aip_sdk.partitions
Partition authoring, resolution, and metric availability for agent-trace dataset versions.
A partition names a typed evaluation target — a session, a trace, or a specific span —
via a re-resolvable PartitionSelector. Create a partition
on a dataset version, list the partitions on a version, resolve a partition to the
concrete canonical objects it selects across the version's traces, and ask which trace
metrics can score it.
aip_sdk.partitions.DatasetMetricAvailability
aip_sdk.partitions.DatasetMetricAvailability
A single metric's verdict for one dataset version. reasons is empty exactly when
available is true.
aip_sdk.partitions.DatasetMetricAvailability.available
aip_sdk.partitions.DatasetMetricAvailability.available: bool
No docstring is defined in the source.
aip_sdk.partitions.DatasetMetricAvailability.model_config
aip_sdk.partitions.DatasetMetricAvailability.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.DatasetMetricAvailability.op_name
aip_sdk.partitions.DatasetMetricAvailability.op_name: str
No docstring is defined in the source.
aip_sdk.partitions.DatasetMetricAvailability.reasons
aip_sdk.partitions.DatasetMetricAvailability.reasons: list[DatasetMetricUnavailableDetail] = Field(default_factory=list[DatasetMetricUnavailableDetail])
No docstring is defined in the source.
aip_sdk.partitions.DatasetMetricUnavailableDetail
aip_sdk.partitions.DatasetMetricUnavailableDetail
One reason a dataset metric is unavailable, with an explanation naming what to fix.
aip_sdk.partitions.DatasetMetricUnavailableDetail.detail
aip_sdk.partitions.DatasetMetricUnavailableDetail.detail: str
No docstring is defined in the source.
aip_sdk.partitions.DatasetMetricUnavailableDetail.model_config
aip_sdk.partitions.DatasetMetricUnavailableDetail.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.DatasetMetricUnavailableDetail.reason_code
aip_sdk.partitions.DatasetMetricUnavailableDetail.reason_code: MetricErrorCode
No docstring is defined in the source.
aip_sdk.partitions.DatasetVersionMetricAvailability
aip_sdk.partitions.DatasetVersionMetricAvailability
Per-metric availability for one dataset version's own rows (not partition occurrences).
has_connection is echoed since it changes every verdict.
aip_sdk.partitions.DatasetVersionMetricAvailability.available_metrics
aip_sdk.partitions.DatasetVersionMetricAvailability.available_metrics: list[str]
The names of the metrics that can score this dataset version, in response order.
aip_sdk.partitions.DatasetVersionMetricAvailability.dataset_id
aip_sdk.partitions.DatasetVersionMetricAvailability.dataset_id: str
No docstring is defined in the source.
aip_sdk.partitions.DatasetVersionMetricAvailability.dataset_version_id
aip_sdk.partitions.DatasetVersionMetricAvailability.dataset_version_id: str
No docstring is defined in the source.
aip_sdk.partitions.DatasetVersionMetricAvailability.has_connection
aip_sdk.partitions.DatasetVersionMetricAvailability.has_connection: bool = True
No docstring is defined in the source.
aip_sdk.partitions.DatasetVersionMetricAvailability.metrics
aip_sdk.partitions.DatasetVersionMetricAvailability.metrics: list[DatasetMetricAvailability] = Field(default_factory=list[DatasetMetricAvailability])
No docstring is defined in the source.
aip_sdk.partitions.DatasetVersionMetricAvailability.model_config
aip_sdk.partitions.DatasetVersionMetricAvailability.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.GetOrCreatePartitionResult
aip_sdk.partitions.GetOrCreatePartitionResult
Outcome of get_or_create_partition().
A plain (partition, created) tuple still unpacks the same way, but named access
(result.partition/result.created) is also available.
Attributes
partitionPartition: The partition — freshly created, or an existing, selector-matching one of the requested name.createdbool:Truewhen this call createdpartition,Falsewhen an existing partition of this name was found and reused instead.
aip_sdk.partitions.GetOrCreatePartitionResult.created
aip_sdk.partitions.GetOrCreatePartitionResult.created: bool
No docstring is defined in the source.
aip_sdk.partitions.GetOrCreatePartitionResult.partition
aip_sdk.partitions.GetOrCreatePartitionResult.partition: Partition
No docstring is defined in the source.
aip_sdk.partitions.MetricAvailability
aip_sdk.partitions.MetricAvailability
A single metric's verdict for one partition on one dataset version.
reasons is empty exactly when available is true. The gates are independent, so a
metric short both a span kind and a ground-truth column reports both; within a gate the
list is not exhaustive, so clearing every listed reason can surface a further one. The
deployment gate contributes only when the report's deployment_checked is true.
aip_sdk.partitions.MetricAvailability.available
aip_sdk.partitions.MetricAvailability.available: bool
No docstring is defined in the source.
aip_sdk.partitions.MetricAvailability.model_config
aip_sdk.partitions.MetricAvailability.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.MetricAvailability.op_name
aip_sdk.partitions.MetricAvailability.op_name: str
No docstring is defined in the source.
aip_sdk.partitions.MetricAvailability.reasons
aip_sdk.partitions.MetricAvailability.reasons: list[MetricUnavailableDetail] = Field(default_factory=list[MetricUnavailableDetail])
No docstring is defined in the source.
aip_sdk.partitions.MetricErrorCode
aip_sdk.partitions.MetricErrorCode
Why a dataset (row-level) metric cannot score a dataset version. Same vocabulary as
the POST /runs gate, so a verdict here matches what a run attempt would hit.
aip_sdk.partitions.MetricErrorCode.OP_MISSING_COLUMNS
aip_sdk.partitions.MetricErrorCode.OP_MISSING_COLUMNS = 'OpMissingColumns'
No docstring is defined in the source.
aip_sdk.partitions.MetricErrorCode.OP_NOT_FOUND
aip_sdk.partitions.MetricErrorCode.OP_NOT_FOUND = 'OpNotFound'
No docstring is defined in the source.
aip_sdk.partitions.MetricErrorCode.OP_SCHEMA_INCOMPATIBLE
aip_sdk.partitions.MetricErrorCode.OP_SCHEMA_INCOMPATIBLE = 'OpSchemaIncompatible'
No docstring is defined in the source.
aip_sdk.partitions.MetricErrorCode.OP_TASK_TYPE_INCOMPATIBLE
aip_sdk.partitions.MetricErrorCode.OP_TASK_TYPE_INCOMPATIBLE = 'OpTaskTypeIncompatible'
No docstring is defined in the source.
aip_sdk.partitions.MetricErrorCode.SCORER_CONTRACT_MISMATCH
aip_sdk.partitions.MetricErrorCode.SCORER_CONTRACT_MISMATCH = 'ScorerContractMismatch'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableDetail
aip_sdk.partitions.MetricUnavailableDetail
One reason a metric is unavailable, with an explanation naming what to fix.
aip_sdk.partitions.MetricUnavailableDetail.detail
aip_sdk.partitions.MetricUnavailableDetail.detail: str
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableDetail.model_config
aip_sdk.partitions.MetricUnavailableDetail.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableDetail.reason_code
aip_sdk.partitions.MetricUnavailableDetail.reason_code: MetricUnavailableReason
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason
aip_sdk.partitions.MetricUnavailableReason
Why a trace metric cannot score a partition.
The first four are properties of the metric against the occurrence rows it would score;
the next four are structural — the partition's granularity, the span kinds its occurrences
carry, or (REQUIRED_SHAPE_UNSUPPORTED) the topology of the traces it resolves to, e.g. an
orchestrator delegating to subagent(s). PARTITION_HAS_NO_OCCURRENCES is a property of the
selector as a whole and answers "why is everything unavailable" for an over-narrowed
partition. OP_NOT_DEPLOYED is a property of the environment: the metric exists and fits,
but its scoring function is not running here, so a run would fail. Ask whoever operates your
platform to deploy it.
AMBIGUOUS_TRACE_GROUND_TRUTH is a property of the ground truth attached to the version.
agent.step_accuracy and agent.tool_selection_accuracy fall back to a session's pooled
ground truth when a trace carries no annotation of its own — correct for a single-trace
session, but a multi-trace one has no single trace whose steps or tool calls should be
compared against a trajectory authored for the whole conversation, so both metrics reject that
occurrence outright at scoring time rather than silently mis-scoring it (a FAILED row, not
a wrong number). Fix by attaching trace-scoped ground truth for the affected traces, or by
scoring the metric on a SESSION partition instead.
Reported only when every occurrence a TRACE partition selects would hit that fallback:
a partition mixing some traces that carry their own ground truth with others that would fall
back to the pooled value still reports available=True — the pooled-fallback traces are
still individually rejected if you run it (each becomes a FAILED row, not a wrong score),
so available=True never hides a bad number; a mixed partition just doesn't warn you in
advance which rows will fail.
aip_sdk.partitions.MetricUnavailableReason.AMBIGUOUS_TRACE_GROUND_TRUTH
aip_sdk.partitions.MetricUnavailableReason.AMBIGUOUS_TRACE_GROUND_TRUTH = 'AmbiguousTraceGroundTruth'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.OP_CONTRACT_INVALID
aip_sdk.partitions.MetricUnavailableReason.OP_CONTRACT_INVALID = 'OpContractInvalid'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.OP_MISSING_COLUMNS
aip_sdk.partitions.MetricUnavailableReason.OP_MISSING_COLUMNS = 'OpMissingColumns'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.OP_NOT_DEPLOYED
aip_sdk.partitions.MetricUnavailableReason.OP_NOT_DEPLOYED = 'OpNotDeployed'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.OP_NOT_FOUND
aip_sdk.partitions.MetricUnavailableReason.OP_NOT_FOUND = 'OpNotFound'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.OP_SCHEMA_INCOMPATIBLE
aip_sdk.partitions.MetricUnavailableReason.OP_SCHEMA_INCOMPATIBLE = 'OpSchemaIncompatible'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.PARTITION_HAS_NO_OCCURRENCES
aip_sdk.partitions.MetricUnavailableReason.PARTITION_HAS_NO_OCCURRENCES = 'PartitionHasNoOccurrences'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.PARTITION_TYPE_UNSUPPORTED
aip_sdk.partitions.MetricUnavailableReason.PARTITION_TYPE_UNSUPPORTED = 'PartitionTypeUnsupported'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.REQUIRED_KINDS_ABSENT
aip_sdk.partitions.MetricUnavailableReason.REQUIRED_KINDS_ABSENT = 'RequiredKindsAbsent'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.REQUIRED_SHAPE_UNSUPPORTED
aip_sdk.partitions.MetricUnavailableReason.REQUIRED_SHAPE_UNSUPPORTED = 'RequiredShapeUnsupported'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.SCORER_CONTRACT_MISMATCH
aip_sdk.partitions.MetricUnavailableReason.SCORER_CONTRACT_MISMATCH = 'ScorerContractMismatch'
No docstring is defined in the source.
aip_sdk.partitions.MetricUnavailableReason.TARGET_KIND_MISMATCH
aip_sdk.partitions.MetricUnavailableReason.TARGET_KIND_MISMATCH = 'TargetKindMismatch'
No docstring is defined in the source.
aip_sdk.partitions.Partition
aip_sdk.partitions.Partition
A partition persisted on a dataset version.
aip_sdk.partitions.Partition.created_at
aip_sdk.partitions.Partition.created_at: datetime | None = None
No docstring is defined in the source.
aip_sdk.partitions.Partition.dataset_version_id
aip_sdk.partitions.Partition.dataset_version_id: str
No docstring is defined in the source.
aip_sdk.partitions.Partition.id
aip_sdk.partitions.Partition.id: str
No docstring is defined in the source.
aip_sdk.partitions.Partition.model_config
aip_sdk.partitions.Partition.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.Partition.name
aip_sdk.partitions.Partition.name: str
No docstring is defined in the source.
aip_sdk.partitions.Partition.partition_type
aip_sdk.partitions.Partition.partition_type: PartitionType
No docstring is defined in the source.
aip_sdk.partitions.Partition.selector
aip_sdk.partitions.Partition.selector: PartitionSelector
No docstring is defined in the source.
aip_sdk.partitions.PartitionMetricAvailability
aip_sdk.partitions.PartitionMetricAvailability
Per-metric availability for one partition on one dataset version.
Carries verdicts only, never span payloads.
deployment_checked is False when the platform could not determine which metrics are
deployed in the environment. The verdicts are then about the metric's contract alone: an
available metric fits this partition but is not confirmed runnable, and no metric can
report OP_NOT_DEPLOYED. It is True on a healthy platform.
ground_truth_checked is the same shape for a different gate: False means the platform
could not read this TRACE partition's ground-truth annotation values, so
agent.step_accuracy/agent.tool_selection_accuracy silently report no occurrence
ambiguous on that one gate rather than a verified answer — see
MetricUnavailableReason.AMBIGUOUS_TRACE_GROUND_TRUTH. Always True when the partition
isn't TRACE-typed or neither of those two metrics was requested, since that gate could never
fire there regardless.
aip_sdk.partitions.PartitionMetricAvailability.available_metrics
aip_sdk.partitions.PartitionMetricAvailability.available_metrics: list[str]
The names of the metrics that can score this partition, in response order.
aip_sdk.partitions.PartitionMetricAvailability.dataset_version_id
aip_sdk.partitions.PartitionMetricAvailability.dataset_version_id: str
No docstring is defined in the source.
aip_sdk.partitions.PartitionMetricAvailability.deployment_checked
aip_sdk.partitions.PartitionMetricAvailability.deployment_checked: bool = True
No docstring is defined in the source.
aip_sdk.partitions.PartitionMetricAvailability.ground_truth_checked
aip_sdk.partitions.PartitionMetricAvailability.ground_truth_checked: bool = True
No docstring is defined in the source.
aip_sdk.partitions.PartitionMetricAvailability.metrics
aip_sdk.partitions.PartitionMetricAvailability.metrics: list[MetricAvailability] = Field(default_factory=list[MetricAvailability])
No docstring is defined in the source.
aip_sdk.partitions.PartitionMetricAvailability.model_config
aip_sdk.partitions.PartitionMetricAvailability.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.PartitionMetricAvailability.partition_id
aip_sdk.partitions.PartitionMetricAvailability.partition_id: str
No docstring is defined in the source.
aip_sdk.partitions.PartitionMetricAvailability.partition_type
aip_sdk.partitions.PartitionMetricAvailability.partition_type: PartitionType
No docstring is defined in the source.
aip_sdk.partitions.PartitionMigrationOutcome
aip_sdk.partitions.PartitionMigrationOutcome
One partition the base version carried, and what became of it on the derived version.
aip_sdk.partitions.PartitionMigrationOutcome.model_config
aip_sdk.partitions.PartitionMigrationOutcome.model_config = ConfigDict(extra='allow')
No docstring is defined in the source.
aip_sdk.partitions.PartitionMigrationOutcome.name
aip_sdk.partitions.PartitionMigrationOutcome.name: str
No docstring is defined in the source.
aip_sdk.partitions.PartitionMigrationOutcome.needs_attention
aip_sdk.partitions.PartitionMigrationOutcome.needs_attention: bool
Whether this target failed to follow the data across, so reason must be acted on.
aip_sdk.partitions.PartitionMigrationOutcome.outcome
aip_sdk.partitions.PartitionMigrationOutcome.outcome: PartitionOutcome
No docstring is defined in the source.
aip_sdk.partitions.PartitionMigrationOutcome.partition_id
aip_sdk.partitions.PartitionMigrationOutcome.partition_id: str | None = None
No docstring is defined in the source.
aip_sdk.partitions.PartitionMigrationOutcome.reason
aip_sdk.partitions.PartitionMigrationOutcome.reason: str | None = None
No docstring is defined in the source.
aip_sdk.partitions.PartitionMigrationOutcome.source_partition_id
aip_sdk.partitions.PartitionMigrationOutcome.source_partition_id: str
No docstring is defined in the source.
aip_sdk.partitions.PartitionOutcome
aip_sdk.partitions.PartitionOutcome
What became of a partition when a new dataset version was derived from the one it sat on.
A partition is pinned to a dataset version, so a derivation carries it across explicitly;
only NAME_COLLISION means the evaluation target did not follow the data.
aip_sdk.partitions.PartitionOutcome.ALREADY_PRESENT
aip_sdk.partitions.PartitionOutcome.ALREADY_PRESENT = 'already_present'
No docstring is defined in the source.
aip_sdk.partitions.PartitionOutcome.MIGRATED
aip_sdk.partitions.PartitionOutcome.MIGRATED = 'migrated'
No docstring is defined in the source.
aip_sdk.partitions.PartitionOutcome.NAME_COLLISION
aip_sdk.partitions.PartitionOutcome.NAME_COLLISION = 'name_collision'
No docstring is defined in the source.
aip_sdk.partitions.create_partition
aip_sdk.partitions.create_partition(dataset_id: str, version_id: str, *, name: str, selector: PartitionSelector, client: APIClient | None = None) -> Partition
Create a partition on a dataset version.
Parameters
dataset_idstr: Dataset ID.version_idstr: Dataset version ID to attach the partition to.namestr: Partition name, unique within the version. Surrounding whitespace is trimmed.selectorPartitionSelector: The selector naming the evaluation target.partition_typeon the created partition is derived from the selector. ASESSIONorTRACEselector carries notarget/ancestry_path; aSPANselector may use them to narrow which spans match, or omit them to select every span.clientAPIClient | None: Optional API client.
Returns
Partition: The created partition.
Raises
InvalidArgumentError: Ifnameis blank or longer than 255 characters, orselectoris inconsistent with its partition type.DuplicatePartitionError: If a partition with this name already exists on the version.AuthError: If credentials are missing or invalid.ForbiddenError: If the caller lacks editor rights on the version's workspace.NotFoundError: If the dataset version does not exist, or belongs to a workspace the caller is not a member of.UnprocessableEntityError: If the version is not an agent-trace version.ResponseParseError: If the platform's response does not match this SDK version.
aip_sdk.partitions.get_dataset_metric_availability
aip_sdk.partitions.get_dataset_metric_availability(dataset_id: str, version_id: str, *, metrics: Iterable[str] | None = None, has_connection: bool = True, client: APIClient | None = None) -> DatasetVersionMetricAvailability
Report which dataset (row-level) metrics can score a dataset version, and why the rest cannot.
Uses the same gate as POST /runs, so an available metric is one the run gate accepts.
For agent-trace datasets, use aip.get_partition_metric_availability() instead.
Parameters
dataset_idstr: Dataset ID.version_idstr: Dataset version ID to judge metrics against.metricsIterable[str] | None: Restrict the report to these metric names. An unmatched name is reported unavailable (OP_NOT_FOUND) rather than omitted.Nonereports every metric visible in the dataset's workspace.has_connectionbool: Whether inference will fill the produced columns (sut_response,retrieved_context) before scoring. Defaults toTrue. PassFalseto judge the version's at-rest columns alone.clientAPIClient | None: Optional API client.
Returns
DatasetVersionMetricAvailability: The per-metric verdicts; readavailable_metricsfor just the names that can run.
Raises
InvalidArgumentError: Ifmetricsis given but empty, or contains a blank name.AuthError: If credentials are missing or invalid.NotFoundError: If the dataset or dataset version does not exist, or the dataset belongs to a workspace the caller is not a member of.UnprocessableEntityError: If the dataset holds agent traces.ResponseParseError: If the platform's response does not match this SDK version.
aip_sdk.partitions.get_or_create_partition
aip_sdk.partitions.get_or_create_partition(dataset_id: str, version_id: str, *, name: str, selector: PartitionSelector, client: APIClient | None = None) -> GetOrCreatePartitionResult
Create a partition, or reuse the existing one if this name is already taken.
For a script that may run more than once against the same dataset version — a demo,
a notebook, a retry — this avoids choosing between failing on
DuplicatePartitionError and hand-rolling a list_partitions() lookup on catch.
Parameters
dataset_idstr: Dataset ID.version_idstr: Dataset version ID to attach the partition to.namestr: Partition name, unique within the version. Surrounding whitespace is trimmed.selectorPartitionSelector: The selector naming the evaluation target. Used only when creating; an existing partition of this name is reused only if its own stored selector matches this one — a name collision with a different selector raises rather than silently redirecting the evaluation to a different target.clientAPIClient | None: Optional API client.
Returns
GetOrCreatePartitionResult: AGetOrCreatePartitionResult(partition, created)—createdisTruewhenGetOrCreatePartitionResult: freshly created,Falsewhen an existing, selector-matching partition of thisGetOrCreatePartitionResult: name was found instead.
Raises
InvalidArgumentError: Ifnameis blank or longer than 255 characters, orselectoris inconsistent with its partition type.AuthError: If credentials are missing or invalid.ForbiddenError: If the caller lacks editor rights on the version's workspace.NotFoundError: If the dataset version does not exist or belongs to a workspace the caller is not a member of, or a name collision is reported but no partition with this name is listed on the version (e.g. it was deleted concurrently — retry).DuplicatePartitionError: If a partition of this name already exists with a different selector — the name is taken by something else, not an equivalent partition safe to reuse.UnprocessableEntityError: If the version is not an agent-trace version.ResponseParseError: If the platform's response does not match this SDK version.
aip_sdk.partitions.get_partition_metric_availability
aip_sdk.partitions.get_partition_metric_availability(dataset_id: str, version_id: str, partition_id: str, *, metrics: Iterable[str] | None = None, client: APIClient | None = None) -> PartitionMetricAvailability
Report which trace metrics can score a partition, and why the rest cannot.
Each metric is judged against the span kinds the partition's occurrences actually carry
and against the occurrence rows it would be scored on, so the verdict reflects this
partition rather than the dataset as a whole. A partition narrowed to nothing reports
PARTITION_HAS_NO_OCCURRENCES rather than an optimistic verdict.
A metric blocked with OP_MISSING_COLUMNS needs ground truth this version does not
carry: attach it with aip.attach_ground_truth(), then re-check against the derived
version and the target's new ID in migrated_partitions. One blocked with
OP_NOT_DEPLOYED fits this partition but is not running in this environment — nothing
about the data will unblock it, so ask whoever operates your platform to deploy it. One
blocked with AMBIGUOUS_TRACE_GROUND_TRUTH (agent.step_accuracy /
agent.tool_selection_accuracy only) is reported only when every occurrence a
TRACE partition selects would score against its session's pooled ground truth for a
session with more than one trace — both metrics reject that occurrence outright at scoring
time rather than silently mis-scoring it, so a mixed partition (some traces with their own
ground truth, some without) can still report available while the pooled-fallback traces
are individually refused as FAILED rows if you run it; available=True never hides a
bad number, a mixed partition just doesn't warn you in advance which rows will fail.
How closely this predicts run admission depends on the partition's granularity. For a
SPAN partition the two agree by construction: run creation resolves the same stored
selector and gates metrics against exactly these occurrences. For a TRACE or
SESSION partition they can diverge — there the partition ID is a coarse provenance
label on the results, since the selector narrows nothing when the whole trace or session
is already the scope, so run creation gates metrics dataset-wide instead. Score the
partition by passing its ID and the reported partition_type in a target to
aip.run(pipeline="trace_metric_invoke", targets=[aip.RunTarget(partition_id=..., partition_type=...)]).
Parameters
dataset_idstr: Dataset ID.version_idstr: Dataset version ID the partition is defined on.partition_idstr: ID of the partition to judge metrics against.metricsIterable[str] | None: Restrict the report to these metric names. A name that matches no trace metric is reported as unavailable withOP_NOT_FOUNDrather than omitted, so a stale selection gets an answer. WhenNone, every trace metric visible in the dataset's workspace is reported.clientAPIClient | None: Optional API client.
Returns
PartitionMetricAvailability: The per-metric verdicts; readavailable_metricsfor just the names that can run.PartitionMetricAvailability: Checkdeployment_checkedbefore treating those names as confirmed runnable — whenPartitionMetricAvailability: it isFalsethe platform could not determine what is deployed, and the verdictsPartitionMetricAvailability: describe the metrics' contracts alone. Checkground_truth_checkedtoo when thePartitionMetricAvailability: partition is TRACE-typed and the report includesagent.step_accuracyorPartitionMetricAvailability:agent.tool_selection_accuracy— when it isFalsethe platform could not read thePartitionMetricAvailability: version's ground-truth values, so neither metric'sAMBIGUOUS_TRACE_GROUND_TRUTHgatePartitionMetricAvailability: was actually evaluated.
Raises
InvalidArgumentError: Ifmetricsis given but empty, or contains a blank name.AuthError: If credentials are missing or invalid.NotFoundError: If the partition or dataset version does not exist, or the version belongs to a workspace the caller is not a member of.UnprocessableEntityError: If the version is not an agent-trace version.ResponseParseError: If the platform's response does not match this SDK version.
aip_sdk.partitions.list_partitions
aip_sdk.partitions.list_partitions(dataset_id: str, version_id: str, *, client: APIClient | None = None) -> list[Partition]
List the partitions defined on a dataset version.
Parameters
dataset_idstr: Dataset ID.version_idstr: Dataset version ID.clientAPIClient | None: Optional API client.
Returns
list[Partition]: The version's partitions, oldest first.
Raises
AuthError: If credentials are missing or invalid.NotFoundError: If the dataset version does not exist.ResponseParseError: If the platform's response does not match this SDK version.
aip_sdk.partitions.resolve_partition
aip_sdk.partitions.resolve_partition(dataset_id: str, version_id: str, partition_id: str, *, client: APIClient | None = None) -> list[ResolvedPartition]
Resolve a partition to the concrete canonical targets it selects.
Each occurrence yields one resolved partition — a SESSION selector resolves to a
single session, a TRACE selector to one trace per trace in the version, and a
SPAN selector to one span per matching span.
Every occurrence is returned in one response, each carrying its full canonical payload
(a span's messages, a trace's spans), so cost scales with the version and the breadth of
the selector rather than with a page size. Narrow the selector — a target and, if
needed, an ancestry_path — when resolving a large trace version. To ask only which
metrics could score the partition, use get_partition_metric_availability(), which
returns verdicts instead of payloads.
Parameters
dataset_idstr: Dataset ID.version_idstr: Dataset version ID.partition_idstr: ID of the partition to resolve.clientAPIClient | None: Optional API client.
Returns
list[ResolvedPartition]: The resolved partitions, each alist[ResolvedPartition]: class:~aip_sdk.trace_models.SessionPartition,list[ResolvedPartition]: class:~aip_sdk.trace_models.TracePartition, orlist[ResolvedPartition]: class:~aip_sdk.trace_models.SpanPartitiontagged bypartition_type.
Raises
AuthError: If credentials are missing or invalid.NotFoundError: If the partition or dataset version does not exist.UnprocessableEntityError: If the version is not an agent-trace version.ResponseParseError: If the platform's response does not match this SDK version.