arize-phoenix 3.0.1__tar.gz → 3.0.3__tar.gz

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: arize-phoenix
-Version: 3.0.1
+Version: 3.0.3
 Summary: ML Observability in your notebook
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues
@@ -23,6 +23,7 @@ Requires-Dist: numpy
 Requires-Dist: openinference-instrumentation-langchain
 Requires-Dist: openinference-instrumentation-llama-index
 Requires-Dist: openinference-instrumentation-openai
+Requires-Dist: openinference-semantic-conventions
 Requires-Dist: opentelemetry-exporter-otlp
 Requires-Dist: opentelemetry-proto
 Requires-Dist: opentelemetry-sdk
@@ -50,7 +51,7 @@ Requires-Dist: hatch; extra == 'dev'
 Requires-Dist: jupyter; extra == 'dev'
 Requires-Dist: langchain>=0.0.334; extra == 'dev'
 Requires-Dist: litellm>=1.0.3; extra == 'dev'
-Requires-Dist: llama-index>=0.9.14; extra == 'dev'
+Requires-Dist: llama-index<0.10.0; extra == 'dev'
 Requires-Dist: nbqa; extra == 'dev'
 Requires-Dist: pandas-stubs<=2.0.2.230605; extra == 'dev'
 Requires-Dist: pre-commit; extra == 'dev'
@@ -64,6 +65,7 @@ Provides-Extra: experimental
 Requires-Dist: tenacity; extra == 'experimental'
 Provides-Extra: llama-index
 Requires-Dist: llama-index==0.9.45; extra == 'llama-index'
+Requires-Dist: openinference-instrumentation-llama-index==0.1.3; extra == 'llama-index'
 Description-Content-Type: text/markdown
 
 <p align="center">

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/pyproject.toml

@@ -43,6 +43,7 @@ dependencies = [
   "opentelemetry-sdk",
   "opentelemetry-proto",
   "opentelemetry-exporter-otlp",
+  "openinference-semantic-conventions",
   "openinference-instrumentation-langchain",
   "openinference-instrumentation-llama-index",
   "openinference-instrumentation-openai",
@@ -64,7 +65,7 @@ dev = [
   "strawberry-graphql[debug-server]==0.208.2",
   "pre-commit",
   "arize[AutoEmbeddings, LLM_Evaluation]",
-  "llama-index>=0.9.14",
+  "llama-index<0.10.0",
   "langchain>=0.0.334",
   "litellm>=1.0.3",
   "google-cloud-aiplatform>=1.3",
@@ -75,6 +76,7 @@ experimental = [
 ]
 llama-index = [
   "llama-index==0.9.45",  # always pin to a version that keeps our notebooks working
+  "openinference-instrumentation-llama-index==0.1.3",
 ]
 
 [project.urls]
@@ -109,7 +111,7 @@ dependencies = [
   "arize",
   "langchain>=0.0.334",
   "litellm>=1.0.3",
-  "llama-index>=0.9.14",
+  "llama-index<0.10.0",
   "openai>=1.0.0",
   "tenacity",
   "nltk==3.8.1",
@@ -129,7 +131,7 @@ dependencies = [
 dependencies = [
   "mypy==1.5.1",
   "pydantic==v1.10.14",  # for mypy
-  "llama-index>=0.9.14",
+  "llama-index<0.10.0",
   "pandas-stubs<=2.0.2.230605",  # version 2.0.3.230814 is causing a dependency conflict.
   "types-psutil",
   "types-tqdm",

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/core/traces.py

@@ -20,12 +20,13 @@ from typing import (
 
 import opentelemetry.proto.trace.v1.trace_pb2 as otlp
 from ddsketch import DDSketch
+from openinference.semconv.trace import SpanAttributes
 from sortedcontainers import SortedKeyList
 from typing_extensions import TypeAlias
 from wrapt import ObjectProxy
 
+import phoenix.trace.schemas
 from phoenix.datetime_utils import right_open_time_range
-from phoenix.trace import semantic_conventions
 from phoenix.trace.otel import decode
 from phoenix.trace.schemas import (
     ATTRIBUTE_PREFIX,
@@ -33,12 +34,10 @@ from phoenix.trace.schemas import (
     CONTEXT_PREFIX,
     ComputedAttributes,
     Span,
-    SpanAttributes,
     SpanID,
     SpanStatusCode,
     TraceID,
 )
-from phoenix.trace.semantic_conventions import RETRIEVAL_DOCUMENTS
 
 END_OF_QUEUE = None  # sentinel value for queue termination
 
@@ -50,9 +49,9 @@ SPAN_ID = CONTEXT_PREFIX + "span_id"
 PARENT_ID = "parent_id"
 START_TIME = "start_time"
 END_TIME = "end_time"
-LLM_TOKEN_COUNT_TOTAL = ATTRIBUTE_PREFIX + semantic_conventions.LLM_TOKEN_COUNT_TOTAL
-LLM_TOKEN_COUNT_PROMPT = ATTRIBUTE_PREFIX + semantic_conventions.LLM_TOKEN_COUNT_PROMPT
-LLM_TOKEN_COUNT_COMPLETION = ATTRIBUTE_PREFIX + semantic_conventions.LLM_TOKEN_COUNT_COMPLETION
+LLM_TOKEN_COUNT_TOTAL = ATTRIBUTE_PREFIX + SpanAttributes.LLM_TOKEN_COUNT_TOTAL
+LLM_TOKEN_COUNT_PROMPT = ATTRIBUTE_PREFIX + SpanAttributes.LLM_TOKEN_COUNT_PROMPT
+LLM_TOKEN_COUNT_COMPLETION = ATTRIBUTE_PREFIX + SpanAttributes.LLM_TOKEN_COUNT_COMPLETION
 
 
 class ReadableSpan(ObjectProxy):  # type: ignore
@@ -73,7 +72,9 @@ class ReadableSpan(ObjectProxy):  # type: ignore
     @property
     def span(self) -> Span:
         span = decode(self._self_otlp_span)
-        span.attributes.update(cast(SpanAttributes, self._self_computed_values))
+        span.attributes.update(
+            cast(phoenix.trace.schemas.SpanAttributes, self._self_computed_values)
+        )
         # TODO: compute latency rank percent (which can change depending on how
         # many spans already ingested).
         return span
@@ -333,9 +334,13 @@ class Traces:
             self._token_count_total -= existing_span[LLM_TOKEN_COUNT_TOTAL] or 0
         self._token_count_total += new_span[LLM_TOKEN_COUNT_TOTAL] or 0
         # Update number of documents
-        num_documents_update = len(new_span.attributes.get(RETRIEVAL_DOCUMENTS) or ())
+        num_documents_update = len(
+            new_span.attributes.get(SpanAttributes.RETRIEVAL_DOCUMENTS) or ()
+        )
         if existing_span:
-            num_documents_update -= len(existing_span.attributes.get(RETRIEVAL_DOCUMENTS) or ())
+            num_documents_update -= len(
+                existing_span.attributes.get(SpanAttributes.RETRIEVAL_DOCUMENTS) or ()
+            )
         if num_documents_update:
             self._num_documents[span_id] += num_documents_update
         # Process previously orphaned spans, if any.

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/experimental/evals/functions/classify.py

@@ -3,6 +3,7 @@ from __future__ import annotations
 import logging
 import warnings
 from collections import defaultdict
+from itertools import product
 from typing import (
     Any,
     DefaultDict,
@@ -18,6 +19,7 @@ from typing import (
 )
 
 import pandas as pd
+from openinference.semconv.trace import DocumentAttributes, SpanAttributes
 from pandas import DataFrame
 from typing_extensions import TypeAlias
 
@@ -40,9 +42,12 @@ from phoenix.experimental.evals.utils import (
     parse_openai_function_call,
     snap_to_rail,
 )
-from phoenix.trace.semantic_conventions import DOCUMENT_CONTENT, INPUT_VALUE, RETRIEVAL_DOCUMENTS
 from phoenix.utilities.logging import printif
 
+DOCUMENT_CONTENT = DocumentAttributes.DOCUMENT_CONTENT
+INPUT_VALUE = SpanAttributes.INPUT_VALUE
+RETRIEVAL_DOCUMENTS = SpanAttributes.RETRIEVAL_DOCUMENTS
+
 logger = logging.getLogger(__name__)
 
 
@@ -54,8 +59,7 @@ Label: TypeAlias = str
 Score: TypeAlias = Optional[float]
 Explanation: TypeAlias = Optional[str]
 Record: TypeAlias = Mapping[str, Any]
-EvaluatorIndex: TypeAlias = int
-RowIndex: TypeAlias = Any
+Index: TypeAlias = int
 
 # snapped_response, explanation, response
 ParsedLLMResponse: TypeAlias = Tuple[Optional[str], Optional[str], str]
@@ -343,8 +347,6 @@ def _get_contents_from_openinference_documents(documents: Iterable[Any]) -> List
 
 
 class RunEvalsPayload(NamedTuple):
-    evaluator_index: EvaluatorIndex
-    row_index: RowIndex
     evaluator: LLMEvaluator
     record: Record
 
@@ -404,23 +406,21 @@ def run_evals(
 
     async def _arun_eval(
         payload: RunEvalsPayload,
-    ) -> Tuple[EvaluatorIndex, RowIndex, Label, Score, Explanation]:
-        label, score, explanation = await payload.evaluator.aevaluate(
+    ) -> Tuple[Label, Score, Explanation]:
+        return await payload.evaluator.aevaluate(
             payload.record,
             provide_explanation=provide_explanation,
             use_function_calling_if_available=use_function_calling_if_available,
         )
-        return payload.evaluator_index, payload.row_index, label, score, explanation
 
     def _run_eval(
         payload: RunEvalsPayload,
-    ) -> Tuple[EvaluatorIndex, RowIndex, Label, Score, Explanation]:
-        label, score, explanation = payload.evaluator.evaluate(
+    ) -> Tuple[Label, Score, Explanation]:
+        return payload.evaluator.evaluate(
             payload.record,
             provide_explanation=provide_explanation,
             use_function_calling_if_available=use_function_calling_if_available,
         )
-        return payload.evaluator_index, payload.row_index, label, score, explanation
 
     executor = get_executor_on_sync_context(
         _run_eval,
@@ -428,24 +428,20 @@ def run_evals(
         concurrency=concurrency,
         tqdm_bar_format=get_tqdm_progress_bar_formatter("run_evals"),
         exit_on_error=True,
-        fallback_return_value=(None, None),
+        fallback_return_value=(None, None, None),
     )
+
+    total_records = len(dataframe)
     payloads = [
-        RunEvalsPayload(
-            evaluator_index=evaluator_index,
-            row_index=row_index,
-            evaluator=evaluator,
-            record=row.to_dict(),
-        )
-        # use the position of the row rather than the dataframe index, which is used
-        # to ensure the output dataframe has the same row order as the input dataframe
-        for row_index, (_, row) in enumerate(dataframe.iterrows())
-        for evaluator_index, evaluator in enumerate(evaluators)
+        RunEvalsPayload(evaluator=evaluator, record=row)
+        for evaluator, (_, row) in product(evaluators, dataframe.iterrows())
     ]
-    eval_results: List[DefaultDict[RowIndex, Dict[ColumnName, Union[Label, Explanation]]]] = [
+    eval_results: List[DefaultDict[Index, Dict[ColumnName, Union[Label, Explanation]]]] = [
         defaultdict(dict) for _ in range(len(evaluators))
     ]
-    for evaluator_index, row_index, label, score, explanation in executor.run(payloads):
+    for index, (label, score, explanation) in enumerate(executor.run(payloads)):
+        evaluator_index = index // total_records
+        row_index = index % total_records
         eval_results[evaluator_index][row_index]["label"] = label
         eval_results[evaluator_index][row_index]["score"] = score
         if provide_explanation:

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/experimental/evals/utils/__init__.py

@@ -32,7 +32,7 @@ def download_benchmark_dataset(task: str, dataset_name: str) -> pd.DataFrame:
         pandas.DataFrame: A pandas dataframe containing the data.
     """
     jsonl_file_name = f"{dataset_name}.jsonl"
-    url = f"http://storage.googleapis.com/arize-assets/phoenix/evals/{task}/{jsonl_file_name}.zip"
+    url = f"http://storage.googleapis.com/arize-phoenix-assets/evals/{task}/{jsonl_file_name}.zip"
     try:
         with urlopen(url) as response:
             zip_byte_stream = BytesIO(response.read())

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/server/api/input_types/SpanSort.py

@@ -4,6 +4,7 @@ from typing import Any, Iterable, Iterator, Optional, Protocol
 
 import pandas as pd
 import strawberry
+from openinference.semconv.trace import SpanAttributes
 from strawberry import UNSET
 from typing_extensions import assert_never
 
@@ -13,7 +14,6 @@ from phoenix.core.traces import (
     START_TIME,
 )
 from phoenix.server.api.types.SortDir import SortDir
-from phoenix.trace import semantic_conventions
 from phoenix.trace.schemas import ComputedAttributes, Span, SpanID
 
 
@@ -22,9 +22,9 @@ class SpanColumn(Enum):
     startTime = START_TIME
     endTime = END_TIME
     latencyMs = ComputedAttributes.LATENCY_MS.value
-    tokenCountTotal = semantic_conventions.LLM_TOKEN_COUNT_TOTAL
-    tokenCountPrompt = semantic_conventions.LLM_TOKEN_COUNT_PROMPT
-    tokenCountCompletion = semantic_conventions.LLM_TOKEN_COUNT_COMPLETION
+    tokenCountTotal = SpanAttributes.LLM_TOKEN_COUNT_TOTAL
+    tokenCountPrompt = SpanAttributes.LLM_TOKEN_COUNT_PROMPT
+    tokenCountCompletion = SpanAttributes.LLM_TOKEN_COUNT_COMPLETION
     cumulativeTokenCountTotal = ComputedAttributes.CUMULATIVE_LLM_TOKEN_COUNT_TOTAL.value
     cumulativeTokenCountPrompt = ComputedAttributes.CUMULATIVE_LLM_TOKEN_COUNT_PROMPT.value
     cumulativeTokenCountCompletion = ComputedAttributes.CUMULATIVE_LLM_TOKEN_COUNT_COMPLETION.value

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/server/api/types/Span.py

@@ -5,6 +5,7 @@ from enum import Enum
 from typing import Any, DefaultDict, Dict, List, Mapping, Optional, Sized, cast
 
 import strawberry
+from openinference.semconv.trace import EmbeddingAttributes, SpanAttributes
 from strawberry import ID, UNSET
 from strawberry.types import Info
 
@@ -15,19 +16,17 @@ from phoenix.server.api.types.DocumentRetrievalMetrics import DocumentRetrievalM
 from phoenix.server.api.types.Evaluation import DocumentEvaluation, SpanEvaluation
 from phoenix.server.api.types.MimeType import MimeType
 from phoenix.trace.schemas import ComputedAttributes, SpanID
-from phoenix.trace.semantic_conventions import (
-    EMBEDDING_EMBEDDINGS,
-    EMBEDDING_VECTOR,
-    EXCEPTION_MESSAGE,
-    INPUT_MIME_TYPE,
-    INPUT_VALUE,
-    LLM_TOKEN_COUNT_COMPLETION,
-    LLM_TOKEN_COUNT_PROMPT,
-    LLM_TOKEN_COUNT_TOTAL,
-    OUTPUT_MIME_TYPE,
-    OUTPUT_VALUE,
-    RETRIEVAL_DOCUMENTS,
-)
+
+EMBEDDING_EMBEDDINGS = SpanAttributes.EMBEDDING_EMBEDDINGS
+EMBEDDING_VECTOR = EmbeddingAttributes.EMBEDDING_VECTOR
+INPUT_MIME_TYPE = SpanAttributes.INPUT_MIME_TYPE
+INPUT_VALUE = SpanAttributes.INPUT_VALUE
+LLM_TOKEN_COUNT_COMPLETION = SpanAttributes.LLM_TOKEN_COUNT_COMPLETION
+LLM_TOKEN_COUNT_PROMPT = SpanAttributes.LLM_TOKEN_COUNT_PROMPT
+LLM_TOKEN_COUNT_TOTAL = SpanAttributes.LLM_TOKEN_COUNT_TOTAL
+OUTPUT_MIME_TYPE = SpanAttributes.OUTPUT_MIME_TYPE
+OUTPUT_VALUE = SpanAttributes.OUTPUT_VALUE
+RETRIEVAL_DOCUMENTS = SpanAttributes.RETRIEVAL_DOCUMENTS
 
 
 @strawberry.enum
@@ -87,7 +86,7 @@ class SpanEvent:
     ) -> "SpanEvent":
         return SpanEvent(
             name=event.name,
-            message=cast(str, event.attributes.get(EXCEPTION_MESSAGE) or ""),
+            message=cast(str, event.attributes.get(trace_schema.EXCEPTION_MESSAGE) or ""),
             timestamp=event.timestamp,
         )
 

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/dsl/filter.py

@@ -1,4 +1,5 @@
 import ast
+import inspect
 import sys
 from dataclasses import dataclass, field
 from difflib import SequenceMatcher
@@ -15,10 +16,10 @@ from typing import (
     cast,
 )
 
+from openinference.semconv import trace
 from typing_extensions import TypeGuard
 
 import phoenix.trace.v1 as pb
-from phoenix.trace import semantic_conventions
 from phoenix.trace.dsl.missing import MISSING
 from phoenix.trace.schemas import COMPUTED_PREFIX, ComputedAttributes, Span, SpanID
 
@@ -137,9 +138,11 @@ def _allowed_replacements() -> Iterator[Tuple[str, ast.expr]]:
         yield "span.context." + source_segment, ast_replacement
 
     for field_name in (
-        getattr(semantic_conventions, variable_name)
-        for variable_name in dir(semantic_conventions)
-        if variable_name.isupper()
+        getattr(klass, attr)
+        for name in dir(trace)
+        if name.endswith("Attributes") and inspect.isclass(klass := getattr(trace, name))
+        for attr in dir(klass)
+        if attr.isupper()
     ):
         source_segment = field_name
         ast_replacement = _ast_replacement(f"span.attributes.get('{field_name}')")

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/dsl/helpers.py

@@ -1,15 +1,15 @@
 from typing import List, Optional, Protocol, Union, cast
 
 import pandas as pd
+from openinference.semconv.trace import DocumentAttributes, SpanAttributes
 
 from phoenix.trace.dsl import SpanQuery
-from phoenix.trace.semantic_conventions import (
-    DOCUMENT_CONTENT,
-    DOCUMENT_SCORE,
-    INPUT_VALUE,
-    OUTPUT_VALUE,
-    RETRIEVAL_DOCUMENTS,
-)
+
+DOCUMENT_CONTENT = DocumentAttributes.DOCUMENT_CONTENT
+DOCUMENT_SCORE = DocumentAttributes.DOCUMENT_SCORE
+INPUT_VALUE = SpanAttributes.INPUT_VALUE
+OUTPUT_VALUE = SpanAttributes.OUTPUT_VALUE
+RETRIEVAL_DOCUMENTS = SpanAttributes.RETRIEVAL_DOCUMENTS
 
 INPUT = {"input": INPUT_VALUE}
 OUTPUT = {"output": OUTPUT_VALUE}

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/dsl/query.py

@@ -19,13 +19,15 @@ from typing import (
 )
 
 import pandas as pd
+from openinference.semconv.trace import SpanAttributes
 
 from phoenix.trace.dsl import SpanFilter
 from phoenix.trace.dsl.filter import SupportsGetSpanEvaluation
 from phoenix.trace.schemas import ATTRIBUTE_PREFIX, CONTEXT_PREFIX, Span
-from phoenix.trace.semantic_conventions import RETRIEVAL_DOCUMENTS
 from phoenix.trace.span_json_encoder import span_to_json
 
+RETRIEVAL_DOCUMENTS = SpanAttributes.RETRIEVAL_DOCUMENTS
+
 _SPAN_ID = "context.span_id"
 _PRESCRIBED_POSITION_PREFIXES = {
     RETRIEVAL_DOCUMENTS: "document_",

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/errors.py

@@ -3,3 +3,7 @@ from phoenix.exceptions import PhoenixException
 
 class InvalidParquetMetadataError(PhoenixException):
     pass
+
+
+class IncompatibleLibraryVersionError(PhoenixException):
+    pass

arize_phoenix-3.0.3/src/phoenix/trace/llama_index/__init__.py

@@ -0,0 +1,3 @@
+from .callback import OpenInferenceTraceCallbackHandler
+
+__all__ = ["OpenInferenceTraceCallbackHandler"]

arize_phoenix-3.0.3/src/phoenix/trace/llama_index/callback.py

@@ -0,0 +1,77 @@
+import logging
+from importlib.metadata import PackageNotFoundError, version
+from importlib.util import find_spec
+from typing import Any
+
+from opentelemetry import trace as trace_api
+from opentelemetry.sdk import trace as trace_sdk
+from opentelemetry.sdk.trace.export import SimpleSpanProcessor
+
+from phoenix.trace.errors import IncompatibleLibraryVersionError
+from phoenix.trace.exporter import _OpenInferenceExporter
+from phoenix.trace.tracer import _show_deprecation_warnings
+
+logger = logging.getLogger(__name__)
+
+LLAMA_INDEX_MODERN_VERSION = (0, 10, 0)
+INSTRUMENTATION_MODERN_VERSION = (1, 0, 0)
+
+
+def _check_instrumentation_compatibility() -> bool:
+    if find_spec("llama_index") is None:
+        raise PackageNotFoundError("Missing `llama-index`. Install with `pip install llama-index`.")
+    # split the version string into a tuple of integers
+    llama_index_version_str = version("llama-index")
+    llama_index_version = tuple(map(int, llama_index_version_str.split(".")[:3]))
+    instrumentation_version_str = version("openinference-instrumentation-llama-index")
+    instrumentation_version = tuple(map(int, instrumentation_version_str.split(".")[:3]))
+    # check if the llama_index version is compatible with the instrumentation version
+    if (
+        llama_index_version < LLAMA_INDEX_MODERN_VERSION
+        and instrumentation_version >= INSTRUMENTATION_MODERN_VERSION
+    ):
+        raise IncompatibleLibraryVersionError(
+            f"llama-index v{llama_index_version_str} is not compatible with "
+            f"openinference-instrumentation-llama-index v{instrumentation_version_str}."
+            "Please either migrate llama-index to at least 0.10.0 or downgrade "
+            "openinference-instrumentation-llama-index via "
+            "`pip install 'openinference-instrumentation-llama-index<1.0.0'`."
+        )
+    elif (
+        llama_index_version >= LLAMA_INDEX_MODERN_VERSION
+        and instrumentation_version < INSTRUMENTATION_MODERN_VERSION
+    ):
+        raise IncompatibleLibraryVersionError(
+            f"llama-index v{llama_index_version_str} is not compatible with "
+            f"openinference-instrumentation-llama-index v{instrumentation_version_str}."
+            "Please upgrade openinference-instrumentation-llama-index to at least 1.0.0"
+            "`pip install 'openinference-instrumentation-llama-index>=1.0.0'`."
+        )
+    # if the versions are compatible, return True
+    return True
+
+
+if _check_instrumentation_compatibility():
+    from openinference.instrumentation.llama_index._callback import (
+        OpenInferenceTraceCallbackHandler as _OpenInferenceTraceCallbackHandler,
+    )
+    from openinference.instrumentation.llama_index.version import (
+        __version__,
+    )
+
+
+class OpenInferenceTraceCallbackHandler(_OpenInferenceTraceCallbackHandler):
+    """Callback handler for storing LLM application trace data in OpenInference format.
+    OpenInference is an open standard for capturing and storing AI model
+    inferences. It enables production LLMapp servers to seamlessly integrate
+    with LLM observability solutions such as Arize and Phoenix.
+
+    For more information on the specification, see
+    https://github.com/Arize-ai/openinference
+    """
+
+    def __init__(self, *args: Any, **kwargs: Any) -> None:
+        _show_deprecation_warnings(self, *args, **kwargs)
+        tracer_provider = trace_sdk.TracerProvider()
+        tracer_provider.add_span_processor(SimpleSpanProcessor(_OpenInferenceExporter()))
+        super().__init__(trace_api.get_tracer(__name__, __version__, tracer_provider))

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/otel.py

@@ -1,3 +1,4 @@
+import inspect
 import json
 from binascii import hexlify, unhexlify
 from datetime import datetime, timezone
@@ -21,12 +22,23 @@ from typing import (
 
 import numpy as np
 import opentelemetry.proto.trace.v1.trace_pb2 as otlp
+from openinference.semconv import trace
+from openinference.semconv.trace import (
+    DocumentAttributes,
+    EmbeddingAttributes,
+    MessageAttributes,
+    SpanAttributes,
+    ToolCallAttributes,
+)
 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
 
-import phoenix.trace.semantic_conventions as sem_conv
 from phoenix.trace.schemas import (
+    EXCEPTION_ESCAPED,
+    EXCEPTION_MESSAGE,
+    EXCEPTION_STACKTRACE,
+    EXCEPTION_TYPE,
     MimeType,
     Span,
     SpanContext,
@@ -37,18 +49,38 @@ from phoenix.trace.schemas import (
     SpanStatusCode,
     TraceID,
 )
-from phoenix.trace.semantic_conventions import (
-    DOCUMENT_METADATA,
-    EXCEPTION_ESCAPED,
-    EXCEPTION_MESSAGE,
-    EXCEPTION_STACKTRACE,
-    EXCEPTION_TYPE,
-    INPUT_MIME_TYPE,
-    LLM_PROMPT_TEMPLATE_VARIABLES,
-    OPENINFERENCE_SPAN_KIND,
-    OUTPUT_MIME_TYPE,
-    TOOL_PARAMETERS,
-)
+
+DOCUMENT_CONTENT = DocumentAttributes.DOCUMENT_CONTENT
+DOCUMENT_ID = DocumentAttributes.DOCUMENT_ID
+DOCUMENT_METADATA = DocumentAttributes.DOCUMENT_METADATA
+EMBEDDING_EMBEDDINGS = SpanAttributes.EMBEDDING_EMBEDDINGS
+EMBEDDING_MODEL_NAME = SpanAttributes.EMBEDDING_MODEL_NAME
+EMBEDDING_TEXT = EmbeddingAttributes.EMBEDDING_TEXT
+EMBEDDING_VECTOR = EmbeddingAttributes.EMBEDDING_VECTOR
+INPUT_MIME_TYPE = SpanAttributes.INPUT_MIME_TYPE
+INPUT_VALUE = SpanAttributes.INPUT_VALUE
+LLM_INPUT_MESSAGES = SpanAttributes.LLM_INPUT_MESSAGES
+LLM_INVOCATION_PARAMETERS = SpanAttributes.LLM_INVOCATION_PARAMETERS
+LLM_MODEL_NAME = SpanAttributes.LLM_MODEL_NAME
+LLM_OUTPUT_MESSAGES = SpanAttributes.LLM_OUTPUT_MESSAGES
+LLM_PROMPTS = SpanAttributes.LLM_PROMPTS
+LLM_TOKEN_COUNT_COMPLETION = SpanAttributes.LLM_TOKEN_COUNT_COMPLETION
+LLM_TOKEN_COUNT_PROMPT = SpanAttributes.LLM_TOKEN_COUNT_PROMPT
+LLM_TOKEN_COUNT_TOTAL = SpanAttributes.LLM_TOKEN_COUNT_TOTAL
+MESSAGE_CONTENT = MessageAttributes.MESSAGE_CONTENT
+MESSAGE_FUNCTION_CALL_ARGUMENTS_JSON = MessageAttributes.MESSAGE_FUNCTION_CALL_ARGUMENTS_JSON
+MESSAGE_FUNCTION_CALL_NAME = MessageAttributes.MESSAGE_FUNCTION_CALL_NAME
+MESSAGE_ROLE = MessageAttributes.MESSAGE_ROLE
+MESSAGE_TOOL_CALLS = MessageAttributes.MESSAGE_TOOL_CALLS
+OPENINFERENCE_SPAN_KIND = SpanAttributes.OPENINFERENCE_SPAN_KIND
+OUTPUT_MIME_TYPE = SpanAttributes.OUTPUT_MIME_TYPE
+OUTPUT_VALUE = SpanAttributes.OUTPUT_VALUE
+RETRIEVAL_DOCUMENTS = SpanAttributes.RETRIEVAL_DOCUMENTS
+TOOL_CALL_FUNCTION_ARGUMENTS_JSON = ToolCallAttributes.TOOL_CALL_FUNCTION_ARGUMENTS_JSON
+TOOL_CALL_FUNCTION_NAME = ToolCallAttributes.TOOL_CALL_FUNCTION_NAME
+TOOL_PARAMETERS = SpanAttributes.TOOL_PARAMETERS
+LLM_PROMPT_TEMPLATE = SpanAttributes.LLM_PROMPT_TEMPLATE
+LLM_PROMPT_TEMPLATE_VARIABLES = SpanAttributes.LLM_PROMPT_TEMPLATE_VARIABLES
 
 
 def decode(otlp_span: otlp.Span) -> Span:
@@ -186,7 +218,13 @@ def _decode_status(otlp_status: otlp.Status) -> Tuple[SpanStatusCode, StatusMess
 
 
 _SEMANTIC_CONVENTIONS: List[str] = sorted(
-    (getattr(sem_conv, name) for name in dir(sem_conv) if name.isupper()),
+    (
+        getattr(klass, attr)
+        for name in dir(trace)
+        if name.endswith("Attributes") and inspect.isclass(klass := getattr(trace, name))
+        for attr in dir(klass)
+        if attr.isupper()
+    ),
     reverse=True,
 )  # sorted so the longer strings go first
 

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/schemas.py

@@ -4,12 +4,10 @@ from enum import Enum
 from typing import Any, Dict, List, Optional, Union
 from uuid import UUID
 
-from phoenix.trace.semantic_conventions import (
-    EXCEPTION_ESCAPED,
-    EXCEPTION_MESSAGE,
-    EXCEPTION_STACKTRACE,
-    EXCEPTION_TYPE,
-)
+EXCEPTION_TYPE = "exception.type"
+EXCEPTION_MESSAGE = "exception.message"
+EXCEPTION_ESCAPED = "exception.escaped"
+EXCEPTION_STACKTRACE = "exception.stacktrace"
 
 
 class SpanStatusCode(Enum):

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/span_json_decoder.py

@@ -2,7 +2,10 @@ import json
 from datetime import datetime
 from typing import Any, Dict, Optional
 
+from openinference.semconv.trace import SpanAttributes
+
 from phoenix.trace.schemas import (
+    EXCEPTION_MESSAGE,
     MimeType,
     Span,
     SpanContext,
@@ -14,11 +17,9 @@ from phoenix.trace.schemas import (
     SpanStatusCode,
     TraceID,
 )
-from phoenix.trace.semantic_conventions import (
-    EXCEPTION_MESSAGE,
-    INPUT_MIME_TYPE,
-    OUTPUT_MIME_TYPE,
-)
+
+INPUT_MIME_TYPE = SpanAttributes.INPUT_MIME_TYPE
+OUTPUT_MIME_TYPE = SpanAttributes.OUTPUT_MIME_TYPE
 
 
 def json_to_attributes(obj: Optional[Dict[str, Any]]) -> Dict[str, Any]:

{arize_phoenix-3.0.1 → arize_phoenix-3.0.3}/src/phoenix/trace/span_json_encoder.py

@@ -5,12 +5,7 @@ from enum import Enum
 from typing import Any, List
 from uuid import UUID
 
-from .schemas import (
-    Span,
-    SpanContext,
-    SpanConversationAttributes,
-    SpanEvent,
-)
+from phoenix.trace.schemas import Span, SpanContext, SpanConversationAttributes, SpanEvent
 
 
 class SpanJSONEncoder(json.JSONEncoder):