arize-phoenix 11.36.0__tar.gz → 11.38.0__tar.gz

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: arize-phoenix
-Version: 11.36.0
+Version: 11.38.0
 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-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/helpers/playground_clients.py

@@ -57,6 +57,7 @@ from phoenix.server.api.types.GenerativeProvider import GenerativeProviderKey
 if TYPE_CHECKING:
     import httpx
     from anthropic.types import MessageParam, TextBlockParam, ToolResultBlockParam
+    from botocore.awsrequest import AWSPreparedRequest  # type: ignore[import-untyped]
     from google.generativeai.types import ContentType
     from openai import AsyncAzureOpenAI, AsyncOpenAI
     from openai.types import CompletionUsage
@@ -546,6 +547,7 @@ class DeepSeekStreamingClient(OpenAIBaseStreamingClient):
         client = AsyncOpenAI(
             api_key=api_key,
             base_url=base_url or "https://api.deepseek.com",
+            default_headers=model.custom_headers or None,
         )
         super().__init__(client=client, model=model, credentials=credentials)
         # DeepSeek uses OpenAI-compatible API but we'll track it as a separate provider
@@ -587,6 +589,7 @@ class XAIStreamingClient(OpenAIBaseStreamingClient):
         client = AsyncOpenAI(
             api_key=api_key,
             base_url=base_url or "https://api.x.ai/v1",
+            default_headers=model.custom_headers or None,
         )
         super().__init__(client=client, model=model, credentials=credentials)
         # xAI uses OpenAI-compatible API but we'll track it as a separate provider
@@ -624,7 +627,11 @@ class OllamaStreamingClient(OpenAIBaseStreamingClient):
         if not base_url:
             raise BadRequest("An Ollama base URL is required for Ollama models")
         api_key = "ollama"
-        client = AsyncOpenAI(api_key=api_key, base_url=base_url)
+        client = AsyncOpenAI(
+            api_key=api_key,
+            base_url=base_url,
+            default_headers=model.custom_headers or None,
+        )
         super().__init__(client=client, model=model, credentials=credentials)
         # Ollama uses OpenAI-compatible API but we'll track it as a separate provider
         # Adding a custom "ollama" provider value to make it distinguishable in traces
@@ -679,6 +686,7 @@ class BedrockStreamingClient(PlaygroundStreamingClient):
         super().__init__(model=model, credentials=credentials)
         self.region = model.region or "us-east-1"
         self.api = "converse"
+        self.custom_headers = model.custom_headers or {}
         self.aws_access_key_id = _get_credential_value(credentials, "AWS_ACCESS_KEY_ID") or getenv(
             "AWS_ACCESS_KEY_ID"
         )
@@ -697,9 +705,24 @@ class BedrockStreamingClient(PlaygroundStreamingClient):
             aws_session_token=self.aws_session_token,
         )
 
+        # Add custom headers support via boto3 event system
+        if self.custom_headers:
+            self._setup_custom_headers(self.client, self.custom_headers)
+
         self._attributes[LLM_PROVIDER] = "aws"
         self._attributes[LLM_SYSTEM] = "aws"
 
+    @staticmethod
+    def _setup_custom_headers(client: Any, custom_headers: Mapping[str, str]) -> None:
+        """Setup custom headers using boto3's event system."""
+        if not custom_headers:
+            return
+
+        def add_custom_headers(request: "AWSPreparedRequest", **kwargs: Any) -> None:
+            request.headers.update(custom_headers)
+
+        client.meta.events.register("before-send.*", add_custom_headers)
+
     @classmethod
     def dependencies(cls) -> list[Dependency]:
         return [Dependency(name="boto3")]
@@ -756,6 +779,9 @@ class BedrockStreamingClient(PlaygroundStreamingClient):
                 aws_secret_access_key=self.aws_secret_access_key,
                 aws_session_token=self.aws_session_token,
             )
+            # Re-setup custom headers for the new client
+            if self.custom_headers:
+                self._setup_custom_headers(self.client, self.custom_headers)
         if self.api == "invoke":
             async for chunk in self._handle_invoke_api(messages, tools, invocation_parameters):
                 yield chunk
@@ -1140,7 +1166,12 @@ class OpenAIStreamingClient(OpenAIBaseStreamingClient):
                 raise BadRequest("An API key is required for OpenAI models")
             api_key = "sk-fake-api-key"
 
-        client = AsyncOpenAI(api_key=api_key, base_url=base_url, timeout=30)
+        client = AsyncOpenAI(
+            api_key=api_key,
+            base_url=base_url,
+            default_headers=model.custom_headers or None,
+            timeout=30,
+        )
         super().__init__(client=client, model=model, credentials=credentials)
         self._attributes[LLM_PROVIDER] = OpenInferenceLLMProviderValues.OPENAI.value
         self._attributes[LLM_SYSTEM] = OpenInferenceLLMSystemValues.OPENAI.value
@@ -1295,6 +1326,7 @@ class AzureOpenAIStreamingClient(OpenAIBaseStreamingClient):
                 api_key=api_key,
                 azure_endpoint=endpoint,
                 api_version=api_version,
+                default_headers=model.custom_headers or None,
             )
         else:
             try:
@@ -1312,6 +1344,7 @@ class AzureOpenAIStreamingClient(OpenAIBaseStreamingClient):
                 ),
                 azure_endpoint=endpoint,
                 api_version=api_version,
+                default_headers=model.custom_headers or None,
             )
         super().__init__(client=client, model=model, credentials=credentials)
         self._attributes[LLM_PROVIDER] = OpenInferenceLLMProviderValues.AZURE.value
@@ -1459,7 +1492,10 @@ class AnthropicStreamingClient(PlaygroundStreamingClient):
         if not api_key:
             raise BadRequest("An API key is required for Anthropic models")
 
-        self.client = anthropic.AsyncAnthropic(api_key=api_key)
+        self.client = anthropic.AsyncAnthropic(
+            api_key=api_key,
+            default_headers=model.custom_headers or None,
+        )
         self.model_name = model.name
         self.rate_limiter = PlaygroundRateLimiter(model.provider_key, anthropic.RateLimitError)
         self.client._client = _HttpxClient(self.client._client, self._attributes)

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/helpers/playground_spans.py

@@ -222,6 +222,7 @@ def get_db_experiment_run(
     *,
     experiment_id: int,
     example_id: int,
+    repetition_number: int,
 ) -> models.ExperimentRun:
     return models.ExperimentRun(
         experiment_id=experiment_id,
@@ -230,7 +231,7 @@ def get_db_experiment_run(
         output=models.ExperimentRunOutput(
             task_output=get_dataset_example_output(db_span),
         ),
-        repetition_number=1,
+        repetition_number=repetition_number,
         start_time=db_span.start_time,
         end_time=db_span.end_time,
         error=db_span.status_message or None,

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/input_types/ChatCompletionInput.py

@@ -26,6 +26,7 @@ class ChatCompletionInput:
     credentials: Optional[list[GenerativeCredentialInput]] = UNSET
     template: Optional[PromptTemplateOptions] = UNSET
     prompt_name: Optional[Identifier] = None
+    repetitions: int
 
 
 @strawberry.input
@@ -36,6 +37,7 @@ class ChatCompletionOverDatasetInput:
     tools: Optional[list[JSON]] = UNSET
     credentials: Optional[list[GenerativeCredentialInput]] = UNSET
     template_format: PromptTemplateFormat = PromptTemplateFormat.MUSTACHE
+    repetitions: int
     dataset_id: GlobalID
     dataset_version_id: Optional[GlobalID] = None
     experiment_name: Optional[str] = None

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/input_types/GenerativeModelInput.py

@@ -2,6 +2,7 @@ from typing import Optional
 
 import strawberry
 from strawberry import UNSET
+from strawberry.scalars import JSON
 
 from phoenix.server.api.types.GenerativeProvider import GenerativeProviderKey
 
@@ -19,3 +20,5 @@ class GenerativeModelInput:
     """ The API version to use for the model. """
     region: Optional[str] = UNSET
     """ The region to use for the model. """
+    custom_headers: Optional[JSON] = UNSET
+    """ Custom headers to use for the model. """

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/input_types/SpanSort.py

@@ -199,7 +199,7 @@ class SpanSort:
                     models.SpanAnnotation.span_rowid == models.Span.id,
                     models.SpanAnnotation.name == eval_name,
                 ),
-            ).order_by(expr)
+            ).order_by(nulls_last(expr))
             return SpanSortConfig(
                 stmt=stmt,
                 orm_expression=eval_result_key.attr.orm_expression,

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/mutations/chat_mutations.py

@@ -112,6 +112,7 @@ class ChatCompletionMutationError:
 @strawberry.type
 class ChatCompletionOverDatasetMutationExamplePayload:
     dataset_example_id: GlobalID
+    repetition_number: int
     experiment_run_id: GlobalID
     result: Union[ChatCompletionMutationPayload, ChatCompletionMutationError]
 
@@ -207,7 +208,12 @@ class ChatCompletionMutationMixin:
         results: list[Union[ChatCompletionMutationPayload, BaseException]] = []
         batch_size = 3
         start_time = datetime.now(timezone.utc)
-        for batch in _get_batches(revisions, batch_size):
+        unbatched_items = [
+            (revision, repetition_number)
+            for revision in revisions
+            for repetition_number in range(1, input.repetitions + 1)
+        ]
+        for batch in _get_batches(unbatched_items, batch_size):
             batch_results = await asyncio.gather(
                 *(
                     cls._chat_completion(
@@ -224,10 +230,11 @@ class ChatCompletionMutationMixin:
                                 variables=revision.input,
                             ),
                             prompt_name=input.prompt_name,
+                            repetitions=repetition_number,
                         ),
                         project_name=project_name,
                     )
-                    for revision in batch
+                    for revision, repetition_number in batch
                 ),
                 return_exceptions=True,
             )
@@ -239,13 +246,13 @@ class ChatCompletionMutationMixin:
             experiment_id=GlobalID(models.Experiment.__name__, str(experiment.id)),
         )
         experiment_runs = []
-        for revision, result in zip(revisions, results):
+        for (revision, repetition_number), result in zip(unbatched_items, results):
             if isinstance(result, BaseException):
                 experiment_run = models.ExperimentRun(
                     experiment_id=experiment.id,
                     dataset_example_id=revision.dataset_example_id,
                     output={},
-                    repetition_number=1,
+                    repetition_number=repetition_number,
                     start_time=start_time,
                     end_time=start_time,
                     error=str(result),
@@ -261,7 +268,7 @@ class ChatCompletionMutationMixin:
                     ),
                     prompt_token_count=db_span.cumulative_llm_token_count_prompt,
                     completion_token_count=db_span.cumulative_llm_token_count_completion,
-                    repetition_number=1,
+                    repetition_number=repetition_number,
                     start_time=db_span.start_time,
                     end_time=db_span.end_time,
                     error=str(result.error_message) if result.error_message else None,
@@ -272,13 +279,16 @@ class ChatCompletionMutationMixin:
             session.add_all(experiment_runs)
             await session.flush()
 
-        for revision, experiment_run, result in zip(revisions, experiment_runs, results):
+        for (revision, repetition_number), experiment_run, result in zip(
+            unbatched_items, experiment_runs, results
+        ):
             dataset_example_id = GlobalID(
                 models.DatasetExample.__name__, str(revision.dataset_example_id)
             )
             experiment_run_id = GlobalID(models.ExperimentRun.__name__, str(experiment_run.id))
             example_payload = ChatCompletionOverDatasetMutationExamplePayload(
                 dataset_example_id=dataset_example_id,
+                repetition_number=repetition_number,
                 experiment_run_id=experiment_run_id,
                 result=result
                 if isinstance(result, ChatCompletionMutationPayload)

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/subscriptions.py

@@ -308,7 +308,7 @@ class Subscription:
                 name=input.experiment_name
                 or _default_playground_experiment_name(input.prompt_name),
                 description=input.experiment_description,
-                repetitions=1,
+                repetitions=input.repetitions,
                 metadata_=input.experiment_metadata or dict(),
                 project_name=project_name,
             )
@@ -327,11 +327,13 @@ class Subscription:
                     llm_client=llm_client,
                     revision=revision,
                     results=results,
+                    repetition_number=repetition_number,
                     experiment_id=experiment.id,
                     project_id=playground_project_id,
                 ),
             )
             for revision in revisions
+            for repetition_number in range(1, input.repetitions + 1)
         ]
         in_progress: list[
             tuple[
@@ -409,6 +411,7 @@ async def _stream_chat_completion_over_dataset_example(
     input: ChatCompletionOverDatasetInput,
     llm_client: PlaygroundStreamingClient,
     revision: models.DatasetExampleRevision,
+    repetition_number: int,
     results: asyncio.Queue[ChatCompletionResult],
     experiment_id: int,
     project_id: int,
@@ -435,7 +438,11 @@ async def _stream_chat_completion_over_dataset_example(
         )
     except TemplateFormatterError as error:
         format_end_time = cast(datetime, normalize_datetime(dt=local_now(), tz=timezone.utc))
-        yield ChatCompletionSubscriptionError(message=str(error), dataset_example_id=example_id)
+        yield ChatCompletionSubscriptionError(
+            message=str(error),
+            dataset_example_id=example_id,
+            repetition_number=repetition_number,
+        )
         await results.put(
             (
                 example_id,
@@ -445,7 +452,7 @@ async def _stream_chat_completion_over_dataset_example(
                     dataset_example_id=revision.dataset_example_id,
                     trace_id=None,
                     output={},
-                    repetition_number=1,
+                    repetition_number=repetition_number,
                     start_time=format_start_time,
                     end_time=format_end_time,
                     error=str(error),
@@ -465,17 +472,24 @@ async def _stream_chat_completion_over_dataset_example(
         ):
             span.add_response_chunk(chunk)
             chunk.dataset_example_id = example_id
+            chunk.repetition_number = repetition_number
             yield chunk
         span.set_attributes(llm_client.attributes)
     db_trace = get_db_trace(span, project_id)
     db_span = get_db_span(span, db_trace)
     db_run = get_db_experiment_run(
-        db_span, db_trace, experiment_id=experiment_id, example_id=revision.dataset_example_id
+        db_span,
+        db_trace,
+        experiment_id=experiment_id,
+        example_id=revision.dataset_example_id,
+        repetition_number=repetition_number,
     )
     await results.put((example_id, db_span, db_run))
     if span.status_message is not None:
         yield ChatCompletionSubscriptionError(
-            message=span.status_message, dataset_example_id=example_id
+            message=span.status_message,
+            dataset_example_id=example_id,
+            repetition_number=repetition_number,
         )
 
 
@@ -511,6 +525,7 @@ async def _chat_completion_result_payloads(
             span=Span(span_rowid=span.id, db_span=span) if span else None,
             experiment_run=to_gql_experiment_run(run),
             dataset_example_id=example_id,
+            repetition_number=run.repetition_number,
         )
 
 

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/types/ChatCompletionSubscriptionPayload.py

@@ -11,6 +11,7 @@ from .Span import Span
 @strawberry.interface
 class ChatCompletionSubscriptionPayload:
     dataset_example_id: Optional[GlobalID] = None
+    repetition_number: Optional[int] = None
 
 
 @strawberry.type

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/types/Project.py

@@ -355,12 +355,16 @@ class Project(Node):
             if sort_config and cursor.sort_column:
                 sort_column = cursor.sort_column
                 compare = operator.lt if sort_config.dir is SortDir.desc else operator.gt
-                stmt = stmt.where(
-                    compare(
-                        tuple_(sort_config.orm_expression, models.Span.id),
-                        (sort_column.value, cursor.rowid),
+                if sort_column.type is CursorSortColumnDataType.NULL:
+                    stmt = stmt.where(sort_config.orm_expression.is_(None))
+                    stmt = stmt.where(compare(models.Span.id, cursor.rowid))
+                else:
+                    stmt = stmt.where(
+                        compare(
+                            tuple_(sort_config.orm_expression, models.Span.id),
+                            (sort_column.value, cursor.rowid),
+                        )
                     )
-                )
             else:
                 stmt = stmt.where(models.Span.id > cursor.rowid)
         stmt = stmt.order_by(cursor_rowid_column)

{arize_phoenix-11.36.0 → arize_phoenix-11.38.0}/src/phoenix/server/api/types/pagination.py

@@ -9,7 +9,7 @@ from strawberry.relay.types import Connection, Edge, NodeType, PageInfo
 from typing_extensions import TypeAlias, assert_never
 
 ID: TypeAlias = int
-CursorSortColumnValue: TypeAlias = Union[str, int, float, datetime]
+CursorSortColumnValue: TypeAlias = Union[str, int, float, datetime, None]
 
 # A type alias for the connection cursor implementation
 CursorString = str
@@ -19,6 +19,7 @@ CURSOR_PREFIX = "connection:"
 
 
 class CursorSortColumnDataType(Enum):
+    NULL = auto()
     STRING = auto()
     INT = auto()
     FLOAT = auto()
@@ -30,6 +31,10 @@ class CursorSortColumn:
     type: CursorSortColumnDataType
     value: CursorSortColumnValue
 
+    def __post_init__(self) -> None:
+        if self.value is None:
+            self.type = CursorSortColumnDataType.NULL
+
     def __str__(self) -> str:
         if isinstance(self.value, str):
             return self.value
@@ -37,7 +42,9 @@ class CursorSortColumn:
             return str(self.value)
         if isinstance(self.value, datetime):
             return self.value.isoformat()
-        assert_never(self.type)
+        if self.value is None:
+            return ""
+        assert_never(self.value)
 
     @classmethod
     def from_string(cls, type: CursorSortColumnDataType, cursor_string: str) -> "CursorSortColumn":
@@ -50,6 +57,8 @@ class CursorSortColumn:
             value = float(cursor_string)
         elif type is CursorSortColumnDataType.DATETIME:
             value = datetime.fromisoformat(cursor_string)
+        elif type is CursorSortColumnDataType.NULL:
+            value = None
         else:
             assert_never(type)
         return cls(type=type, value=value)

arize_phoenix-11.38.0/src/phoenix/server/static/.vite/manifest.json

@@ -0,0 +1,93 @@
+{
+  "_components-BQPHTBfv.js": {
+    "file": "assets/components-BQPHTBfv.js",
+    "name": "components",
+    "imports": [
+      "_vendor-BdjZxMii.js",
+      "_vendor-arizeai-CHYlS8jV.js",
+      "_pages-C0Y17J0T.js",
+      "_vendor-codemirror-Di6t4HnH.js",
+      "_vendor-three-BLWp5bic.js"
+    ]
+  },
+  "_pages-C0Y17J0T.js": {
+    "file": "assets/pages-C0Y17J0T.js",
+    "name": "pages",
+    "imports": [
+      "_vendor-BdjZxMii.js",
+      "_vendor-arizeai-CHYlS8jV.js",
+      "_components-BQPHTBfv.js",
+      "_vendor-codemirror-Di6t4HnH.js",
+      "_vendor-recharts-C9wCDYj3.js"
+    ]
+  },
+  "_vendor-BGzfc4EU.css": {
+    "file": "assets/vendor-BGzfc4EU.css",
+    "src": "_vendor-BGzfc4EU.css"
+  },
+  "_vendor-BdjZxMii.js": {
+    "file": "assets/vendor-BdjZxMii.js",
+    "name": "vendor",
+    "imports": [
+      "_vendor-three-BLWp5bic.js"
+    ],
+    "css": [
+      "assets/vendor-BGzfc4EU.css"
+    ]
+  },
+  "_vendor-arizeai-CHYlS8jV.js": {
+    "file": "assets/vendor-arizeai-CHYlS8jV.js",
+    "name": "vendor-arizeai",
+    "imports": [
+      "_vendor-BdjZxMii.js"
+    ]
+  },
+  "_vendor-codemirror-Di6t4HnH.js": {
+    "file": "assets/vendor-codemirror-Di6t4HnH.js",
+    "name": "vendor-codemirror",
+    "imports": [
+      "_vendor-BdjZxMii.js",
+      "_vendor-shiki-MNnmOotP.js"
+    ],
+    "dynamicImports": [
+      "_vendor-shiki-MNnmOotP.js",
+      "_vendor-shiki-MNnmOotP.js",
+      "_vendor-shiki-MNnmOotP.js"
+    ]
+  },
+  "_vendor-recharts-C9wCDYj3.js": {
+    "file": "assets/vendor-recharts-C9wCDYj3.js",
+    "name": "vendor-recharts",
+    "imports": [
+      "_vendor-BdjZxMii.js"
+    ]
+  },
+  "_vendor-shiki-MNnmOotP.js": {
+    "file": "assets/vendor-shiki-MNnmOotP.js",
+    "name": "vendor-shiki",
+    "isDynamicEntry": true,
+    "imports": [
+      "_vendor-BdjZxMii.js"
+    ]
+  },
+  "_vendor-three-BLWp5bic.js": {
+    "file": "assets/vendor-three-BLWp5bic.js",
+    "name": "vendor-three"
+  },
+  "index.tsx": {
+    "file": "assets/index-BL5BMgJU.js",
+    "name": "index",
+    "src": "index.tsx",
+    "isEntry": true,
+    "imports": [
+      "_vendor-BdjZxMii.js",
+      "_vendor-arizeai-CHYlS8jV.js",
+      "_pages-C0Y17J0T.js",
+      "_components-BQPHTBfv.js",
+      "_vendor-three-BLWp5bic.js",
+      "_vendor-codemirror-Di6t4HnH.js",
+      "_vendor-shiki-MNnmOotP.js",
+      "_vendor-recharts-C9wCDYj3.js"
+    ]
+  }
+}