Migrate Pandas Client

The Pandas Client is used for batch logging of LLM traces, evaluations, and traditional ML model predictions using pandas DataFrames.

from arize.pandas.logger import Client
client = Client(...)

from arize import ArizeClient
client = ArizeClient(...)

log

The log() method migrates from client.log() to client.ml.log().

Parameter Reference

This table provides a complete mapping of all parameters between v7 and v8, including which parameters were removed, renamed, or remain unchanged.

Parameter	v7	v8	Changes
`space_id`	Client init	Required per call	Must pass explicitly
`model_id`	Required	Required	Renamed to `model_name`
`dataframe`	Required	Required	—
`schema`	Required	Required	—
`environment`	Required	Required	—
`model_type`	Required	Required	—
`model_version`	Optional	Optional	—
`batch_id`	Optional	Optional	—
`validate`	Optional	Optional	—
`metrics_validation`	Optional	Optional	—
`surrogate_explainability`	Optional	Optional	—
`timeout`	Optional	Optional	—
`path`	Optional	Optional	Renamed to `tmp_dir` for clarity
`sync`	Optional	❌ Removed	No longer supported
`verbose`	Optional	❌ Removed	Use Python logging config instead

Side-by-Side Comparison

See the complete migration in action with this example showing both client initialization and logging model predictions.

from arize.pandas.logger import Client
from arize.utils.types import Environments, ModelTypes, Schema

# Client initialization
client = Client(
    api_key="your-api-key",
    space_id="your-space-id"
)

# Logging model predictions
response = client.log(
    dataframe=predictions_df,
    schema=schema,
    environment=Environments.PRODUCTION,
    model_id="my-model",
    model_type=ModelTypes.BINARY_CLASSIFICATION,
    model_version="v1.0",
    batch_id="batch-123",
    metrics_validation=[Metrics.CLASSIFICATION],
    validate=True,
    path="/tmp/arize",
    surrogate_explainability=False,
    timeout=30.0,
    sync=False,
    verbose=True
)

from arize import ArizeClient
from arize.ml.types import Environments, ModelTypes, Schema

# Client initialization
client = ArizeClient(api_key="your-api-key")

# Logging model predictions
response = client.ml.log(
    space_id="your-space-id",  # Now required per call
    model_name="my-model",  # Renamed from model_id
    dataframe=predictions_df,
    schema=schema,
    environment=Environments.PRODUCTION,
    model_type=ModelTypes.BINARY_CLASSIFICATION,
    model_version="v1.0",
    batch_id="batch-123",
    metrics_validation=[Metrics.CLASSIFICATION],
    validate=True,
    tmp_dir="/tmp/arize",  # Renamed from path
    surrogate_explainability=False,
    timeout=30.0
    # sync parameter removed
    # verbose parameter removed
)

log_spans

The log_spans() method migrates from client.log_spans() to client.spans.log().

Parameter Reference

This table provides a complete mapping of all parameters between v7 and v8, including which parameters were removed, renamed, or remain unchanged.

Parameter	v7	v8	Notes / Changes
`space_id`	Client init	Required per call	Must pass explicitly
`project_name`	Optional	Required	Must pass explicitly
`dataframe`	Required	Required	—
`evals_dataframe`	Optional	Optional	—
`datetime_format`	Optional	Optional	—
`validate`	Optional	Optional	—
`timeout`	Optional	Optional	—
`model_id`	Deprecated	❌ Removed	Use `project_name` instead
`model_version`	Optional	❌ Removed	No longer supported
`path`	Optional	Optional	Renamed to `tmp_dir` for clarity
`verbose`	Optional	❌ Removed	Use Python logging config instead

Side-by-Side Comparison

See the complete migration in action with this example showing both client initialization and logging spans with all common parameters.

from arize.pandas.logger import Client

# Client initialization
client = Client(
    api_key="your-api-key",
    space_id="your-space-id"
)

# Logging spans
response = client.log_spans(
    dataframe=spans_df,
    model_id="my-project",  # or project_name
    model_version="v1.0",   # optional
    evals_dataframe=evals_df,  # optional
    datetime_format="%Y-%m-%dT%H:%M:%S.%f+00:00",
    validate=True,
    path="/tmp/arize",
    timeout=30.0,
    verbose=True,
    project_name="my-project"  # alternative to model_id
)

from arize import ArizeClient

# Client initialization
client = ArizeClient(api_key="your-api-key")

# Logging spans
response = client.spans.log(
    space_id="your-space-id",  # Now required per call
    project_name="my-project",  # Required, no model_id alternative
    dataframe=spans_df,
    evals_dataframe=evals_df,  # optional
    datetime_format="%Y-%m-%dT%H:%M:%S.%f+00:00",
    validate=True,
    tmp_dir="/tmp/arize",  # renamed from 'path'
    timeout=30.0
    # verbose parameter removed
    # model_version removed
)

log_evaluations & log_evaluations_sync

In v7, there were two separate methods for logging evaluations: log_evaluations() (async HTTP) and log_evaluations_sync() (sync gRPC). In v8, both consolidate into a single update_evaluations() method that uses gRPC by default with an optional force_http parameter. Both v7 methods (log_evaluations and log_evaluations_sync) now use the same v8 method: update_evaluations(), which uses Arrow Flight (gRPC) by default, matching the behavior of v7’s log_evaluations_sync(). Use force_http=True if you need HTTP transport (not recommended for large payloads).

Parameter Reference

This table provides a complete mapping of all parameters between v7 and v8, including which parameters were removed, renamed, or remain unchanged.

Parameter	v7	v8	Notes / Changes
`space_id`	Client init	Required per call	Must pass explicitly
`project_name`	Optional	Required	Must pass explicitly
`dataframe`	Required	Required	—
`validate`	Optional	Optional	—
`timeout`	Optional	Optional	—
`model_id`	Deprecated	❌ Removed	Use `project_name` instead
`model_version`	Optional	❌ Removed	No longer supported
`verbose`	Optional	❌ Removed	Use Python logging config instead
`tmp_dir`	N/A	✅ New	Temporary directory for file storage
`force_http`	N/A	✅ New	Force HTTP transport (defaults to `False`)

Side-by-Side Comparison

See the complete migration in action with this example showing both client initialization and logging evaluations synchronously.

log_evaluations_sync

from arize.pandas.logger import Client

# Client initialization
client = Client(
    api_key="your-api-key",
    space_id="your-space-id"
)

# Logging evaluations (sync via gRPC)
response = client.log_evaluations_sync(
    dataframe=evals_df,
    model_id="my-project",  # or project_name
    model_version="v1.0",   # optional
    validate=True,
    timeout=30.0,
    verbose=True,
    project_name="my-project"  # alternative to model_id
)

from arize import ArizeClient

# Client initialization
client = ArizeClient(api_key="your-api-key")

# Logging evaluations (gRPC by default)
response = client.spans.update_evaluations(
    space_id="your-space-id",  # Now required per call
    project_name="my-project",  # Required, no model_id alternative
    dataframe=evals_df,
    validate=True,
    timeout=30.0,
    tmp_dir="/tmp/arize",  # optional, new parameter
    force_http=False  # Use gRPC like v7's log_evaluations_sync
    # verbose parameter removed
    # model_version removed
)

log_evaluations

from arize.pandas.logger import Client

# Client initialization
client = Client(
    api_key="your-api-key",
    space_id="your-space-id"
)

# Logging evaluations
response = client.log_evaluations(
    dataframe=evals_df,
    model_id="my-project",  # or project_name
    model_version="v1.0",   # optional
    validate=True,
    path="/tmp/arize",
    timeout=30.0,
    verbose=True,
    project_name="my-project"  # alternative to model_id
)

from arize import ArizeClient

# Client initialization
client = ArizeClient(api_key="your-api-key")

# Logging evaluations
response = client.spans.update_evaluations(
    space_id="your-space-id",  # Now required per call
    project_name="my-project",  # Required, no model_id alternative
    dataframe=evals_df,
    validate=True,
    tmp_dir="/tmp/arize",  # renamed from 'path'
    timeout=30.0,
    force_http=True  # Use HTTP like v7's log_evaluations
    # verbose parameter removed
    # model_version removed
)

log_annotations

The log_annotations() method migrates from client.log_annotations() to client.spans.update_annotations().

Parameter Reference

This table provides a complete mapping of all parameters between v7 and v8, including which parameters were removed, renamed, or remain unchanged.

Parameter	v7	v8	Notes / Changes
`space_id`	Client init	Required per call	Must pass explicitly
`project_name`	Required	Required	—
`dataframe`	Required	Required	—
`validate`	Optional	Optional	—
`verbose`	Optional	❌ Removed	Use Python logging config instead

Side-by-Side Comparison

See the complete migration in action with this example showing both client initialization and logging annotations.

from arize.pandas.logger import Client

# Client initialization
client = Client(
    api_key="your-api-key",
    space_id="your-space-id"
)

# Logging annotations
response = client.log_annotations(
    dataframe=annotations_df,
    project_name="my-project",
    validate=True,
    verbose=True
)

from arize import ArizeClient

# Client initialization
client = ArizeClient(api_key="your-api-key")

# Logging annotations
response = client.spans.update_annotations(
    space_id="your-space-id",  # Now required per call
    project_name="my-project",
    dataframe=annotations_df,
    validate=True
    # verbose parameter removed
)

log_metadata

The log_metadata() method migrates from client.log_metadata() to client.spans.update_metadata().

Parameter Reference

This table provides a complete mapping of all parameters between v7 and v8, including which parameters were removed, renamed, or remain unchanged.

Parameter	v7	v8	Changes
`space_id`	Client init	Required per call	Must pass explicitly
`project_name`	Required	Required	—
`dataframe`	Required	Required	—
`patch_document_column_name`	Optional	Optional	—
`validate`	Optional	Optional	—
`verbose`	Optional	❌ Removed	Use Python logging config instead

Side-by-Side Comparison

See the complete migration in action with this example showing both client initialization and logging metadata updates.

from arize.pandas.logger import Client

# Client initialization
client = Client(
    api_key="your-api-key",
    space_id="your-space-id"
)

# Logging metadata
response = client.log_metadata(
    dataframe=metadata_df,
    project_name="my-project",
    patch_document_column_name="patch_document",
    validate=True,
    verbose=True
)

from arize import ArizeClient

# Client initialization
client = ArizeClient(api_key="your-api-key")

# Logging metadata
response = client.spans.update_metadata(
    space_id="your-space-id",  # Now required per call
    project_name="my-project",
    dataframe=metadata_df,
    patch_document_column_name="patch_document",
    validate=True
    # verbose parameter removed
)

API reference

API clients

log

Parameter Reference

Side-by-Side Comparison

log_spans

Parameter Reference

Side-by-Side Comparison

log_evaluations & log_evaluations_sync

Parameter Reference

Side-by-Side Comparison

log_evaluations_sync

log_evaluations

log_annotations

Parameter Reference

Side-by-Side Comparison

log_metadata

Parameter Reference

Side-by-Side Comparison

​log

​Parameter Reference

​Side-by-Side Comparison

​log_spans

​Parameter Reference

​Side-by-Side Comparison

​log_evaluations & log_evaluations_sync

​Parameter Reference

​Side-by-Side Comparison

​log_evaluations_sync

​log_evaluations

​log_annotations

​Parameter Reference

​Side-by-Side Comparison

​log_metadata

​Parameter Reference

​Side-by-Side Comparison

log

Parameter Reference

Side-by-Side Comparison

log_spans

Parameter Reference

Side-by-Side Comparison

log_evaluations & log_evaluations_sync

Parameter Reference

Side-by-Side Comparison

log_evaluations_sync

log_evaluations

log_annotations

Parameter Reference

Side-by-Side Comparison

log_metadata

Parameter Reference

Side-by-Side Comparison