arize-phoenix 5.11.0__tar.gz → 5.12.0__tar.gz

{arize_phoenix-5.11.0 → arize_phoenix-5.12.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: arize-phoenix
-Version: 5.11.0
+Version: 5.12.0
 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-5.11.0 → arize_phoenix-5.12.0}/src/phoenix/server/api/helpers/playground_clients.py

@@ -131,8 +131,8 @@ class PlaygroundRateLimiter(RateLimiter, KeyedSingleton):
                 await self._throttler.async_wait_until_ready()
                 request_start_time = time.time()
                 maybe_coroutine = fn(*args, **kwargs)
-                if inspect.iscoroutine(maybe_coroutine):
-                    return await maybe_coroutine  # type: ignore
+                if inspect.isawaitable(maybe_coroutine):
+                    return await maybe_coroutine  # type: ignore[no-any-return]
                 else:
                     return maybe_coroutine
             except self._rate_limit_error:
@@ -144,10 +144,11 @@ class PlaygroundRateLimiter(RateLimiter, KeyedSingleton):
                             try:
                                 request_start_time = time.time()
                                 await self._throttler.async_wait_until_ready()
-                                if inspect.iscoroutinefunction(fn):
-                                    return await fn(*args, **kwargs)  # type: ignore
+                                maybe_coroutine = fn(*args, **kwargs)
+                                if inspect.isawaitable(maybe_coroutine):
+                                    return await maybe_coroutine  # type: ignore[no-any-return]
                                 else:
-                                    return fn(*args, **kwargs)
+                                    return maybe_coroutine
                             except self._rate_limit_error:
                                 self._throttler.on_rate_limit_error(
                                     request_start_time, verbose=self._verbose
@@ -295,12 +296,14 @@ class OpenAIStreamingClient(PlaygroundStreamingClient):
             BoundedFloatInvocationParameter(
                 invocation_name="frequency_penalty",
                 label="Frequency Penalty",
+                default_value=0.0,
                 min_value=-2.0,
                 max_value=2.0,
             ),
             BoundedFloatInvocationParameter(
                 invocation_name="presence_penalty",
                 label="Presence Penalty",
+                default_value=0.0,
                 min_value=-2.0,
                 max_value=2.0,
             ),
@@ -313,12 +316,14 @@ class OpenAIStreamingClient(PlaygroundStreamingClient):
                 invocation_name="top_p",
                 canonical_name=CanonicalParameterName.TOP_P,
                 label="Top P",
+                default_value=1.0,
                 min_value=0.0,
                 max_value=1.0,
             ),
             IntInvocationParameter(
                 invocation_name="seed",
                 canonical_name=CanonicalParameterName.RANDOM_SEED,
+                default_value=0,
                 label="Seed",
             ),
             JSONInvocationParameter(
@@ -345,7 +350,11 @@ class OpenAIStreamingClient(PlaygroundStreamingClient):
         from openai.types.chat import ChatCompletionStreamOptionsParam
 
         # Convert standard messages to OpenAI messages
-        openai_messages = [self.to_openai_chat_completion_param(*message) for message in 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)
         tool_call_ids: dict[int, str] = {}
         token_usage: Optional["CompletionUsage"] = None
         throttled_create = self.rate_limiter._alimit(self.client.chat.completions.create)
@@ -395,7 +404,7 @@ class OpenAIStreamingClient(PlaygroundStreamingClient):
         content: JSONScalarType,
         tool_call_id: Optional[str] = None,
         tool_calls: Optional[list[JSONScalarType]] = None,
-    ) -> "ChatCompletionMessageParam":
+    ) -> Optional["ChatCompletionMessageParam"]:
         from openai.types.chat import (
             ChatCompletionAssistantMessageParam,
             ChatCompletionSystemMessageParam,
@@ -487,6 +496,7 @@ class OpenAIO1StreamingClient(OpenAIStreamingClient):
                 invocation_name="seed",
                 canonical_name=CanonicalParameterName.RANDOM_SEED,
                 label="Seed",
+                default_value=0,
             ),
             JSONInvocationParameter(
                 invocation_name="tool_choice",
@@ -495,65 +505,7 @@ class OpenAIO1StreamingClient(OpenAIStreamingClient):
             ),
         ]
 
-    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
-        unfiltered_openai_messages = [
-            self.to_openai_o1_chat_completion_param(*message) for message in messages
-        ]
-
-        # filter out unsupported messages
-        openai_messages: list[ChatCompletionMessageParam] = [
-            message for message in unfiltered_openai_messages if message is not None
-        ]
-
-        tool_call_ids: dict[int, str] = {}
-
-        throttled_create = self.rate_limiter._alimit(self.client.chat.completions.create)
-        response = await throttled_create(
-            messages=openai_messages,
-            model=self.model_name,
-            tools=tools or NOT_GIVEN,
-            **invocation_parameters,
-        )
-
-        choice = response.choices[0]
-        message = choice.message
-        content = message.content
-
-        text_chunk = TextChunk(content=content)
-        yield text_chunk
-
-        if (tool_calls := message.tool_calls) is not None:
-            for tool_call_index, tool_call in enumerate(tool_calls):
-                tool_call_id = (
-                    tool_call.id
-                    if tool_call.id is not None
-                    else tool_call_ids.get(tool_call_index, f"tool_call_{tool_call_index}")
-                )
-                tool_call_ids[tool_call_index] = tool_call_id
-                if (function := tool_call.function) is not None:
-                    tool_call_chunk = ToolCallChunk(
-                        id=tool_call_id,
-                        function=FunctionCallChunk(
-                            name=function.name or "",
-                            arguments=function.arguments or "",
-                        ),
-                    )
-                    yield tool_call_chunk
-
-        if (usage := response.usage) is not None:
-            self._attributes.update(dict(self._llm_token_counts(usage)))
-
-    def to_openai_o1_chat_completion_param(
+    def to_openai_chat_completion_param(
         self,
         role: ChatCompletionMessageRole,
         content: JSONScalarType,
@@ -640,8 +592,12 @@ class AzureOpenAIStreamingClient(OpenAIStreamingClient):
     provider_key=GenerativeProviderKey.ANTHROPIC,
     model_names=[
         PROVIDER_DEFAULT,
+        "claude-3-5-sonnet-latest",
+        "claude-3-5-haiku-latest",
+        "claude-3-5-sonnet-20241022",
+        "claude-3-5-haiku-20241022",
         "claude-3-5-sonnet-20240620",
-        "claude-3-opus-20240229",
+        "claude-3-opus-latest",
         "claude-3-sonnet-20240229",
         "claude-3-haiku-20240307",
     ],
@@ -674,12 +630,14 @@ class AnthropicStreamingClient(PlaygroundStreamingClient):
                 invocation_name="max_tokens",
                 canonical_name=CanonicalParameterName.MAX_COMPLETION_TOKENS,
                 label="Max Tokens",
+                default_value=1024,
                 required=True,
             ),
             BoundedFloatInvocationParameter(
                 invocation_name="temperature",
                 canonical_name=CanonicalParameterName.TEMPERATURE,
                 label="Temperature",
+                default_value=0.0,
                 min_value=0.0,
                 max_value=1.0,
             ),
@@ -692,6 +650,7 @@ class AnthropicStreamingClient(PlaygroundStreamingClient):
                 invocation_name="top_p",
                 canonical_name=CanonicalParameterName.TOP_P,
                 label="Top P",
+                default_value=1.0,
                 min_value=0.0,
                 max_value=1.0,
             ),
@@ -867,21 +826,25 @@ class GeminiStreamingClient(PlaygroundStreamingClient):
             FloatInvocationParameter(
                 invocation_name="presence_penalty",
                 label="Presence Penalty",
+                default_value=0.0,
             ),
             FloatInvocationParameter(
                 invocation_name="frequency_penalty",
                 label="Frequency Penalty",
+                default_value=0.0,
             ),
             BoundedFloatInvocationParameter(
                 invocation_name="top_p",
                 canonical_name=CanonicalParameterName.TOP_P,
                 label="Top P",
+                default_value=1.0,
                 min_value=0.0,
                 max_value=1.0,
             ),
             BoundedFloatInvocationParameter(
                 invocation_name="top_k",
                 label="Top K",
+                default_value=1.0,
                 min_value=0.0,
                 max_value=1.0,
             ),
@@ -889,6 +852,7 @@ class GeminiStreamingClient(PlaygroundStreamingClient):
                 invocation_name="seed",
                 canonical_name=CanonicalParameterName.RANDOM_SEED,
                 label="Seed",
+                default_value=0,
             ),
         ]
 
@@ -923,6 +887,13 @@ class GeminiStreamingClient(PlaygroundStreamingClient):
         chat = client.start_chat(history=gemini_message_history)
         stream = await chat.send_message_async(**gemini_params)
         async for event in stream:
+            self._attributes.update(
+                {
+                    LLM_TOKEN_COUNT_PROMPT: event.usage_metadata.prompt_token_count,
+                    LLM_TOKEN_COUNT_COMPLETION: event.usage_metadata.candidates_token_count,
+                    LLM_TOKEN_COUNT_TOTAL: event.usage_metadata.total_token_count,
+                }
+            )
             yield TextChunk(content=event.text)
 
     def _build_gemini_messages(

{arize_phoenix-5.11.0 → arize_phoenix-5.12.0}/src/phoenix/server/api/mutations/chat_mutations.py

@@ -25,6 +25,7 @@ from typing_extensions import assert_never
 from phoenix.datetime_utils import local_now, normalize_datetime
 from phoenix.db import models
 from phoenix.db.helpers import get_dataset_example_revisions
+from phoenix.server.api.auth import IsNotReadOnly
 from phoenix.server.api.context import Context
 from phoenix.server.api.exceptions import BadRequest, CustomGraphQLError, NotFound
 from phoenix.server.api.helpers.playground_clients import (
@@ -67,6 +68,7 @@ from phoenix.utilities.json import jsonify
 from phoenix.utilities.template_formatters import (
     FStringTemplateFormatter,
     MustacheTemplateFormatter,
+    NoOpFormatter,
     TemplateFormatter,
 )
 
@@ -117,7 +119,7 @@ class ChatCompletionOverDatasetMutationPayload:
 
 @strawberry.type
 class ChatCompletionMutationMixin:
-    @strawberry.mutation
+    @strawberry.mutation(permission_classes=[IsNotReadOnly])  # type: ignore
     @classmethod
     async def chat_completion_over_dataset(
         cls,
@@ -272,7 +274,7 @@ class ChatCompletionMutationMixin:
             payload.examples.append(example_payload)
         return payload
 
-    @strawberry.mutation
+    @strawberry.mutation(permission_classes=[IsNotReadOnly])  # type: ignore
     @classmethod
     async def chat_completion(
         cls, info: Info[Context, None], input: ChatCompletionInput
@@ -481,6 +483,8 @@ def _template_formatter(template_language: TemplateLanguage) -> TemplateFormatte
         return MustacheTemplateFormatter()
     if template_language is TemplateLanguage.F_STRING:
         return FStringTemplateFormatter()
+    if template_language is TemplateLanguage.NONE:
+        return NoOpFormatter()
     assert_never(template_language)
 
 

{arize_phoenix-5.11.0 → arize_phoenix-5.12.0}/src/phoenix/server/api/subscriptions.py

@@ -1,11 +1,11 @@
 import asyncio
 import logging
-from asyncio import FIRST_COMPLETED, Queue, QueueEmpty, Task, create_task, wait, wait_for
 from collections.abc import AsyncIterator, Iterator
 from datetime import datetime, timedelta, timezone
 from typing import (
     Any,
     AsyncGenerator,
+    Coroutine,
     Iterable,
     Mapping,
     Optional,
@@ -24,6 +24,7 @@ from typing_extensions import TypeAlias, assert_never
 
 from phoenix.datetime_utils import local_now, normalize_datetime
 from phoenix.db import models
+from phoenix.server.api.auth import IsNotReadOnly
 from phoenix.server.api.context import Context
 from phoenix.server.api.exceptions import BadRequest, CustomGraphQLError, NotFound
 from phoenix.server.api.helpers.playground_clients import (
@@ -63,6 +64,7 @@ from phoenix.server.types import DbSessionFactory
 from phoenix.utilities.template_formatters import (
     FStringTemplateFormatter,
     MustacheTemplateFormatter,
+    NoOpFormatter,
     TemplateFormatter,
     TemplateFormatterError,
 )
@@ -86,7 +88,7 @@ PLAYGROUND_PROJECT_NAME = "playground"
 
 @strawberry.type
 class Subscription:
-    @strawberry.subscription
+    @strawberry.subscription(permission_classes=[IsNotReadOnly])  # type: ignore
     async def chat_completion(
         self, info: Info[Context, None], input: ChatCompletionInput
     ) -> AsyncIterator[ChatCompletionSubscriptionPayload]:
@@ -161,7 +163,7 @@ class Subscription:
         info.context.event_queue.put(SpanInsertEvent(ids=(playground_project_id,)))
         yield ChatCompletionSubscriptionResult(span=to_gql_span(db_span))
 
-    @strawberry.subscription
+    @strawberry.subscription(permission_classes=[IsNotReadOnly])  # type: ignore
     async def chat_completion_over_dataset(
         self, info: Info[Context, None], input: ChatCompletionOverDatasetInput
     ) -> AsyncIterator[ChatCompletionSubscriptionPayload]:
@@ -286,7 +288,7 @@ class Subscription:
             experiment=to_gql_experiment(experiment)
         )  # eagerly yields experiment so it can be linked by consumers of the subscription
 
-        results: Queue[ChatCompletionResult] = Queue()
+        results: asyncio.Queue[ChatCompletionResult] = asyncio.Queue()
         not_started: list[tuple[DatasetExampleID, ChatStream]] = [
             (
                 GlobalID(DatasetExample.__name__, str(revision.dataset_example_id)),
@@ -302,7 +304,11 @@ class Subscription:
             for revision in revisions
         ]
         in_progress: list[
-            tuple[Optional[DatasetExampleID], ChatStream, Task[ChatCompletionSubscriptionPayload]]
+            tuple[
+                Optional[DatasetExampleID],
+                ChatStream,
+                asyncio.Task[ChatCompletionSubscriptionPayload],
+            ]
         ] = []
         max_in_progress = 3
         write_batch_size = 10
@@ -314,7 +320,9 @@ class Subscription:
                 task = _create_task_with_timeout(stream)
                 in_progress.append((ex_id, stream, task))
             async_tasks_to_run = [task for _, _, task in in_progress]
-            completed_tasks, _ = await wait(async_tasks_to_run, return_when=FIRST_COMPLETED)
+            completed_tasks, _ = await asyncio.wait(
+                async_tasks_to_run, return_when=asyncio.FIRST_COMPLETED
+            )
             for completed_task in completed_tasks:
                 idx = [task for _, _, task in in_progress].index(completed_task)
                 example_id, stream, _ = in_progress[idx]
@@ -326,7 +334,7 @@ class Subscription:
                     del in_progress[idx]  # removes timed-out stream
                     if example_id is not None:
                         yield ChatCompletionSubscriptionError(
-                            message="Timed out", dataset_example_id=example_id
+                            message="Playground task timed out", dataset_example_id=example_id
                         )
                 except Exception as error:
                     del in_progress[idx]  # removes failed stream
@@ -367,7 +375,7 @@ async def _stream_chat_completion_over_dataset_example(
     input: ChatCompletionOverDatasetInput,
     llm_client: PlaygroundStreamingClient,
     revision: models.DatasetExampleRevision,
-    results: Queue[ChatCompletionResult],
+    results: asyncio.Queue[ChatCompletionResult],
     experiment_id: int,
     project_id: int,
 ) -> ChatStream:
@@ -469,23 +477,52 @@ def _is_result_payloads_stream(
 
 def _create_task_with_timeout(
     iterable: AsyncIterator[GenericType], timeout_in_seconds: int = 90
-) -> Task[GenericType]:
-    return create_task(wait_for(_as_coroutine(iterable), timeout=timeout_in_seconds))
+) -> asyncio.Task[GenericType]:
+    return asyncio.create_task(
+        _wait_for(
+            _as_coroutine(iterable),
+            timeout=timeout_in_seconds,
+            timeout_message="Playground task timed out",
+        )
+    )
+
+
+async def _wait_for(
+    coro: Coroutine[None, None, GenericType],
+    timeout: float,
+    timeout_message: Optional[str] = None,
+) -> GenericType:
+    """
+    A function that imitates asyncio.wait_for, but allows the task to be
+    cancelled with a custom message.
+    """
+    task = asyncio.create_task(coro)
+    done, pending = await asyncio.wait([task], timeout=timeout)
+    assert len(done) + len(pending) == 1
+    if done:
+        task = done.pop()
+        return task.result()
+    task = pending.pop()
+    task.cancel(msg=timeout_message)
+    try:
+        return await task
+    except asyncio.CancelledError:
+        raise asyncio.TimeoutError()
 
 
-async def _drain(queue: Queue[GenericType]) -> list[GenericType]:
+async def _drain(queue: asyncio.Queue[GenericType]) -> list[GenericType]:
     values: list[GenericType] = []
     while not queue.empty():
         values.append(await queue.get())
     return values
 
 
-def _drain_no_wait(queue: Queue[GenericType]) -> list[GenericType]:
+def _drain_no_wait(queue: asyncio.Queue[GenericType]) -> list[GenericType]:
     values: list[GenericType] = []
     while True:
         try:
             values.append(queue.get_nowait())
-        except QueueEmpty:
+        except asyncio.QueueEmpty:
             break
     return values
 
@@ -526,6 +563,8 @@ def _template_formatter(template_language: TemplateLanguage) -> TemplateFormatte
         return MustacheTemplateFormatter()
     if template_language is TemplateLanguage.F_STRING:
         return FStringTemplateFormatter()
+    if template_language is TemplateLanguage.NONE:
+        return NoOpFormatter()
     assert_never(template_language)
 
 

{arize_phoenix-5.11.0 → arize_phoenix-5.12.0}/src/phoenix/server/api/types/TemplateLanguage.py

@@ -5,5 +5,6 @@ import strawberry
 
 @strawberry.enum
 class TemplateLanguage(Enum):
+    NONE = "NONE"
     MUSTACHE = "MUSTACHE"
     F_STRING = "F_STRING"

arize_phoenix-5.12.0/src/phoenix/server/static/.vite/manifest.json

@@ -0,0 +1,78 @@
+{
+  "_components-72cQL1d1.js": {
+    "file": "assets/components-72cQL1d1.js",
+    "name": "components",
+    "imports": [
+      "_vendor-DexmGnha.js",
+      "_pages-DFAkBAUh.js",
+      "_vendor-arizeai--Q3ol330.js",
+      "_vendor-codemirror-B4bYvWa6.js",
+      "_vendor-three-DwGkEfCM.js"
+    ]
+  },
+  "_pages-DFAkBAUh.js": {
+    "file": "assets/pages-DFAkBAUh.js",
+    "name": "pages",
+    "imports": [
+      "_vendor-DexmGnha.js",
+      "_vendor-arizeai--Q3ol330.js",
+      "_components-72cQL1d1.js",
+      "_vendor-recharts-B4ZzJhNh.js",
+      "_vendor-codemirror-B4bYvWa6.js"
+    ]
+  },
+  "_vendor-!~{003}~.js": {
+    "file": "assets/vendor-DxkFTwjz.css",
+    "src": "_vendor-!~{003}~.js"
+  },
+  "_vendor-DexmGnha.js": {
+    "file": "assets/vendor-DexmGnha.js",
+    "name": "vendor",
+    "imports": [
+      "_vendor-three-DwGkEfCM.js"
+    ],
+    "css": [
+      "assets/vendor-DxkFTwjz.css"
+    ]
+  },
+  "_vendor-arizeai--Q3ol330.js": {
+    "file": "assets/vendor-arizeai--Q3ol330.js",
+    "name": "vendor-arizeai",
+    "imports": [
+      "_vendor-DexmGnha.js"
+    ]
+  },
+  "_vendor-codemirror-B4bYvWa6.js": {
+    "file": "assets/vendor-codemirror-B4bYvWa6.js",
+    "name": "vendor-codemirror",
+    "imports": [
+      "_vendor-DexmGnha.js"
+    ]
+  },
+  "_vendor-recharts-B4ZzJhNh.js": {
+    "file": "assets/vendor-recharts-B4ZzJhNh.js",
+    "name": "vendor-recharts",
+    "imports": [
+      "_vendor-DexmGnha.js"
+    ]
+  },
+  "_vendor-three-DwGkEfCM.js": {
+    "file": "assets/vendor-three-DwGkEfCM.js",
+    "name": "vendor-three"
+  },
+  "index.tsx": {
+    "file": "assets/index-BowjltW-.js",
+    "name": "index",
+    "src": "index.tsx",
+    "isEntry": true,
+    "imports": [
+      "_vendor-DexmGnha.js",
+      "_vendor-arizeai--Q3ol330.js",
+      "_pages-DFAkBAUh.js",
+      "_components-72cQL1d1.js",
+      "_vendor-three-DwGkEfCM.js",
+      "_vendor-recharts-B4ZzJhNh.js",
+      "_vendor-codemirror-B4bYvWa6.js"
+    ]
+  }
+}