arize-phoenix 11.9.0__tar.gz → 11.10.1__tar.gz

{arize_phoenix-11.9.0 → arize_phoenix-11.10.1}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: arize-phoenix
-Version: 11.9.0
+Version: 11.10.1
 Summary: AI Observability and Evaluation
 Project-URL: Documentation, https://arize.com/docs/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues

{arize_phoenix-11.9.0 → arize_phoenix-11.10.1}/src/phoenix/server/api/helpers/playground_clients.py

@@ -20,7 +20,7 @@ from openinference.semconv.trace import (
 )
 from strawberry import UNSET
 from strawberry.scalars import JSON as JSONScalarType
-from typing_extensions import TypeAlias, assert_never
+from typing_extensions import TypeAlias, assert_never, override
 
 from phoenix.config import getenv
 from phoenix.evals.models.rate_limiters import (
@@ -437,9 +437,9 @@ class OpenAIBaseStreamingClient(PlaygroundStreamingClient):
         if role is ChatCompletionMessageRole.TOOL:
             if tool_call_id is None:
                 raise ValueError("tool_call_id is required for tool messages")
-        return ChatCompletionToolMessageParam(
-            {"content": content, "role": "tool", "tool_call_id": tool_call_id}
-        )
+            return ChatCompletionToolMessageParam(
+                {"content": content, "role": "tool", "tool_call_id": tool_call_id}
+            )
         assert_never(role)
 
     def to_openai_tool_call_param(
@@ -1140,27 +1140,28 @@ class OpenAIStreamingClient(OpenAIBaseStreamingClient):
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.OPENAI.value
 
 
-@register_llm_client(
-    provider_key=GenerativeProviderKey.OPENAI,
-    model_names=[
-        "o1",
-        "o1-pro",
-        "o1-2024-12-17",
-        "o1-pro-2025-03-19",
-        "o1-mini",
-        "o1-mini-2024-09-12",
-        "o1-preview",
-        "o1-preview-2024-09-12",
-        "o3",
-        "o3-pro",
-        "o3-2025-04-16",
-        "o3-mini",
-        "o3-mini-2025-01-31",
-        "o4-mini",
-        "o4-mini-2025-04-16",
-    ],
-)
-class OpenAIReasoningStreamingClient(OpenAIStreamingClient):
+_OPENAI_REASONING_MODELS = [
+    "o1",
+    "o1-pro",
+    "o1-2024-12-17",
+    "o1-pro-2025-03-19",
+    "o1-mini",
+    "o1-mini-2024-09-12",
+    "o1-preview",
+    "o1-preview-2024-09-12",
+    "o3",
+    "o3-pro",
+    "o3-2025-04-16",
+    "o3-mini",
+    "o3-mini-2025-01-31",
+    "o4-mini",
+    "o4-mini-2025-04-16",
+]
+
+
+class OpenAIReasoningReasoningModelsMixin:
+    """Mixin class for OpenAI-style reasoning model clients (o1, o3 series)."""
+
     @classmethod
     def supported_invocation_parameters(cls) -> list[InvocationParameter]:
         return [
@@ -1191,6 +1192,16 @@ class OpenAIReasoningStreamingClient(OpenAIStreamingClient):
             ),
         ]
 
+
+@register_llm_client(
+    provider_key=GenerativeProviderKey.OPENAI,
+    model_names=_OPENAI_REASONING_MODELS,
+)
+class OpenAIReasoningNonStreamingClient(
+    OpenAIReasoningReasoningModelsMixin,
+    OpenAIStreamingClient,
+):
+    @override
     async def chat_completion_create(
         self,
         messages: list[
@@ -1283,46 +1294,11 @@ class OpenAIReasoningStreamingClient(OpenAIStreamingClient):
         if role is ChatCompletionMessageRole.TOOL:
             if tool_call_id is None:
                 raise ValueError("tool_call_id is required for tool messages")
-        return ChatCompletionToolMessageParam(
-            {"content": content, "role": "tool", "tool_call_id": tool_call_id}
-        )
+            return ChatCompletionToolMessageParam(
+                {"content": content, "role": "tool", "tool_call_id": tool_call_id}
+            )
         assert_never(role)
 
-    @staticmethod
-    def _llm_token_counts(usage: "CompletionUsage") -> Iterator[tuple[str, Any]]:
-        yield LLM_TOKEN_COUNT_PROMPT, usage.prompt_tokens
-        yield LLM_TOKEN_COUNT_COMPLETION, usage.completion_tokens
-        yield LLM_TOKEN_COUNT_TOTAL, usage.total_tokens
-
-        if hasattr(usage, "prompt_tokens_details") and usage.prompt_tokens_details is not None:
-            prompt_details = usage.prompt_tokens_details
-            if (
-                hasattr(prompt_details, "cached_tokens")
-                and prompt_details.cached_tokens is not None
-            ):
-                yield LLM_TOKEN_COUNT_PROMPT_DETAILS_CACHE_READ, prompt_details.cached_tokens
-            if hasattr(prompt_details, "audio_tokens") and prompt_details.audio_tokens is not None:
-                yield LLM_TOKEN_COUNT_PROMPT_DETAILS_AUDIO, prompt_details.audio_tokens
-
-        if (
-            hasattr(usage, "completion_tokens_details")
-            and usage.completion_tokens_details is not None
-        ):
-            completion_details = usage.completion_tokens_details
-            if (
-                hasattr(completion_details, "reasoning_tokens")
-                and completion_details.reasoning_tokens is not None
-            ):
-                yield (
-                    LLM_TOKEN_COUNT_COMPLETION_DETAILS_REASONING,
-                    completion_details.reasoning_tokens,
-                )
-            if (
-                hasattr(completion_details, "audio_tokens")
-                and completion_details.audio_tokens is not None
-            ):
-                yield LLM_TOKEN_COUNT_COMPLETION_DETAILS_AUDIO, completion_details.audio_tokens
-
 
 @register_llm_client(
     provider_key=GenerativeProviderKey.AZURE_OPENAI,
@@ -1376,6 +1352,113 @@ class AzureOpenAIStreamingClient(OpenAIBaseStreamingClient):
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.OPENAI.value
 
 
+@register_llm_client(
+    provider_key=GenerativeProviderKey.AZURE_OPENAI,
+    model_names=_OPENAI_REASONING_MODELS,
+)
+class AzureOpenAIReasoningNonStreamingClient(
+    OpenAIReasoningReasoningModelsMixin,
+    AzureOpenAIStreamingClient,
+):
+    @override
+    async def chat_completion_create(
+        self,
+        messages: list[
+            tuple[ChatCompletionMessageRole, str, Optional[str], Optional[list[JSONScalarType]]]
+        ],
+        tools: list[JSONScalarType],
+        **invocation_parameters: Any,
+    ) -> AsyncIterator[ChatCompletionChunk]:
+        from openai import NOT_GIVEN
+
+        # Convert standard messages to OpenAI messages
+        openai_messages = []
+        for message in messages:
+            openai_message = self.to_openai_chat_completion_param(*message)
+            if openai_message is not None:
+                openai_messages.append(openai_message)
+
+        throttled_create = self.rate_limiter._alimit(self.client.chat.completions.create)
+        response = await throttled_create(
+            messages=openai_messages,
+            model=self.model_name,
+            stream=False,
+            tools=tools or NOT_GIVEN,
+            **invocation_parameters,
+        )
+
+        if response.usage is not None:
+            self._attributes.update(dict(self._llm_token_counts(response.usage)))
+
+        choice = response.choices[0]
+        if choice.message.content:
+            yield TextChunk(content=choice.message.content)
+
+        if choice.message.tool_calls:
+            for tool_call in choice.message.tool_calls:
+                yield ToolCallChunk(
+                    id=tool_call.id,
+                    function=FunctionCallChunk(
+                        name=tool_call.function.name,
+                        arguments=tool_call.function.arguments,
+                    ),
+                )
+
+    def to_openai_chat_completion_param(
+        self,
+        role: ChatCompletionMessageRole,
+        content: JSONScalarType,
+        tool_call_id: Optional[str] = None,
+        tool_calls: Optional[list[JSONScalarType]] = None,
+    ) -> Optional["ChatCompletionMessageParam"]:
+        from openai.types.chat import (
+            ChatCompletionAssistantMessageParam,
+            ChatCompletionDeveloperMessageParam,
+            ChatCompletionToolMessageParam,
+            ChatCompletionUserMessageParam,
+        )
+
+        if role is ChatCompletionMessageRole.USER:
+            return ChatCompletionUserMessageParam(
+                {
+                    "content": content,
+                    "role": "user",
+                }
+            )
+        if role is ChatCompletionMessageRole.SYSTEM:
+            return ChatCompletionDeveloperMessageParam(
+                {
+                    "content": content,
+                    "role": "developer",
+                }
+            )
+        if role is ChatCompletionMessageRole.AI:
+            if tool_calls is None:
+                return ChatCompletionAssistantMessageParam(
+                    {
+                        "content": content,
+                        "role": "assistant",
+                    }
+                )
+            else:
+                return ChatCompletionAssistantMessageParam(
+                    {
+                        "content": content,
+                        "role": "assistant",
+                        "tool_calls": [
+                            self.to_openai_tool_call_param(tool_call) for tool_call in tool_calls
+                        ],
+                    }
+                )
+        if role is ChatCompletionMessageRole.TOOL:
+            if tool_call_id is None:
+                raise ValueError("tool_call_id is required for tool messages")
+            return ChatCompletionToolMessageParam(
+                {"content": content, "role": "tool", "tool_call_id": tool_call_id}
+            )
+        assert_never(role)
+
+
 @register_llm_client(
     provider_key=GenerativeProviderKey.ANTHROPIC,
     model_names=[

{arize_phoenix-11.9.0 → arize_phoenix-11.10.1}/src/phoenix/server/api/helpers/playground_registry.py

@@ -1,4 +1,4 @@
-from typing import TYPE_CHECKING, Any, Callable, Optional, Union
+from typing import TYPE_CHECKING, Any, Callable, Optional, Sequence, Union
 
 from phoenix.server.api.types.GenerativeProvider import GenerativeProviderKey
 
@@ -59,7 +59,7 @@ PLAYGROUND_CLIENT_REGISTRY: PlaygroundClientRegistry = PlaygroundClientRegistry(
 
 def register_llm_client(
     provider_key: GenerativeProviderKey,
-    model_names: list[ModelName],
+    model_names: Sequence[ModelName],
 ) -> Callable[[type["PlaygroundStreamingClient"]], type["PlaygroundStreamingClient"]]:
     def decorator(cls: type["PlaygroundStreamingClient"]) -> type["PlaygroundStreamingClient"]:
         provider_registry = PLAYGROUND_CLIENT_REGISTRY._registry.setdefault(provider_key, {})

{arize_phoenix-11.9.0 → arize_phoenix-11.10.1}/src/phoenix/server/api/types/Project.py

@@ -7,7 +7,7 @@ from typing import TYPE_CHECKING, Annotated, Any, ClassVar, Literal, Optional, c
 import strawberry
 from aioitertools.itertools import groupby, islice
 from openinference.semconv.trace import SpanAttributes
-from sqlalchemy import and_, desc, distinct, exists, func, or_, select
+from sqlalchemy import and_, case, desc, distinct, exists, func, or_, select
 from sqlalchemy.dialects import postgresql, sqlite
 from sqlalchemy.sql.elements import ColumnElement
 from sqlalchemy.sql.expression import tuple_
@@ -719,6 +719,7 @@ class Project(Node):
         info: Info[Context, None],
         time_range: TimeRange,
         time_bin_config: Optional[TimeBinConfig] = UNSET,
+        filter_condition: Optional[str] = UNSET,
     ) -> SpanCountTimeSeries:
         if time_range.start is None:
             raise BadRequest("Start time is required")
@@ -742,7 +743,17 @@ class Project(Node):
                 field = "year"
         bucket = date_trunc(dialect, field, models.Span.start_time, utc_offset_minutes)
         stmt = (
-            select(bucket, func.count(models.Span.id))
+            select(
+                bucket,
+                func.count(models.Span.id).label("total_count"),
+                func.sum(case((models.Span.status_code == "OK", 1), else_=0)).label("ok_count"),
+                func.sum(case((models.Span.status_code == "ERROR", 1), else_=0)).label(
+                    "error_count"
+                ),
+                func.sum(case((models.Span.status_code == "UNSET", 1), else_=0)).label(
+                    "unset_count"
+                ),
+            )
             .join_from(models.Span, models.Trace)
             .where(models.Trace.project_rowid == self.project_rowid)
             .group_by(bucket)
@@ -752,21 +763,31 @@ class Project(Node):
             stmt = stmt.where(time_range.start <= models.Span.start_time)
         if time_range.end:
             stmt = stmt.where(models.Span.start_time < time_range.end)
+        if filter_condition:
+            span_filter = SpanFilter(condition=filter_condition)
+            stmt = span_filter(stmt)
 
         data = {}
         async with info.context.db() as session:
-            async for t, v in await session.stream(stmt):
+            async for t, total_count, ok_count, error_count, unset_count in await session.stream(
+                stmt
+            ):
                 timestamp = _as_datetime(t)
-                data[timestamp] = TimeSeriesDataPoint(timestamp=timestamp, value=v)
+                data[timestamp] = SpanCountTimeSeriesDataPoint(
+                    timestamp=timestamp,
+                    ok_count=ok_count,
+                    error_count=error_count,
+                    unset_count=unset_count,
+                    total_count=total_count,
+                )
 
         data_timestamps: list[datetime] = [data_point.timestamp for data_point in data.values()]
         min_time = min([*data_timestamps, time_range.start])
         max_time = max(
             [
                 *data_timestamps,
-                *([time_range.end] if time_range.end else []),
+                *([time_range.end] if time_range.end else [datetime.now(timezone.utc)]),
             ],
-            default=datetime.now(timezone.utc),
         )
         for timestamp in get_timestamp_range(
             start_time=min_time,
@@ -775,7 +796,7 @@ class Project(Node):
             utc_offset_minutes=utc_offset_minutes,
         ):
             if timestamp not in data:
-                data[timestamp] = TimeSeriesDataPoint(timestamp=timestamp)
+                data[timestamp] = SpanCountTimeSeriesDataPoint(timestamp=timestamp)
         return SpanCountTimeSeries(data=sorted(data.values(), key=lambda x: x.timestamp))
 
     @strawberry.field
@@ -828,9 +849,8 @@ class Project(Node):
         max_time = max(
             [
                 *data_timestamps,
-                *([time_range.end] if time_range.end else []),
+                *([time_range.end] if time_range.end else [datetime.now(timezone.utc)]),
             ],
-            default=datetime.now(timezone.utc),
         )
         for timestamp in get_timestamp_range(
             start_time=min_time,
@@ -917,9 +937,8 @@ class Project(Node):
         max_time = max(
             [
                 *data_timestamps,
-                *([time_range.end] if time_range.end else []),
+                *([time_range.end] if time_range.end else [datetime.now(timezone.utc)]),
             ],
-            default=datetime.now(timezone.utc),
         )
         for timestamp in get_timestamp_range(
             start_time=min_time,
@@ -1025,9 +1044,8 @@ class Project(Node):
         max_time = max(
             [
                 *data_timestamps,
-                *([time_range.end] if time_range.end else []),
+                *([time_range.end] if time_range.end else [datetime.now(timezone.utc)]),
             ],
-            default=datetime.now(timezone.utc),
         )
         for timestamp in get_timestamp_range(
             start_time=min_time,
@@ -1111,9 +1129,8 @@ class Project(Node):
         max_time = max(
             [
                 *data_timestamps,
-                *([time_range.end] if time_range.end else []),
+                *([time_range.end] if time_range.end else [datetime.now(timezone.utc)]),
             ],
-            default=datetime.now(timezone.utc),
         )
         for timestamp in get_timestamp_range(
             start_time=min_time,
@@ -1195,9 +1212,8 @@ class Project(Node):
         max_time = max(
             [
                 *data_timestamps,
-                *([time_range.end] if time_range.end else []),
+                *([time_range.end] if time_range.end else [datetime.now(timezone.utc)]),
             ],
-            default=datetime.now(timezone.utc),
         )
         for timestamp in get_timestamp_range(
             start_time=min_time,
@@ -1281,9 +1297,8 @@ class Project(Node):
         max_time = max(
             [
                 *score_timestamps,
-                *([time_range.end] if time_range.end else []),
+                *([time_range.end] if time_range.end else [datetime.now(timezone.utc)]),
             ],
-            default=datetime.now(timezone.utc),
         )
         data: dict[datetime, SpanAnnotationScoreTimeSeriesDataPoint] = {
             timestamp: SpanAnnotationScoreTimeSeriesDataPoint(
@@ -1313,8 +1328,17 @@ class Project(Node):
 
 
 @strawberry.type
-class SpanCountTimeSeries(TimeSeries):
-    """A time series of span count"""
+class SpanCountTimeSeriesDataPoint:
+    timestamp: datetime
+    ok_count: Optional[int] = None
+    error_count: Optional[int] = None
+    unset_count: Optional[int] = None
+    total_count: Optional[int] = None
+
+
+@strawberry.type
+class SpanCountTimeSeries:
+    data: list[SpanCountTimeSeriesDataPoint]
 
 
 @strawberry.type

arize_phoenix-11.10.1/src/phoenix/server/static/.vite/manifest.json

@@ -0,0 +1,87 @@
+{
+  "_components-XAeml0-1.js": {
+    "file": "assets/components-XAeml0-1.js",
+    "name": "components",
+    "imports": [
+      "_vendor-DhvamIr8.js",
+      "_pages-CPfaxiKa.js",
+      "_vendor-arizeai-4fVwwnrI.js",
+      "_vendor-codemirror-DRfFHb57.js",
+      "_vendor-three-C5WAXd5r.js"
+    ]
+  },
+  "_pages-CPfaxiKa.js": {
+    "file": "assets/pages-CPfaxiKa.js",
+    "name": "pages",
+    "imports": [
+      "_vendor-DhvamIr8.js",
+      "_vendor-arizeai-4fVwwnrI.js",
+      "_components-XAeml0-1.js",
+      "_vendor-codemirror-DRfFHb57.js",
+      "_vendor-recharts-w6bSawXG.js"
+    ]
+  },
+  "_vendor-CqDb5u4o.css": {
+    "file": "assets/vendor-CqDb5u4o.css",
+    "src": "_vendor-CqDb5u4o.css"
+  },
+  "_vendor-DhvamIr8.js": {
+    "file": "assets/vendor-DhvamIr8.js",
+    "name": "vendor",
+    "imports": [
+      "_vendor-three-C5WAXd5r.js"
+    ],
+    "css": [
+      "assets/vendor-CqDb5u4o.css"
+    ]
+  },
+  "_vendor-arizeai-4fVwwnrI.js": {
+    "file": "assets/vendor-arizeai-4fVwwnrI.js",
+    "name": "vendor-arizeai",
+    "imports": [
+      "_vendor-DhvamIr8.js"
+    ]
+  },
+  "_vendor-codemirror-DRfFHb57.js": {
+    "file": "assets/vendor-codemirror-DRfFHb57.js",
+    "name": "vendor-codemirror",
+    "imports": [
+      "_vendor-DhvamIr8.js",
+      "_vendor-shiki-CplrhwOk.js"
+    ]
+  },
+  "_vendor-recharts-w6bSawXG.js": {
+    "file": "assets/vendor-recharts-w6bSawXG.js",
+    "name": "vendor-recharts",
+    "imports": [
+      "_vendor-DhvamIr8.js"
+    ]
+  },
+  "_vendor-shiki-CplrhwOk.js": {
+    "file": "assets/vendor-shiki-CplrhwOk.js",
+    "name": "vendor-shiki",
+    "imports": [
+      "_vendor-DhvamIr8.js"
+    ]
+  },
+  "_vendor-three-C5WAXd5r.js": {
+    "file": "assets/vendor-three-C5WAXd5r.js",
+    "name": "vendor-three"
+  },
+  "index.tsx": {
+    "file": "assets/index-D7EtHUpz.js",
+    "name": "index",
+    "src": "index.tsx",
+    "isEntry": true,
+    "imports": [
+      "_vendor-DhvamIr8.js",
+      "_vendor-arizeai-4fVwwnrI.js",
+      "_pages-CPfaxiKa.js",
+      "_components-XAeml0-1.js",
+      "_vendor-three-C5WAXd5r.js",
+      "_vendor-codemirror-DRfFHb57.js",
+      "_vendor-shiki-CplrhwOk.js",
+      "_vendor-recharts-w6bSawXG.js"
+    ]
+  }
+}