose & Chat Models

These are your all-rounders, great for conversation, coding, and creative tasks.

Specialized & Tool-Use Models

These models are designed for specific jobs beyond just chatting.

Add the below code to your task, and specify which model will work the best.

# IMPORTANT: Replace the model name below with one from your list from Task 0 that best suits your purpose!
MODEL_TO_USE = "llama3" # <-- CHANGE THIS IF NEEDED

Task 1: The First Connection

Goal: Successfully connect to the API and get your first response from the AI.

What you’ll learn: How to make a basic API call using the openai library and receive a generated response from the NRP AI endpoint. This is the “Hello, World!” of using this API.

Instructions:

  1. Run Task 0 first and get a model name from the output.
  2. Copy that model name and paste it into the MODEL_TO_USE variable in the code below.
Spoiler warning 
        

```python
# ==============================================================================
# Task 1: The First Connection
# Goal: Successfully connect to the API and get your first response.
# ==============================================================================
print("\n--- Task 1: The First Connection ---")

print(f"Attempting to use model: {MODEL_TO_USE}")
try:
    completion = client.chat.completions.create(
        model=MODEL_TO_USE,
        messages=[
            {"role": "system", "content": "You are a helpful assistant."},
            {"role": "user", "content": "Tell me a fun fact about the history of Oklahoma City."}
        ]
    )
    print(completion.choices[0].message.content)
except APIError as e:
    print(f"An API error occurred: {e}")
```

Task 2: The AI Persona Switch

Goal: Make the AI adopt two different personalities using the system prompt. This shows how you can control the AI’s tone and behavior.

Spoiler warning 
        

```python
# ==============================================================================
# Task 2: The AI Persona Switch
# Goal: Make the AI adopt two different personalities using the "system" prompt.
# ==============================================================================
print("\n--- Task 2: The AI Persona Switch ---")
try:
    # Pirate Persona
    pirate_completion = client.chat.completions.create(
        model=MODEL_TO_USE,
        messages=[
            {"role": "system", "content": "You are a sarcastic, world-weary pirate. You answer all questions as such."},
            {"role": "user", "content": "What should I bring to a hackathon?"}
        ]
    )
    print("PIRATE SAYS:")
    print(pirate_completion.choices[0].message.content)

    # Robot Persona
    robot_completion = client.chat.completions.create(
        model=MODEL_TO_USE,
        messages=[
            {"role": "system", "content": "You are BEEP-BOOP, a cheerful robot assistant. You are enthusiastic and helpful."},
            {"role": "user", "content": "What should I bring to a hackathon?"}
        ]
    )
    print("\nROBOT SAYS:")
    print(robot_completion.choices[0].message.content)
except APIError as e:
    print(f"An API error occurred: {e}")
```

Task 3: The Code Explainer

Goal: Get the AI to explain a piece of Python code in plain English. This is a great way to use AI as a learning and debugging tool.

Spoiler warning 
        

```python
# ==============================================================================
# Task 3: The Code Explainer
# Goal: Get the AI to explain a piece of Python code in plain English.
# ==============================================================================
print("\n--- Task 4: The Code Explainer ---")
try:
    code_snippet = "short_names = [name for name in names if len(name) < 5]"
    code_explanation = client.chat.completions.create(
        model=MODEL_TO_USE,
        messages=[
            {"role": "system", "content": "You are an expert Python programmer who excels at explaining complex code to beginners in simple terms."},
            {"role": "user", "content": f"Please explain what this line of Python code does: ```{code_snippet}```"}
        ]
    )
    print(code_explanation.choices[0].message.content)
except APIError as e:
    print(f"An API error occurred: {e}")
```

Task 4: The Data Extractor (JSON Mode)

Goal: Turn a messy sentence into a clean, structured JSON object. This is one of the most powerful and practical uses for LLMs.

Heads-up: This uses response_format={“type”: “json_object”}, which not all models support. meta-llama/Llama-3-8b-chat-hf is a good choice if it’s on your list.

Spoiler warning 
        

```python
# ==============================================================================
# Task 5: The Data Extractor (JSON Mode)
# Goal: Turn a messy sentence into a clean, structured JSON object.
# ==============================================================================
print("\n--- Task 5: The Data Extractor (JSON Mode) ---")
try:
    # Note: For JSON mode to work reliably, you must instruct the model in the prompt!
    # Not all models on the server may support JSON mode. Llama 3 is a good choice.
    json_extraction = client.chat.completions.create(
        model=MODEL_TO_USE,
        response_format={"type": "json_object"}, # This powerful feature asks the AI to only output valid JSON
        messages=[
            {"role": "system", "content": "You are a data processing robot. Extract information from the user's text and respond only with a valid JSON object."},
            {"role": "user", "content": "The project showcase for Oklathon starts Sunday, July 20th at 9:00 AM. It's located at StarSpace46."}
        ]
    )
    print("Extracted JSON data:")
    print(json_extraction.choices[0].message.content)
except APIError as e:
    print(f"An API error occurred: {e}")
except Exception as e:  
```