wait_visible / waitVisible waits for a terminal persisted trace, including an error trace. Evaluation wait requires usable scores for every requested trace/metric pair; terminal status alone is insufficient. Missing, duplicate, nonnumeric, or failed evidence does not become a passing result.
On this page
Evaluations, PlatformClient, Traces.Example
This creates a real evaluation run for an existing trace.trace_id is a persisted task trace UUID from the same project.
Evaluations
Constructor
Methods
metrics
metrics
Returns the supported trace metric metadata from the platform.
create
create
Creates one batch trace-evaluation run. Deduplicates trace IDs and metric names and requires nonempty inputs. Creation is never automatically retried.
get
get
Fetches one evaluation run and verifies its identity. A RUNNING or PENDING result is not complete evidence.
scores
scores
Fetches the score records for an evaluation run; use wait to require a complete usable set.
wait
wait
Poll one existing run; terminal status alone is not usable evidence.
PlatformClient
Constructor
Fields and properties
Methods
aclose
aclose
decode
decode
decode_many
decode_many
request
request
Traces
Constructor
Methods
get
get
Fetches one trace with span evidence and verifies that the returned trace ID matches the requested ID.
list
list
Returns a page of trace summaries. Limit must be 1–200 and offset must be nonnegative.
iterate
iterate
Iterates trace summaries in pages, stopping at the configured maximum or the end of the server result.
wait_visible
wait_visible
Wait for persisted, finished evidence, including task ERROR traces.
Signatures and fields are generated from the SDK source. See the import index for available entry points.

