Single-Step Classification

import json
import os
from openai import OpenAI
from flashlearn.skills.classification import ClassificationSkill
from flashlearn.utils import imdb_reviews_50k

def main():
    # Step 1: Setup your provider
    # Uncomment and set your API key if needed:
    # os.environ["OPENAI_API_KEY"] = 'YOUR API KEY'

    # Example using the OpenAI client (you could also configure DeepSeek or another provider)
    client = OpenAI()

    # Step 2: Load sample data (a list of dictionaries with "review" and "sentiment" keys)
    reviews = imdb_reviews_50k(sample=100)

    # Step 3: Initialize the ClassificationSkill for sentiment analysis.
    # Here, we specify the model, client, allowed categories, and the system prompt.
    skill = ClassificationSkill(
        model_name="gpt-4o-mini",   # or use your preferred model
        client=client,
        categories=["positive", "negative"],
        max_categories=1,
        system_prompt="We want to classify short movie reviews by sentiment."
    )

    # Step 4: Prepare classification tasks:
    # Remove the original sentiment key, retaining only the review text.
    removed_sentiment = [{'review': x['review']} for x in reviews]
    tasks = skill.create_tasks(removed_sentiment)

    # Step 7: (Optional) Save tasks to a JSONL file for offline processing or auditing.
    with open('tasks.jsonl', 'w') as jsonl_file:
        for entry in tasks:
            jsonl_file.write(json.dumps(entry) + '\n')

    # Step 5: Execute classification tasks in real time.
    # FlashLearn processes these tasks concurrently.
    results = skill.run_tasks_in_parallel(tasks)

    # Step 6: Map results back to input data and check accuracy.
    correct = 0
    for i, review in enumerate(reviews):
        reviews[i]['category'] = results[str(i)]['categories']
        if reviews[i]['sentiment'] == results[str(i)]['categories']:
            correct += 1
    print(f'Accuracy: {round(correct / len(reviews), 2)} %')

    # Step 7: Save the final results to a JSONL file.
    with open('results.jsonl', 'w') as jsonl_file:
        for entry in reviews:
            jsonl_file.write(json.dumps(entry) + '\n')

    # Step 8: Save the skill configuration for future use.
    skill.save("BinaryClassificationSkill.json")

if __name__ == "__main__":
    main()