1
%pip install gretel-client opik pandas --upgrade --quiet

1
import os
2
import getpass
3
import opik
4
import pandas as pd
5
6
print("🔐 Setting up authentication...")
7
8
# Set up Gretel API key
9
if "GRETEL_API_KEY" not in os.environ:
10
    os.environ["GRETEL_API_KEY"] = getpass.getpass("Enter your Gretel API key: ")
11
12
# Set up Opik (will prompt for API key if not configured)
13
opik.configure()
14
15
print("✅ Authentication completed!")

1
from gretel_client.navigator_client import Gretel  # Use navigator_client instead!
2
from gretel_client.data_designer import columns as C
3
from gretel_client.data_designer import params as P
4
5
print("🤖 Setting up Q&A dataset generation with Gretel Data Designer...")
6
7
# Initialize Data Designer using the navigator_client and factory method
8
gretel_navigator = Gretel()  # This creates the navigator client
9
dd = gretel_navigator.data_designer.new(model_suite="apache-2.0")
10
11
# Add topic column (categorical sampler)
12
dd.add_column(
13
    C.SamplerColumn(
14
        name="topic",
15
        type=P.SamplerType.CATEGORY,
16
        params=P.CategorySamplerParams(
17
            values=[
18
                "neural networks", "deep learning", "machine learning", "NLP", 
19
                "computer vision", "reinforcement learning", "AI ethics", "data science"
20
            ]
21
        )
22
    )
23
)
24
25
# Add difficulty column
26
dd.add_column(
27
    C.SamplerColumn(
28
        name="difficulty",
29
        type=P.SamplerType.CATEGORY,
30
        params=P.CategorySamplerParams(
31
            values=["beginner", "intermediate", "advanced"]
32
        )
33
    )
34
)
35
36
# Add question column (LLM-generated)
37
dd.add_column(
38
    C.LLMTextColumn(
39
        name="question",
40
        prompt=(
41
            "Generate a challenging, specific question about {{ topic }} "
42
            "at {{ difficulty }} level. The question should be clear, focused, "
43
            "and something a student or practitioner might actually ask."
44
        )
45
    )
46
)
47
48
# Add answer column (LLM-generated)
49
dd.add_column(
50
    C.LLMTextColumn(
51
        name="answer",
52
        prompt=(
53
            "Provide a clear, accurate, and comprehensive answer to this {{ difficulty }}-level "
54
            "question about {{ topic }}: '{{ question }}'. The answer should be educational "
55
            "and directly address all aspects of the question."
56
        )
57
    )
58
)
59
60
print("📊 Generating Q&A dataset...")
61
62
# Generate the dataset
63
workflow_run = dd.create(num_records=20, wait_until_done=True)
64
synthetic_df = workflow_run.dataset.df
65
66
print(f"✅ Generated {len(synthetic_df)} Q&A pairs!")
67
print(f"\n📊 Dataset shape: {synthetic_df.shape}")
68
print(f"📋 Columns: {list(synthetic_df.columns)}")
69
70
# Display first few rows
71
print("\n📄 Sample data:")
72
synthetic_df.head(3)

1
def convert_to_opik_format(df):
2
    """Convert Gretel Q&A data to Opik dataset format"""
3
    opik_items = []
4
    
5
    for _, row in df.iterrows():
6
        # Create Opik dataset item
7
        item = {
8
            "input": {
9
                "question": row["question"]
10
            },
11
            "expected_output": row["answer"],
12
            "metadata": {
13
                "topic": row.get("topic", "AI/ML"),
14
                "difficulty": row.get("difficulty", "unknown"),
15
                "source": "gretel_navigator"
16
            }
17
        }
18
        opik_items.append(item)
19
    
20
    return opik_items
21
22
print("🔄 Converting to Opik format...")
23
24
opik_data = convert_to_opik_format(synthetic_df)
25
26
print(f"✅ Converted {len(opik_data)} items to Opik format!")
27
print("\n📋 Sample converted item:")
28
import json
29
print(json.dumps(opik_data[0], indent=2))

1
print("📤 Pushing dataset to Opik...")
2
3
# Initialize Opik client
4
opik_client = opik.Opik()
5
6
# Create the dataset
7
dataset_name = "gretel-ai-qa-dataset"
8
dataset = opik_client.get_or_create_dataset(
9
    name=dataset_name,
10
    description="Synthetic Q&A dataset generated using Gretel Data Designer for AI/ML evaluation"
11
)
12
13
# Insert the data
14
dataset.insert(opik_data)
15
16
print(f"✅ Successfully created dataset: {dataset.name}")
17
print(f"🆔 Dataset ID: {dataset.id}")
18
print(f"📊 Total items: {len(opik_data)}")

1
print("🔍 Verifying dataset creation...")
2
3
# Try to retrieve the dataset
4
try:
5
    retrieved_dataset = opik_client.get_dataset(dataset_name)
6
    print(f"✅ Dataset verified: {retrieved_dataset.name}")
7
    print(f"🆔 Dataset ID: {retrieved_dataset.id}")
8
    
9
    print(f"\n🎯 Next steps:")
10
    print(f"1. Go to https://www.comet.com")
11
    print(f"2. Navigate to Opik → Datasets")
12
    print(f"3. Find your dataset: {dataset_name}")
13
    print(f"4. Use it to evaluate your AI models!")
14
    
15
except Exception as e:
16
    print(f"❌ Could not verify dataset: {e}")
17
    print("Please check your Opik configuration and try again.")

1
# Example: Simple Q&A model evaluation
2
@opik.track
3
def simple_qa_model(input_data):
4
    """A simple example model that generates responses to questions"""
5
    question = input_data.get('question', '')
6
    
7
    # This is just an example - replace with your actual model
8
    if 'neural network' in question.lower():
9
        return "A neural network is a computational model inspired by biological neural networks."
10
    elif 'machine learning' in question.lower():
11
        return "Machine learning is a subset of AI that enables systems to learn from data."
12
    else:
13
        return "This is a complex AI/ML topic that requires detailed explanation."
14
15
print("🧪 Example model evaluation setup:")
16
print(f"Dataset: {dataset_name}")
17
print("Model: simple_qa_model (replace with your actual model)")
18
print("\n💡 To run evaluation, uncomment and run the following code:")
19
print("\n🎉 Integration complete! Your Gretel-generated dataset is ready for model evaluation in Opik.")

1
%%capture
2
%pip install -U gretel-client

1
import pandas as pd
2
from gretel_client.navigator_client import Gretel
3
4
# Initialize Gretel client
5
gretel = Gretel(api_key="prompt")
6
7
# Option 1: Use Gretel's sample ecommerce dataset (has 200+ records)
8
my_data_source = "https://gretel-datasets.s3.us-west-2.amazonaws.com/ecommerce_customers.csv"
9
10
# Option 2: Create your own Q&A dataset (needs 200+ records for holdout)
11
# For demonstration, we'll create a larger dataset
12
sample_questions = [
13
    'What is machine learning?',
14
    'How do neural networks work?',
15
    'What is the difference between AI and ML?',
16
    'Explain deep learning concepts',
17
    'What are the applications of NLP?'
18
] * 50  # Repeat to get 250 records
19
20
sample_answers = [
21
    'Machine learning is a subset of AI that enables systems to learn from data.',
22
    'Neural networks are computational models inspired by biological neural networks.',
23
    'AI is the broader concept while ML is a specific approach to achieve AI.',
24
    'Deep learning uses multi-layer neural networks to model complex patterns.',
25
    'NLP applications include chatbots, translation, sentiment analysis, and text generation.'
26
] * 50  # Repeat to get 250 records
27
28
sample_data = {
29
    'question': sample_questions,
30
    'answer': sample_answers,
31
    'topic': (['ML', 'Neural Networks', 'AI/ML', 'Deep Learning', 'NLP'] * 50),
32
    'difficulty': (['beginner', 'intermediate', 'beginner', 'advanced', 'intermediate'] * 50)
33
}
34
35
original_df = pd.DataFrame(sample_data)
36
print(f"📄 Original dataset: {len(original_df)} records")
37
print(original_df.head())
38
39
# Important: Gretel requires at least 200 records to use holdout
40
if len(original_df) < 200:
41
    print("⚠️ Warning: Dataset has less than 200 records. Holdout will be disabled.")

1
# For quick demo with small dataset - disable holdout and transform
2
synthetic_dataset = gretel.safe_synthetic_dataset \
3
    .from_data_source(original_df, holdout=None) \
4
    .synthesize(num_records=5) \
5
    .create()
6
7
# Wait for completion and get results
8
synthetic_dataset.wait_until_done()
9
synthetic_df_safe = synthetic_dataset.dataset.df
10
11
print(f"✅ Generated {len(synthetic_df_safe)} synthetic Q&A pairs using Safe Synthetics!")
12
print(synthetic_df_safe.head())

1
# Preview synthetic data
2
print("🔍 Synthetic dataset preview:")
3
print(synthetic_dataset.dataset.df.head())
4
5
# View quality report table
6
print("📊 Quality Report Summary:")
7
print(synthetic_dataset.report.table)
8
9
# View detailed HTML report in notebook
10
# synthetic_dataset.report.display_in_notebook()
11
12
# Access workflow details
13
print("\n🔧 Workflow Configuration:")
14
print(synthetic_dataset.config_yaml)
15
16
# List all workflow steps
17
print("\n📋 Workflow Steps:")
18
for step in synthetic_dataset.steps:
19
    print(f"- {step.name}")

1
def convert_to_opik_format(df):
2
    """Convert Gretel Q&A data to Opik dataset format"""
3
    opik_items = []
4
    
5
    for _, row in df.iterrows():
6
        # Create Opik dataset item
7
        item = {
8
            "input": {
9
                "question": row["question"]
10
            },
11
            "expected_output": row["answer"],
12
            "metadata": {
13
                "topic": row.get("topic", "AI/ML"),
14
                "difficulty": row.get("difficulty", "unknown"),
15
                "source": "gretel_navigator"
16
            }
17
        }
18
        opik_items.append(item)
19
    
20
    return opik_items
21
22
# Initialize Opik client if not already defined
23
opik_client = opik.Opik()
24
# Convert and upload to Opik (same process as before)
25
opik_data_safe = convert_to_opik_format(synthetic_df_safe)
26
27
# Create dataset in Opik
28
dataset_safe = opik_client.get_or_create_dataset(
29
    name="gretel-safe-synthetics-qa-dataset",
30
    description="Synthetic Q&A dataset generated using Gretel Safe Synthetics"
31
)
32
33
dataset_safe.insert(opik_data_safe)
34
print(f"✅ Safe Synthetics dataset created: {dataset_safe.name}")