arize-phoenix 4.0.1__tar.gz → 4.0.3__tar.gz

{arize_phoenix-4.0.1 → arize_phoenix-4.0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: arize-phoenix
-Version: 4.0.1
+Version: 4.0.3
 Summary: AI Observability and Evaluation
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues

{arize_phoenix-4.0.1 → arize_phoenix-4.0.3}/src/phoenix/db/bulk_inserter.py

@@ -3,7 +3,7 @@ import logging
 from dataclasses import dataclass, field
 from datetime import datetime, timezone
 from itertools import islice
-from time import perf_counter, time
+from time import perf_counter
 from typing import (
     Any,
     AsyncContextManager,
@@ -48,21 +48,20 @@ class BulkInserter:
         cache_for_dataloaders: Optional[CacheForDataLoaders] = None,
         initial_batch_of_spans: Optional[Iterable[Tuple[Span, str]]] = None,
         initial_batch_of_evaluations: Optional[Iterable[pb.Evaluation]] = None,
-        run_interval_in_seconds: float = 2,
+        sleep: float = 0.1,
         max_num_per_transaction: int = 1000,
         enable_prometheus: bool = False,
     ) -> None:
         """
         :param db: A function to initiate a new database session.
         :param initial_batch_of_spans: Initial batch of spans to insert.
-        :param run_interval_in_seconds: The time interval between the starts of each
-        bulk insert. If there's nothing to insert, the inserter goes back to sleep.
+        :param sleep: The time to sleep between bulk insertions
         :param max_num_per_transaction: The maximum number of items to insert in a single
         transaction. Multiple transactions will be used if there are more items in the batch.
         """
         self._db = db
         self._running = False
-        self._run_interval_seconds = run_interval_in_seconds
+        self._sleep = sleep
         self._max_num_per_transaction = max_num_per_transaction
         self._spans: List[Tuple[Span, str]] = (
             [] if initial_batch_of_spans is None else list(initial_batch_of_spans)
@@ -98,11 +97,10 @@ class BulkInserter:
 
     async def _bulk_insert(self) -> None:
         spans_buffer, evaluations_buffer = None, None
-        next_run_at = time() + self._run_interval_seconds
+        # start first insert immediately if the inserter has not run recently
         while self._spans or self._evaluations or self._running:
-            await asyncio.sleep(next_run_at - time())
-            next_run_at = time() + self._run_interval_seconds
             if not (self._spans or self._evaluations):
+                await asyncio.sleep(self._sleep)
                 continue
             # It's important to grab the buffers at the same time so there's
             # no race condition, since an eval insertion will fail if the span
@@ -128,6 +126,7 @@ class BulkInserter:
                 evaluations_buffer = None
             for project_rowid in transaction_result.updated_project_rowids:
                 self._last_updated_at_by_project[project_rowid] = datetime.now(timezone.utc)
+            await asyncio.sleep(self._sleep)
 
     async def _insert_spans(self, spans: List[Tuple[Span, str]]) -> TransactionResult:
         transaction_result = TransactionResult()

{arize_phoenix-4.0.1 → arize_phoenix-4.0.3}/src/phoenix/trace/trace_dataset.py

@@ -16,6 +16,7 @@ from pyarrow import Schema, Table, parquet
 
 from phoenix.config import DATASET_DIR, GENERATED_DATASET_NAME_PREFIX, TRACE_DATASET_DIR
 from phoenix.datetime_utils import normalize_timestamps
+from phoenix.trace.attributes import unflatten
 from phoenix.trace.errors import InvalidParquetMetadataError
 from phoenix.trace.schemas import ATTRIBUTE_PREFIX, CONTEXT_PREFIX, Span
 from phoenix.trace.span_evaluations import Evaluations, SpanEvaluations
@@ -161,13 +162,13 @@ class TraceDataset:
         for _, row in self.dataframe.iterrows():
             is_attribute = row.index.str.startswith(ATTRIBUTE_PREFIX)
             attribute_keys = row.index[is_attribute]
-            attributes = (
+            attributes = unflatten(
                 row.loc[is_attribute]
                 .rename(
                     {key: key[len(ATTRIBUTE_PREFIX) :] for key in attribute_keys},
                 )
                 .dropna()
-                .to_dict()
+                .items()
             )
             is_context = row.index.str.startswith(CONTEXT_PREFIX)
             context_keys = row.index[is_context]

arize_phoenix-4.0.3/src/phoenix/version.py

@@ -0,0 +1 @@
+__version__ = "4.0.3"

arize_phoenix-4.0.1/src/phoenix/version.py

@@ -1 +0,0 @@
-__version__ = "4.0.1"