Fount-Core 0.1.7.4

No project description provided

Fount SDK API Reference

Quick Reference

Client Initialization

from fount import Fount

# Basic usage
client = Fount()

# With custom transport
client = Fount(transport=custom_httpx_client)

# As context manager
with Fount() as client:
    # auto cleanup
    pass

Environment Variables

• FOUNT_API_KEY - Your API key (required)
• FOUNT_BASE_URL - API base URL (default: https://fountondev.datapoem.ai/api/v1/)

Class: Fount

Methods

upload_dataframe(dataframe, name=None) -> Dataset

Upload a pandas DataFrame.

Parameters:

• dataframe (pd.DataFrame): DataFrame to upload
• name (str, optional): Dataset name

Returns: Dataset object with ID

---

upload_csv(pathname, name) -> Dataset

Upload a CSV file.

Parameters:

• pathname (str): Path to CSV file
• name (str, optional): Dataset name

Returns: Dataset object with ID

---

upload_excel(pathname, sheet_name, name) -> Dataset

Upload Excel file (not implemented).

Parameters:

• pathname (str): Path to Excel file
• sheet_name (str): Sheet to extract
• name (str, optional): Dataset name

---

train(dataset, categorical_cols, date_column, target_columns, validation_data_required, validation_split, time_granularity, **kwargs) -> TrainingJob

Train a model.

Parameters:

• dataset (Dataset): Training dataset
• categorical_cols (List[str]): Categorical column names
• date_column (str): Date column name
• target_columns (List[str]): Target column names
• validation_data_required (bool): Create validation set
• validation_split (float): Validation split ratio (0-1)
• time_granularity (str): Time interval or frequency
• **kwargs: Additional parameters (epochs, learning_rate, etc.)

Returns: TrainingJob object

---

tune(categorical_cols, date_column, target_columns, validation_data_required, validation_split, time_granularity, **kwargs) -> TuningJob

Hyperparameter tuning.

Parameters:

• categorical_cols (List[str]): Categorical column names
• date_column (str): Date column name
• target_columns (List[str]): Target column names
• validation_data_required (bool): Create validation set
• validation_split (float): Validation split ratio (0-1)
• time_granularity (str): Time interval or frequency
• **kwargs: Additional parameters (param_grid, n_trials, etc.)

Returns: TuningJob object

---

inference(model_name, batch_size, **kwargs) -> InferenceJob

Run batch inference.

Parameters:

• model_name (str): Model name/ID
• batch_size (int): Batch size
• **kwargs: Additional parameters (input_dataset, output_format, etc.)

Returns: InferenceJob object

---

close() -> None

Close client and release resources.

---

Job Classes

All job classes (TrainingJob, TuningJob, InferenceJob) have:

Methods

status() -> Dict

Get job status.

Returns: Dict with:

• state: Current state (pending, running, completed, failed)
• progress: Progress percentage
• message: Status message

metrics() -> Dict

Get job results/metrics.

Returns: Dict with job-specific metrics

Exceptions

• SDKError - Base SDK exception
• AuthenticationError - Auth failures
• RateLimitError - Rate limit exceeded
• UploadError - Upload failures
• TrainingError - Training failures
• InferenceError - Inference failures
• TuningError - Tuning failures

Common Patterns

Training Pipeline

# Upload data
df = pd.read_csv("data.csv")
dataset = client.upload_dataframe(df)

# Train model
job = client.train(
    dataset=dataset,
    categorical_cols=["cat1", "cat2"],
    date_column="date",
    target="target",
    validation_data_required=True,
    validation_split=0.2
)

# Monitor progress
while True:
    status = job.status()
    if status["state"] in ["completed", "failed"]:
        break
    time.sleep(30)

# Get results
if status["state"] == "completed":
    metrics = job.metrics()

Batch Inference

# Run inference
job = client.inference(
    model_name="my_model",
    batch_size=1000,
    input_dataset=dataset.id
)

# Get predictions
results = job.metrics()
predictions = results["predictions"]

Error Handling

from fount.errors import AuthenticationError, UploadError

try:
    dataset = client.upload_dataframe(df)
except AuthenticationError:
    print("Invalid API key")
except UploadError as e:
    print(f"Upload failed: {e}")