pip install fenic

import fenic as fc

session = fc.Session.get_or_create(fc.SessionConfig())

import fenic as fc

session = fc.Session.get_or_create(fc.SessionConfig())

# Read a CSV file from a public dataset
df = session.read.csv("hf://datasets/datasets-examples/doc-formats-csv-1/data.csv")

# Read Parquet files using glob patterns
df = session.read.parquet("hf://datasets/cais/mmlu/astronomy/*.parquet")

# Read from a specific dataset revision
df = session.read.parquet("hf://datasets/datasets-examples/doc-formats-csv-1@~parquet/**/*.parquet")

# Read multiple CSV files with schema merging
df = session.read.csv("hf://datasets/username/dataset_name/*.csv", merge_schemas=True)

# Read multiple Parquet files with schema merging
df = session.read.parquet("hf://datasets/username/dataset_name/*.parquet", merge_schemas=True)

export HF_TOKEN="your_hugging_face_token_here"

hf://{repo_type}/{repo_id}/{path_to_file}

hf://{repo_type}/{repo_id}@{revision}/{path_to_file}

# Mix HF and local files in one read call
df = session.read.parquet([
    "hf://datasets/cais/mmlu/astronomy/*.parquet",
    "file:///local/data/*.parquet",
    "./relative/path/data.parquet"
])

import fenic as fc

session = fc.Session.get_or_create(fc.SessionConfig())

# Load IMDB dataset from Hugging Face
df = session.read.parquet("hf://datasets/imdb/plain_text/train-*.parquet")

# Filter and select
positive_reviews = df.filter(fc.col("label") == 1).select("text", "label")

# Group by and aggregate
label_counts = df.group_by("label").agg(
    fc.count("*").alias("count")
)

import fenic as fc

# Requires OPENAI_API_KEY to be set for language and embedding calls
session = fc.Session.get_or_create(
    fc.SessionConfig(
        semantic=fc.SemanticConfig(
            language_models={
                "gpt-4o-mini": fc.OpenAILanguageModel(
                    model_name="gpt-4o-mini",
                    rpm=60,
                    tpm=60000,
                )
            },
            embedding_models={
                "text-embedding-3-small": fc.OpenAIEmbeddingModel(
                    model_name="text-embedding-3-small",
                    rpm=60,
                    tpm=60000,
                )
            },
        )
    )
)

# Load a text dataset from Hugging Face
df = session.read.parquet("hf://datasets/imdb/plain_text/train-00000-of-00001.parquet")

# Add embeddings to text columns
df_with_embeddings = df.select(
    "*",
    fc.semantic.embed(fc.col("text")).alias("embedding")
)

# Apply semantic functions for sentiment analysis
df_analyzed = df_with_embeddings.select(
    "*",
    fc.semantic.analyze_sentiment(
        fc.col("text"),
        model_alias="gpt-4o-mini",  # Optional: specify model
    ).alias("sentiment")
)

import fenic as fc

# Requires OPENAI_API_KEY to be set for semantic calls
session = fc.Session.get_or_create(
    fc.SessionConfig(
        semantic=fc.SemanticConfig(
            language_models={
                "gpt-4o-mini": fc.OpenAILanguageModel(
                    model_name="gpt-4o-mini",
                    rpm=60,
                    tpm=60000,
                )
            },
        )
    )
)

# Load MMLU astronomy subset from Hugging Face
df = session.read.parquet("hf://datasets/cais/mmlu/astronomy/*.parquet")

# Process the data
processed_df = (df
    # Filter for specific criteria
    .filter(fc.col("subject") == "astronomy")
    # Select relevant columns
    .select("question", "choices", "answer")
    # Add difficulty analysis using semantic.map
    .select(
        "*",
        fc.semantic.map(
            "Rate the difficulty of this question from 1-5: {{question}}",
            question=fc.col("question"),
            model_alias="gpt-4o-mini"  # Optional: specify model
        ).alias("difficulty")
    )
)

# Show results
processed_df.show()