arize-phoenix 4.0.0__tar.gz → 4.0.2__tar.gz

{arize_phoenix-4.0.0 → arize_phoenix-4.0.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: arize-phoenix
-Version: 4.0.0
+Version: 4.0.2
 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-4.0.0 → arize_phoenix-4.0.2}/src/phoenix/db/bulk_inserter.py

@@ -63,6 +63,10 @@ class BulkInserter:
         self._db = db
         self._running = False
         self._run_interval_seconds = run_interval_in_seconds
+
+        # tracks time between insertions to improve responsiveness for small batches
+        self._last_insertion_time = time() - self._run_interval_seconds
+
         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)
@@ -98,10 +102,10 @@ 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
+            next_run_at = self._last_insertion_time + self._run_interval_seconds
+            await asyncio.sleep(max(next_run_at - time(), 0))
             if not (self._spans or self._evaluations):
                 continue
             # It's important to grab the buffers at the same time so there's
@@ -128,6 +132,7 @@ 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)
+            self._last_insertion_time = time()
 
     async def _insert_spans(self, spans: List[Tuple[Span, str]]) -> TransactionResult:
         transaction_result = TransactionResult()

{arize_phoenix-4.0.0 → arize_phoenix-4.0.2}/src/phoenix/db/insertion/helpers.py

@@ -30,11 +30,8 @@ def insert_stmt(
         raise ValueError(
             "Both `constraint` and `column_names` must be provided or omitted at the same time."
         )
-    if (
-        dialect is SupportedSQLDialect.POSTGRESQL
-        and constraint is None
-        or dialect is SupportedSQLDialect.SQLITE
-        and not column_names
+    if (dialect is SupportedSQLDialect.POSTGRESQL and constraint is None) or (
+        dialect is SupportedSQLDialect.SQLITE and not column_names
     ):
         return insert(table).values(values)
     if dialect is SupportedSQLDialect.POSTGRESQL:

{arize_phoenix-4.0.0 → arize_phoenix-4.0.2}/src/phoenix/trace/otel.py

@@ -17,7 +17,11 @@ from typing import (
 
 import numpy as np
 import opentelemetry.proto.trace.v1.trace_pb2 as otlp
-from openinference.semconv.trace import DocumentAttributes, SpanAttributes
+from openinference.semconv.trace import (
+    DocumentAttributes,
+    OpenInferenceMimeTypeValues,
+    SpanAttributes,
+)
 from opentelemetry.proto.common.v1.common_pb2 import AnyValue, ArrayValue, KeyValue
 from opentelemetry.util.types import Attributes, AttributeValue
 from typing_extensions import TypeAlias, assert_never
@@ -70,6 +74,12 @@ def decode_otlp_span(otlp_span: otlp.Span) -> Span:
     status_code, status_message = _decode_status(otlp_span.status)
     events = [_decode_event(event) for event in otlp_span.events]
 
+    if (input_value := get_attribute_value(attributes, INPUT_VALUE)) and not isinstance(
+        input_value, str
+    ):
+        attributes["input"]["value"] = json.dumps(input_value)
+        attributes["input"]["mime_type"] = OpenInferenceMimeTypeValues.JSON.value
+
     return Span(
         name=otlp_span.name,
         context=SpanContext(

{arize_phoenix-4.0.0 → arize_phoenix-4.0.2}/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-4.0.2/src/phoenix/version.py

@@ -0,0 +1 @@
+__version__ = "4.0.2"

arize_phoenix-4.0.0/src/phoenix/version.py

@@ -1 +0,0 @@
-__version__ = "4.0.0"