Skip to main content
The Exporter functionality has been restructured in v8. Instead of a dedicated ArizeExportClient, export methods are now integrated directly into the unified ArizeClient via resource-specific methods. 
from arize.exporter import ArizeExportClient

client = ArizeExportClient(api_key="your-api-key")

​
Exporting Spans/Traces (LLM Data)

In v8, span export methods are available on client.spans.

​
export_model_to_df() for Spans

The export_model_to_df() method for tracing data migrates to client.spans.export_to_df().

​
Parameter Reference

Parameterv7v8Changes
space_idRequiredRequired—
model_idRequiredRequiredRenamed to project_name
project_nameN/A✅ RequiredRenamed from model_id
environmentRequired❌ RemovedAlways TRACING for spans
start_timeRequiredRequired—
end_timeRequiredRequired—
include_actualsOptional❌ RemovedNot applicable to spans
model_versionOptional❌ RemovedNot applicable to spans
batch_idOptional❌ RemovedNot applicable to spans
whereOptionalOptional—
similarity_search_paramsOptionalOptional—
columnsOptionalOptional—
stream_chunk_sizeOptionalOptional—
parallelize_exportsOptional❌ RemovedNo longer supported

​
Side-by-Side Comparison

from arize.exporter import ArizeExportClient
from arize.utils.types import Environments
from datetime import datetime

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

# Export spans/traces to DataFrame
df = client.export_model_to_df(
    space_id="your-space-id",
    model_id="my-llm-project",
    environment=Environments.TRACING,
    start_time=datetime(2024, 1, 1),
    end_time=datetime(2024, 1, 31),
    where="span.name = 'generate'",
    columns=["span_id", "parent_span_id", "span.name"],
    stream_chunk_size=1000,
    parallelize_exports=True
)

​
export_model_to_parquet() for Spans

The export_model_to_parquet() method for tracing data migrates to client.spans.export_to_parquet().

​
Parameter Reference

Parameterv7v8Changes
pathRequiredRequired—
space_idRequiredRequired—
model_idRequiredRequiredRenamed to project_name
project_nameN/A✅ RequiredRenamed from model_id
environmentRequired❌ RemovedAlways TRACING for spans
start_timeRequiredRequired—
end_timeRequiredRequired—
include_actualsOptional❌ RemovedNot applicable to spans
model_versionOptional❌ RemovedNot applicable to spans
batch_idOptional❌ RemovedNot applicable to spans
whereOptionalOptional—
similarity_search_paramsOptionalOptional—
columnsOptionalOptional—
stream_chunk_sizeOptionalOptional—
parallelize_exportsOptional❌ RemovedNo longer supported

​
Side-by-Side Comparison

from arize.exporter import ArizeExportClient
from arize.utils.types import Environments
from datetime import datetime

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

# Export spans/traces to Parquet
client.export_model_to_parquet(
    path="/path/to/output.parquet",
    space_id="your-space-id",
    model_id="my-llm-project",
    environment=Environments.TRACING,
    start_time=datetime(2024, 1, 1),
    end_time=datetime(2024, 1, 31),
    where="span.name = 'generate'",
    columns=["span_id", "parent_span_id", "span.name"],
    stream_chunk_size=1000,
    parallelize_exports=True
)

​
Exporting Models (Traditional ML Data)

In v8, model export methods are available on client.ml.

​
export_model_to_df() for Models

The export_model_to_df() method for traditional ML models migrates to client.ml.export_to_df().

​
Parameter Reference

Parameterv7v8Changes
space_idRequiredRequired—
model_idRequiredRequiredRenamed to model_name
model_nameN/A✅ RequiredRenamed from model_id
environmentRequiredRequired—
start_timeRequiredRequired—
end_timeRequiredRequired—
include_actualsOptionalOptional—
model_versionOptionalOptional—
batch_idOptionalOptional—
whereOptionalOptional—
similarity_search_paramsOptionalOptional—
columnsOptionalOptional—
stream_chunk_sizeOptionalOptional—
parallelize_exportsOptional❌ RemovedNo longer supported

​
Side-by-Side Comparison

from arize.exporter import ArizeExportClient
from arize.utils.types import Environments
from datetime import datetime

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

# Export model data to DataFrame
df = client.export_model_to_df(
    space_id="your-space-id",
    model_id="fraud-detection",
    environment=Environments.PRODUCTION,
    start_time=datetime(2024, 1, 1),
    end_time=datetime(2024, 1, 31),
    include_actuals=True,
    model_version="v1.0",
    where="prediction_score > 0.8",
    columns=["prediction_id", "prediction_label", "actual_label"],
    stream_chunk_size=1000,
    parallelize_exports=True
)

​
export_model_to_parquet() for Models

The export_model_to_parquet() method for traditional ML models migrates to client.ml.export_to_parquet().

​
Parameter Reference

Parameterv7v8Changes
pathRequiredRequired—
space_idRequiredRequired—
model_idRequiredRequiredRenamed to model_name
model_nameN/A✅ RequiredRenamed from model_id
environmentRequiredRequired—
start_timeRequiredRequired—
end_timeRequiredRequired—
include_actualsOptionalOptional—
model_versionOptionalOptional—
batch_idOptionalOptional—
whereOptionalOptional—
similarity_search_paramsOptionalOptional—
columnsOptionalOptional—
stream_chunk_sizeOptionalOptional—
parallelize_exportsOptional❌ RemovedNo longer supported

​
Side-by-Side Comparison

from arize.exporter import ArizeExportClient
from arize.utils.types import Environments
from datetime import datetime

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

# Export model data to Parquet
client.export_model_to_parquet(
    path="/path/to/output.parquet",
    space_id="your-space-id",
    model_id="fraud-detection",
    environment=Environments.PRODUCTION,
    start_time=datetime(2024, 1, 1),
    end_time=datetime(2024, 1, 31),
    include_actuals=True,
    model_version="v1.0",
    where="prediction_score > 0.8",
    columns=["prediction_id", "prediction_label", "actual_label"],
    stream_chunk_size=1000,
    parallelize_exports=True
)