arize-phoenix 7.8.0__py3-none-any.whl → 7.9.0__py3-none-any.whl

{arize_phoenix-7.8.0.dist-info → arize_phoenix-7.9.0.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: arize-phoenix
-Version: 7.8.0
+Version: 7.9.0
 Summary: AI Observability and Evaluation
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues
@@ -127,7 +127,7 @@ Description-Content-Type: text/markdown
 
 <p align="center">
     <a target="_blank" href="https://phoenix.arize.com" style="background:none">
-        <img alt="phoenix banner" src="https://github.com/Arize-ai/phoenix-assets/blob/main/images/socal/github-large-banner-phoenix.jpg?raw=true" width="auto" height="auto"></img>
+        <img alt="phoenix banner" src="https://github.com/Arize-ai/phoenix-assets/blob/main/images/socal/github-large-banner-phoenix-v2.jpg?raw=true" width="auto" height="auto"></img>
     </a>
     <br/>
     <br/>
@@ -160,6 +160,7 @@ Phoenix is an open-source AI observability platform designed for experimentation
 - [**_Evaluation_**](https://docs.arize.com/phoenix/evaluation/llm-evals) - Leverage LLMs to benchmark your application's performance using response and retrieval evals.
 - [**_Datasets_**](https://docs.arize.com/phoenix/datasets-and-experiments/overview-datasets) - Create versioned datasets of examples for experimentation, evaluation, and fine-tuning.
 - [**_Experiments_**](https://docs.arize.com/phoenix/datasets-and-experiments/overview-datasets#experiments) - Track and evaluate changes to prompts, LLMs, and retrieval.
+- [**_Playground_**](https://docs.arize.com/phoenix/prompt-engineering/overview-prompts)- Optemize prompts, compare models, adjust parameters, and replay traced LLM calls.
 
 Phoenix is vendor and language agnostic with out-of-the-box support for popular frameworks (🦙[LlamaIndex](https://docs.arize.com/phoenix/tracing/integrations-tracing/llamaindex), 🦜⛓[LangChain](https://docs.arize.com/phoenix/tracing/integrations-tracing/langchain), [Haystack](https://docs.arize.com/phoenix/tracing/integrations-tracing/haystack), 🧩[DSPy](https://docs.arize.com/phoenix/tracing/integrations-tracing/dspy)) and LLM providers ([OpenAI](https://docs.arize.com/phoenix/tracing/integrations-tracing/openai), [Bedrock](https://docs.arize.com/phoenix/tracing/integrations-tracing/bedrock), [MistralAI](https://docs.arize.com/phoenix/tracing/integrations-tracing/mistralai), [VertexAI](https://docs.arize.com/phoenix/tracing/integrations-tracing/vertexai), [LiteLLM](https://docs.arize.com/phoenix/tracing/integrations-tracing/litellm), and more). For details on auto-instrumentation, check out the [OpenInference](https://github.com/Arize-ai/openinference) project.
 
@@ -225,6 +226,7 @@ Phoenix is built on top of OpenTelemetry and is vendor, language, and framework
 | [Groq](https://docs.arize.com/phoenix/tracing/integrations-tracing/groq) | `openinference-instrumentation-groq` | [![PyPI Version](https://img.shields.io/pypi/v/openinference-instrumentation-groq.svg)](https://pypi.python.org/pypi/openinference-instrumentation-groq) |
 | [Instructor](https://docs.arize.com/phoenix/tracing/integrations-tracing/instructor) | `openinference-instrumentation-instructor` | [![PyPI Version](https://img.shields.io/pypi/v/openinference-instrumentation-instructor.svg)](https://pypi.python.org/pypi/openinference-instrumentation-instructor) |
 | [Anthropic](https://docs.arize.com/phoenix/tracing/integrations-tracing/anthropic) | `openinference-instrumentation-anthropic` | [![PyPI Version](https://img.shields.io/pypi/v/openinference-instrumentation-anthropic.svg)](https://pypi.python.org/pypi/openinference-instrumentation-anthropic) |
+| [Smolagents](https://huggingface.co/docs/smolagents/en/tutorials/inspect_runs) | `openinference-instrumentation-smolagents` | [![PyPI Version](https://img.shields.io/pypi/v/openinference-instrumentation-smolagents.svg)](https://pypi.python.org/pypi/openinference-instrumentation-smolagents) |
 
 ### JavaScript
 

{arize_phoenix-7.8.0.dist-info → arize_phoenix-7.9.0.dist-info}/RECORD

@@ -6,7 +6,7 @@ phoenix/exceptions.py,sha256=n2L2KKuecrdflB9MsCdAYCiSEvGJptIsfRkXMoJle7A,169
 phoenix/py.typed,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
 phoenix/services.py,sha256=kpW1WL0kiB8XJsO6XycvZVJ-lBkNoenhQ7atCvBoSe8,5365
 phoenix/settings.py,sha256=ht-0oN-sMV6SPXrk7Tu1EZlngpAYkGNLYPhO8DyrdQI,661
-phoenix/version.py,sha256=vZw91QDi0kVrQT2yAQVzBUsHnFXKxQOJTwxJLeLNYNo,22
+phoenix/version.py,sha256=cR9GZGw5gRT-1yq69vN7Bfk48p6j47mGqQA4xv9soHw,22
 phoenix/core/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/core/embedding_dimension.py,sha256=zKGbcvwOXgLf-yrJBpQyKtd-LEOPRKHnUToyAU8Owis,87
 phoenix/core/model.py,sha256=qBFraOtmwCCnWJltKNP18DDG0mULXigytlFsa6YOz6k,4837
@@ -102,7 +102,7 @@ phoenix/server/api/utils.py,sha256=quCBRcusc6PUq9tJq7M8PgwFZp7nXgVAxtbw8feribY,8
 phoenix/server/api/dataloaders/__init__.py,sha256=nt5Lbs34q05GSvNCT2IGrkRW0kg65L8h3bZelzgpkYg,3909
 phoenix/server/api/dataloaders/annotation_summaries.py,sha256=2sHmIDX7n8tuPeBTs9bMKtlMKWn_Ph9awTZqmwn2Owc,5505
 phoenix/server/api/dataloaders/average_experiment_run_latency.py,sha256=GLFoFAbztOH-0FVzzZ77mATIO63UcjB50j3qXiNi-tE,1811
-phoenix/server/api/dataloaders/dataset_example_revisions.py,sha256=YG2VSWOkM7q_-RO8OwE8bHYX8d3_OH34SeaD4pvEt0w,5115
+phoenix/server/api/dataloaders/dataset_example_revisions.py,sha256=xF7M2dg3UmjhdCrscnztCIBBI0cg3RF48IIqvilpc18,4623
 phoenix/server/api/dataloaders/dataset_example_spans.py,sha256=z_MFquqAcJ9wat7BBp7MVeJ9BYuu2EZEdaog52iWDno,1390
 phoenix/server/api/dataloaders/document_evaluation_summaries.py,sha256=9fdROnzp-mymggHwNvpRkCk93LUFxxLy55-j3HP_2HY,5565
 phoenix/server/api/dataloaders/document_evaluations.py,sha256=KCnCItJ2DQOCHvxFp1KK2AStPN1akGlAGOEzEAu6c6I,1246
@@ -135,7 +135,7 @@ phoenix/server/api/dataloaders/cache/two_tier_cache.py,sha256=cmo8FUT3E91R139IEz
 phoenix/server/api/helpers/__init__.py,sha256=m2-xaSPqUiSs91k62JaRDjFNfl-1byxBfY-m_Vxw16U,272
 phoenix/server/api/helpers/dataset_helpers.py,sha256=AMlKY9_e0wnTrTSSQemM5NHfnpwARSytx-m9YK6f6bY,8421
 phoenix/server/api/helpers/experiment_run_filters.py,sha256=DOnVwrmn39eAkk2mwuZP8kIcAnR5jrOgllEwWSjsw94,29893
-phoenix/server/api/helpers/playground_clients.py,sha256=RJNl2WbeAXzK_Am_g9RJsacnLqmC0J0w-TV5cjfq0cg,36832
+phoenix/server/api/helpers/playground_clients.py,sha256=JU42O_Og_LgpjJyH1kndae9xUqHezoOs34Mb2RSkaRo,38827
 phoenix/server/api/helpers/playground_registry.py,sha256=CPLMziFB2wmr-dfbx7VbzO2f8YIG_k5RftzvGXYGQ1w,2570
 phoenix/server/api/helpers/playground_spans.py,sha256=qGk7V7IZK7EkRE1mvZyROpLN5kgOahOZifFzUWmqYFc,16546
 phoenix/server/api/input_types/AddExamplesToDatasetInput.py,sha256=mIQz0S_z8YdrktKIY6RCvtNJ2yZF9pYvTGgasUsI-54,430
@@ -160,7 +160,7 @@ phoenix/server/api/input_types/DimensionFilter.py,sha256=eBYcn7ECSJQlEePvbStqkHB
 phoenix/server/api/input_types/DimensionInput.py,sha256=Vfx5FmiMKey4-EHDQsQRPzSAMRJMN5oVMLDUl4NKAa8,164
 phoenix/server/api/input_types/GenerativeModelInput.py,sha256=h_9dNkz-LBgOLKQ5_ijch4UNGiDb4x5CCC96WyverSg,551
 phoenix/server/api/input_types/Granularity.py,sha256=PiOFWmSMs_w7ZZMNfEWPzIY6S6guI83Vy4NXXSEewGo,2310
-phoenix/server/api/input_types/InvocationParameters.py,sha256=Lg6-SKsx72Lb-jzjJLSfPfmyt8bPz_8JtLpv5-TDcZQ,5154
+phoenix/server/api/input_types/InvocationParameters.py,sha256=I_FvPvY2sXxNd-kFSB96-1DOW7lo-VbqVWzppU_5hy8,5196
 phoenix/server/api/input_types/PatchAnnotationInput.py,sha256=NWhkcbcGNPwfOYsN3wm5YFNNrSc5T-8Y5my74RK99HE,520
 phoenix/server/api/input_types/PatchDatasetExamplesInput.py,sha256=_uMqkAInhLDvzUSASl6HgLNulTsekMcYzyd5J6LF90I,884
 phoenix/server/api/input_types/PatchDatasetInput.py,sha256=OURtTVY8Z_oFEDtKwT1LCMaOK5D4QYo5TVQ6mDrex-g,328
@@ -337,9 +337,9 @@ phoenix/utilities/project.py,sha256=auVpARXkDb-JgeX5f2aStyFIkeKvGwN9l7qrFeJMVxI,
 phoenix/utilities/re.py,sha256=x8Xbk-Wa6qDMAtUd_7JtZvKtrYEuMY-bchB0n163_5c,2006
 phoenix/utilities/span_store.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
 phoenix/utilities/template_formatters.py,sha256=gh9PJD6WEGw7TEYXfSst1UR4pWWwmjxMLrDVQ_CkpkQ,2779
-arize_phoenix-7.8.0.dist-info/METADATA,sha256=EM8prb9RNrEFDt2ZmeTjMnwammFcOhmf-LMhDooHnYU,23052
-arize_phoenix-7.8.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
-arize_phoenix-7.8.0.dist-info/entry_points.txt,sha256=Pgpn8Upxx9P8z8joPXZWl2LlnAlGc3gcQoVchb06X1Q,94
-arize_phoenix-7.8.0.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
-arize_phoenix-7.8.0.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
-arize_phoenix-7.8.0.dist-info/RECORD,,
+arize_phoenix-7.9.0.dist-info/METADATA,sha256=Dm7BoR78lSCEm7pc2Lar9DzCeqHvudTrG4veoMMrOSI,23522
+arize_phoenix-7.9.0.dist-info/WHEEL,sha256=qtCwoSJWgHk21S1Kb4ihdzI2rlJ1ZKaIurTj_ngOhyQ,87
+arize_phoenix-7.9.0.dist-info/entry_points.txt,sha256=Pgpn8Upxx9P8z8joPXZWl2LlnAlGc3gcQoVchb06X1Q,94
+arize_phoenix-7.9.0.dist-info/licenses/IP_NOTICE,sha256=JBqyyCYYxGDfzQ0TtsQgjts41IJoa-hiwDrBjCb9gHM,469
+arize_phoenix-7.9.0.dist-info/licenses/LICENSE,sha256=HFkW9REuMOkvKRACuwLPT0hRydHb3zNg-fdFt94td18,3794
+arize_phoenix-7.9.0.dist-info/RECORD,,

phoenix/server/api/dataloaders/dataset_example_revisions.py

@@ -1,6 +1,6 @@
 from typing import Optional, Union
 
-from sqlalchemy import and_, case, func, null, or_, select
+from sqlalchemy import Integer, case, func, or_, select, union
 from sqlalchemy.sql.expression import literal
 from strawberry.dataloader import DataLoader
 from typing_extensions import TypeAlias
@@ -20,74 +20,50 @@ class DatasetExampleRevisionsDataLoader(DataLoader[Key, Result]):
     def __init__(self, db: DbSessionFactory) -> None:
         super().__init__(
             load_fn=self._load_fn,
-            max_batch_size=200,  # needed to prevent the size of the query from getting too large
+            # Setting max_batch_size to prevent the size of the query from getting too large.
+            # The maximum number of terms is SQLITE_MAX_COMPOUND_SELECT which defaults to 500.
+            # This is needed because of the compound select query below used in transferring
+            # the input data to the database. SQLite in fact has better ways to transfer data,
+            # but unfortunately they're not made available in sqlalchemy yet.
+            max_batch_size=200,
         )
         self._db = db
 
     async def _load_fn(self, keys: list[Key]) -> list[Union[Result, NotFound]]:
-        example_and_version_ids = tuple(
-            set(
-                (example_id, version_id)
-                for example_id, version_id in keys
-                if version_id is not None
-            )
-        )
-        versionless_example_ids = tuple(
-            set(example_id for example_id, version_id in keys if version_id is None)
-        )
-        resolved_example_and_version_ids = (
-            (
+        # sqlalchemy has limited SQLite support for VALUES, so use UNION ALL instead.
+        # For details, see https://github.com/sqlalchemy/sqlalchemy/issues/7228
+        keys_subquery = union(
+            *(
                 select(
-                    models.DatasetExample.id.label("example_id"),
-                    models.DatasetVersion.id.label("version_id"),
-                )
-                .select_from(models.DatasetExample)
-                .join(
-                    models.DatasetVersion,
-                    onclause=literal(True),  # cross join
-                )
-                .where(
-                    or_(
-                        *(
-                            and_(
-                                models.DatasetExample.id == example_id,
-                                models.DatasetVersion.id == version_id,
-                            )
-                            for example_id, version_id in example_and_version_ids
-                        )
-                    )
+                    literal(example_id, Integer).label("example_id"),
+                    literal(version_id, Integer).label("version_id"),
                 )
+                for example_id, version_id in keys
             )
-            .union(
-                select(
-                    models.DatasetExample.id.label("example_id"), null().label("version_id")
-                ).where(models.DatasetExample.id.in_(versionless_example_ids))
-            )
-            .subquery()
-        )
+        ).subquery()
         revision_ids = (
             select(
-                resolved_example_and_version_ids.c.example_id,
-                resolved_example_and_version_ids.c.version_id,
+                keys_subquery.c.example_id,
+                keys_subquery.c.version_id,
                 func.max(models.DatasetExampleRevision.id).label("revision_id"),
             )
-            .select_from(resolved_example_and_version_ids)
+            .select_from(keys_subquery)
             .join(
                 models.DatasetExampleRevision,
-                onclause=resolved_example_and_version_ids.c.example_id
+                onclause=keys_subquery.c.example_id
                 == models.DatasetExampleRevision.dataset_example_id,
             )
             .where(
                 or_(
-                    resolved_example_and_version_ids.c.version_id.is_(None),
-                    models.DatasetExampleRevision.dataset_version_id
-                    <= resolved_example_and_version_ids.c.version_id,
+                    # This query gets the latest `revision_id` for each example:
+                    # - If `version_id` is NOT given, it finds the maximum `revision_id`.
+                    # - If `version_id` is given, it finds the highest `revision_id` whose
+                    #   `version_id` is less than or equal to the one specified.
+                    keys_subquery.c.version_id.is_(None),
+                    models.DatasetExampleRevision.dataset_version_id <= keys_subquery.c.version_id,
                 )
             )
-            .group_by(
-                resolved_example_and_version_ids.c.example_id,
-                resolved_example_and_version_ids.c.version_id,
-            )
+            .group_by(keys_subquery.c.example_id, keys_subquery.c.version_id)
         ).subquery()
         query = (
             select(

phoenix/server/api/helpers/playground_clients.py

@@ -37,6 +37,7 @@ from phoenix.server.api.input_types.InvocationParameters import (
     InvocationParameter,
     InvocationParameterInput,
     JSONInvocationParameter,
+    StringInvocationParameter,
     StringListInvocationParameter,
     extract_parameter,
     validate_invocation_parameters,
@@ -501,6 +502,11 @@ class OpenAIO1StreamingClient(OpenAIStreamingClient):
     @classmethod
     def supported_invocation_parameters(cls) -> list[InvocationParameter]:
         return [
+            StringInvocationParameter(
+                invocation_name="reasoning_effort",
+                label="Reasoning Effort",
+                canonical_name=CanonicalParameterName.REASONING_EFFORT,
+            ),
             IntInvocationParameter(
                 invocation_name="max_completion_tokens",
                 canonical_name=CanonicalParameterName.MAX_COMPLETION_TOKENS,
@@ -523,6 +529,49 @@ 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
+        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,
@@ -532,6 +581,7 @@ class OpenAIO1StreamingClient(OpenAIStreamingClient):
     ) -> Optional["ChatCompletionMessageParam"]:
         from openai.types.chat import (
             ChatCompletionAssistantMessageParam,
+            ChatCompletionDeveloperMessageParam,
             ChatCompletionToolMessageParam,
             ChatCompletionUserMessageParam,
         )
@@ -544,7 +594,12 @@ class OpenAIO1StreamingClient(OpenAIStreamingClient):
                 }
             )
         if role is ChatCompletionMessageRole.SYSTEM:
-            return None  # System messages are not supported for o1 models
+            return ChatCompletionDeveloperMessageParam(
+                {
+                    "content": content,
+                    "role": "developer",
+                }
+            )
         if role is ChatCompletionMessageRole.AI:
             if tool_calls is None:
                 return ChatCompletionAssistantMessageParam(

phoenix/server/api/input_types/InvocationParameters.py

@@ -15,6 +15,7 @@ class CanonicalParameterName(str, Enum):
     RANDOM_SEED = "random_seed"
     TOOL_CHOICE = "tool_choice"
     RESPONSE_FORMAT = "response_format"
+    REASONING_EFFORT = "reasoning_effort"
 
 
 @strawberry.enum

phoenix/version.py

@@ -1 +1 @@
-__version__ = "7.8.0"
+__version__ = "7.9.0"