arize-phoenix 4.5.0__tar.gz → 4.6.2__tar.gz

{arize_phoenix-4.5.0 → arize_phoenix-4.6.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: arize-phoenix
-Version: 4.5.0
+Version: 4.6.2
 Summary: AI Observability and Evaluation
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues
@@ -20,25 +20,28 @@ Requires-Python: <3.13,>=3.8
 Requires-Dist: aioitertools
 Requires-Dist: aiosqlite
 Requires-Dist: alembic<2,>=1.3.0
+Requires-Dist: arize-phoenix-evals>=0.13.1
 Requires-Dist: cachetools
 Requires-Dist: grpcio
 Requires-Dist: hdbscan>=0.8.33
+Requires-Dist: httpx
 Requires-Dist: jinja2
 Requires-Dist: numpy<2
 Requires-Dist: openinference-instrumentation
 Requires-Dist: openinference-instrumentation-langchain>=0.1.12
 Requires-Dist: openinference-instrumentation-llama-index>=1.2.0
 Requires-Dist: openinference-instrumentation-openai>=0.1.4
-Requires-Dist: openinference-semantic-conventions>=0.1.5
+Requires-Dist: openinference-semantic-conventions>=0.1.9
 Requires-Dist: opentelemetry-exporter-otlp
 Requires-Dist: opentelemetry-proto>=1.12.0
 Requires-Dist: opentelemetry-sdk
 Requires-Dist: opentelemetry-semantic-conventions
-Requires-Dist: pandas
+Requires-Dist: pandas>=1.0
 Requires-Dist: protobuf<6.0,>=3.20
 Requires-Dist: psutil
 Requires-Dist: pyarrow
-Requires-Dist: requests
+Requires-Dist: python-multipart
+Requires-Dist: pyyaml
 Requires-Dist: scikit-learn
 Requires-Dist: scipy
 Requires-Dist: sqlalchemy[asyncio]<3,>=2.0.4
@@ -77,6 +80,7 @@ Requires-Dist: llama-index>=0.10.3; extra == 'dev'
 Requires-Dist: nbqa; extra == 'dev'
 Requires-Dist: pandas-stubs==2.0.3.230814; (python_version < '3.9') and extra == 'dev'
 Requires-Dist: pandas-stubs==2.2.2.240603; (python_version >= '3.9') and extra == 'dev'
+Requires-Dist: pandas>=1.0; extra == 'dev'
 Requires-Dist: pre-commit; extra == 'dev'
 Requires-Dist: prometheus-client; extra == 'dev'
 Requires-Dist: psycopg[binary]; extra == 'dev'
@@ -86,12 +90,16 @@ Requires-Dist: pytest-postgresql; extra == 'dev'
 Requires-Dist: pytest==8.2.2; extra == 'dev'
 Requires-Dist: ruff==0.4.9; extra == 'dev'
 Requires-Dist: strawberry-graphql[debug-server,opentelemetry]==0.235.0; extra == 'dev'
+Requires-Dist: tabulate; extra == 'dev'
+Requires-Dist: types-tabulate; extra == 'dev'
 Provides-Extra: evals
-Requires-Dist: arize-phoenix-evals>=0.3.0; extra == 'evals'
 Provides-Extra: experimental
-Requires-Dist: tenacity; extra == 'experimental'
 Provides-Extra: llama-index
-Requires-Dist: llama-index==0.10.44; extra == 'llama-index'
+Requires-Dist: llama-index-agent-openai==0.2.7; extra == 'llama-index'
+Requires-Dist: llama-index-embeddings-openai==0.1.10; extra == 'llama-index'
+Requires-Dist: llama-index-llms-openai==0.1.24; extra == 'llama-index'
+Requires-Dist: llama-index-readers-file==0.1.25; extra == 'llama-index'
+Requires-Dist: llama-index==0.10.51; extra == 'llama-index'
 Provides-Extra: pg
 Requires-Dist: asyncpg; extra == 'pg'
 Description-Content-Type: text/markdown
@@ -127,7 +135,7 @@ Description-Content-Type: text/markdown
 
 Phoenix is an open-source AI observability platform designed for experimentation, evaluation, and troubleshooting. It provides:
 
--   **_Tracing_** - Trace your LLM application's runtime using using OpenTelemetry-based instrumentation.
+-   **_Tracing_** - Trace your LLM application's runtime using OpenTelemetry-based instrumentation.
 -   **_Evaluation_** - Leverage LLMs to benchmark your application's performance using response and retrieval evals.
 -   **_Inference Analysis_** - Visualize inferences and embeddings using dimensionality reduction and clustering to identify drift and performance degradation.
 

{arize_phoenix-4.5.0 → arize_phoenix-4.6.2}/README.md

@@ -29,7 +29,7 @@
 
 Phoenix is an open-source AI observability platform designed for experimentation, evaluation, and troubleshooting. It provides:
 
--   **_Tracing_** - Trace your LLM application's runtime using using OpenTelemetry-based instrumentation.
+-   **_Tracing_** - Trace your LLM application's runtime using OpenTelemetry-based instrumentation.
 -   **_Evaluation_** - Leverage LLMs to benchmark your application's performance using response and retrieval evals.
 -   **_Inference Analysis_** - Visualize inferences and embeddings using dimensionality reduction and clustering to identify drift and performance degradation.
 

{arize_phoenix-4.5.0 → arize_phoenix-4.6.2}/examples/manually-instrumented-chatbot/frontend/requirements.txt

@@ -87,7 +87,7 @@ referencing==0.34.0
     # via
     #   jsonschema
     #   jsonschema-specifications
-requests==2.31.0
+requests==2.32.2
     # via streamlit
 rich==13.7.1
     # via streamlit
@@ -110,7 +110,7 @@ toml==0.10.2
     # via streamlit
 toolz==0.12.1
     # via altair
-tornado==6.4
+tornado==6.4.1
     # via streamlit
 typing-extensions==4.11.0
     # via

{arize_phoenix-4.5.0 → arize_phoenix-4.6.2}/pyproject.toml

@@ -24,7 +24,7 @@ classifiers = [
 dependencies = [
   "scikit-learn",
   "numpy<2",  # https://github.com/scikit-learn-contrib/hdbscan/issues/642
-  "pandas",
+  "pandas>=1.0",
   "jinja2",
   "umap-learn",
   "hdbscan>=0.8.33",
@@ -41,12 +41,12 @@ dependencies = [
   "protobuf>=3.20, <6.0",
   "grpcio",
   "tqdm",
-  "requests",
+  "httpx",
   "opentelemetry-sdk",
   "opentelemetry-proto>=1.12.0",  # needed to avoid this issue: https://github.com/Arize-ai/phoenix/issues/2695
   "opentelemetry-exporter-otlp",
   "opentelemetry-semantic-conventions",
-  "openinference-semantic-conventions>=0.1.5",
+  "openinference-semantic-conventions>=0.1.9",
   "openinference-instrumentation",
   "openinference-instrumentation-langchain>=0.1.12",
   "openinference-instrumentation-llama-index>=1.2.0",
@@ -57,6 +57,9 @@ dependencies = [
   "aioitertools",
   "sqlean.py>=3.45.1",
   "cachetools",
+  "python-multipart",  # see https://www.starlette.io/#dependencies
+  "arize-phoenix-evals>=0.13.1",
+  "pyyaml", # for OpenAPI
 ]
 dynamic = ["version"]
 
@@ -67,6 +70,9 @@ dev = [
   "jupyter",
   "nbqa",
   "ruff==0.4.9",
+  "pandas>=1.0",
+  "tabulate",  # used by DataFrame.to_markdown()
+  "types-tabulate",
   "pandas-stubs==2.2.2.240603; python_version>='3.9'",
   "pandas-stubs==2.0.3.230814; python_version<'3.9'",
   "pytest==8.2.2",
@@ -85,14 +91,14 @@ dev = [
   "anthropic",
   "prometheus_client",
 ]
-evals = [
-  "arize-phoenix-evals>=0.3.0",
-]
-experimental = [
-  "tenacity",
-]
+evals = []
+experimental = []
 llama-index = [
-  "llama-index==0.10.44",  # always pin to a version that keeps our notebooks working
+  "llama-index==0.10.51",  # always pin to a version that keeps our notebooks working
+  "llama-index-readers-file==0.1.25",
+  "llama-index-llms-openai==0.1.24",
+  "llama-index-embeddings-openai==0.1.10",
+  "llama-index-agent-openai==0.2.7",
 ]
 pg = [
   "asyncpg",
@@ -150,7 +156,6 @@ dependencies = [
   "litellm>=1.0.3",
   "openai>=1.0.0",
   "tenacity",
-  "requests",
   "protobuf==3.20",  # version minimum (for tests)
   "responses",
   "tiktoken",
@@ -166,11 +171,11 @@ dependencies = [
 dependencies = [
   "mypy==1.10.0",
   "tenacity",
-  "pandas-stubs==2.2.2.240603; python_version>='3.9'",
-  "pandas-stubs==2.0.3.230814; python_version<'3.9'",
+  "pandas>=1.0",
+  "pandas-stubs==2.0.3.230814",
+  "types-tabulate",
   "types-psutil",
   "types-tqdm",
-  "types-requests",
   "types-protobuf",
   "types-setuptools",
   "types-cachetools",
@@ -187,14 +192,21 @@ dependencies = [
   "opentelemetry-instrumentation-grpc",
   "py-grpc-prometheus",
   "strawberry-graphql[opentelemetry]==0.235.0",  # need to pin version because we're monkey-patching
+  "requests",  # this is needed to type-check third-party packages
 ]
 
+[[tool.hatch.envs.type.matrix]]
+python = ["3.8", "3.9", "3.12"]
+
 [tool.hatch.envs.style]
 detached = true
 dependencies = [
   "ruff==0.4.9",
 ]
 
+[[tool.hatch.envs.style.matrix]]
+python = ["3.8", "3.9", "3.12"]
+
 [tool.hatch.envs.notebooks]
 detached = true
 dependencies = [
@@ -204,6 +216,7 @@ dependencies = [
 [tool.hatch.envs.docs]
 detached = true
 dependencies = [
+  "pyment",
   "interrogate",
 ]
 
@@ -292,6 +305,14 @@ dependencies = [
 [tool.hatch.envs.gql.scripts]
 build = 'strawberry export-schema phoenix.server.api.schema:schema > app/schema.graphql'
 
+[tool.hatch.envs.openapi]
+dependencies = [
+  "pyyaml",
+]
+
+[tool.hatch.envs.openapi.scripts]
+build = 'python -m phoenix.server.api.openapi.main > schemas/openapi.yaml'
+
 [tool.hatch.envs.proto]
 detached = true
 dependencies = [
@@ -326,6 +347,7 @@ disallow_untyped_defs = true
 disallow_incomplete_defs  = true
 strict = true
 exclude = [
+  "api_reference",
   "packages/",
   "src/phoenix/evals/",
   "dist/",
@@ -347,6 +369,7 @@ module = [
   "wrapt",
   "langchain.*",
   "litellm",
+  "litellm.*",
   "nest_asyncio",
   "opentelemetry.*",
   "pyarrow",

{arize_phoenix-4.5.0 → arize_phoenix-4.6.2}/src/phoenix/__init__.py

@@ -1,9 +1,3 @@
-import sys
-from importlib.abc import Loader, MetaPathFinder
-from importlib.machinery import ModuleSpec
-from types import ModuleType
-from typing import Any, Optional
-
 from .inferences.fixtures import ExampleInferences, load_example
 from .inferences.inferences import Inferences
 from .inferences.schema import EmbeddingColumnNames, RetrievalEmbeddingColumnNames, Schema
@@ -57,24 +51,3 @@ __all__ = [
     "Client",
     "evals",
 ]
-
-
-class PhoenixEvalsFinder(MetaPathFinder):
-    def find_spec(self, fullname: Any, path: Any, target: Any = None) -> Optional[ModuleSpec]:
-        if fullname == "phoenix.evals":
-            return ModuleSpec(fullname, PhoenixEvalsLoader())
-        return None
-
-
-class PhoenixEvalsLoader(Loader):
-    def create_module(self, spec: ModuleSpec) -> None:
-        return None
-
-    def exec_module(self, module: ModuleType) -> None:
-        raise ImportError(
-            "The optional `phoenix.evals` package is not installed. "
-            "Please install `phoenix` with the `evals` extra: `pip install 'arize-phoenix[evals]'`."
-        )
-
-
-sys.meta_path.append(PhoenixEvalsFinder())

{arize_phoenix-4.5.0 → arize_phoenix-4.6.2}/src/phoenix/config.py

@@ -2,7 +2,9 @@ import os
 import tempfile
 from logging import getLogger
 from pathlib import Path
-from typing import List, Optional
+from typing import Dict, List, Optional
+
+from .utilities.re import parse_env_headers
 
 logger = getLogger(__name__)
 
@@ -12,6 +14,12 @@ ENV_PHOENIX_GRPC_PORT = "PHOENIX_GRPC_PORT"
 ENV_PHOENIX_HOST = "PHOENIX_HOST"
 ENV_PHOENIX_HOST_ROOT_PATH = "PHOENIX_HOST_ROOT_PATH"
 ENV_NOTEBOOK_ENV = "PHOENIX_NOTEBOOK_ENV"
+ENV_PHOENIX_CLIENT_HEADERS = "PHOENIX_CLIENT_HEADERS"
+"""
+The headers to include in Phoenix client requests.
+Note: This overrides OTEL_EXPORTER_OTLP_HEADERS in the case where
+phoenix.trace instrumentors are used.
+"""
 ENV_PHOENIX_COLLECTOR_ENDPOINT = "PHOENIX_COLLECTOR_ENDPOINT"
 """
 The endpoint traces and evals are sent to. This must be set if the Phoenix
@@ -105,15 +113,15 @@ GRPC_PORT = 4317
 """The port the gRPC server will run on after launch_app is called.
 The default network port for OTLP/gRPC is 4317.
 See https://opentelemetry.io/docs/specs/otlp/#otlpgrpc-default-port"""
-GENERATED_DATASET_NAME_PREFIX = "phoenix_dataset_"
+GENERATED_INFERENCES_NAME_PREFIX = "phoenix_inferences_"
 """The prefix of datasets that are auto-assigned a name."""
 WORKING_DIR = get_working_dir()
-"""The work directory for saving, loading, and exporting datasets."""
+"""The work directory for saving, loading, and exporting data."""
 
 ROOT_DIR = WORKING_DIR
 EXPORT_DIR = ROOT_DIR / "exports"
-DATASET_DIR = ROOT_DIR / "datasets"
-TRACE_DATASET_DIR = ROOT_DIR / "trace_datasets"
+INFERENCES_DIR = ROOT_DIR / "inferences"
+TRACE_DATASETS_DIR = ROOT_DIR / "trace_datasets"
 
 
 def ensure_working_dir() -> None:
@@ -126,8 +134,8 @@ def ensure_working_dir() -> None:
         for path in (
             ROOT_DIR,
             EXPORT_DIR,
-            DATASET_DIR,
-            TRACE_DATASET_DIR,
+            INFERENCES_DIR,
+            TRACE_DATASETS_DIR,
         ):
             path.mkdir(parents=True, exist_ok=True)
     except Exception as e:
@@ -219,4 +227,31 @@ def get_env_enable_prometheus() -> bool:
     )
 
 
+def get_env_client_headers() -> Optional[Dict[str, str]]:
+    if headers_str := os.getenv(ENV_PHOENIX_CLIENT_HEADERS):
+        return parse_env_headers(headers_str)
+    return None
+
+
+def get_base_url() -> str:
+    host = get_env_host()
+    if host == "0.0.0.0":
+        host = "127.0.0.1"
+    base_url = get_env_collector_endpoint() or f"http://{host}:{get_env_port()}"
+    return base_url if base_url.endswith("/") else base_url + "/"
+
+
+def get_web_base_url() -> str:
+    """Return the web UI base URL.
+
+    Returns:
+        str: the web UI base URL
+    """
+    from phoenix.session.session import active_session
+
+    if session := active_session():
+        return session.url
+    return get_base_url()
+
+
 DEFAULT_PROJECT_NAME = "default"

{arize_phoenix-4.5.0 → arize_phoenix-4.6.2}/src/phoenix/core/model.py

@@ -7,73 +7,73 @@ from .embedding_dimension import EmbeddingDimension
 
 
 def _get_embedding_dimensions(
-    primary_dataset: Inferences, reference_dataset: Optional[Inferences]
+    primary_inferences: Inferences, reference_inferences: Optional[Inferences]
 ) -> List[EmbeddingDimension]:
     embedding_dimensions: List[EmbeddingDimension] = []
     embedding_features: EmbeddingFeatures = {}
 
     primary_embedding_features: Optional[EmbeddingFeatures] = (
-        primary_dataset.schema.embedding_feature_column_names
+        primary_inferences.schema.embedding_feature_column_names
     )
     if primary_embedding_features is not None:
         embedding_features.update(primary_embedding_features)
     primary_prompt_column_names: Optional[EmbeddingColumnNames] = (
-        primary_dataset.schema.prompt_column_names
+        primary_inferences.schema.prompt_column_names
     )
     if primary_prompt_column_names is not None:
         embedding_features.update({"prompt": primary_prompt_column_names})
     primary_response_column_names: Optional[Union[str, EmbeddingColumnNames]] = (
-        primary_dataset.schema.response_column_names
+        primary_inferences.schema.response_column_names
     )
     if isinstance(primary_response_column_names, EmbeddingColumnNames):
         embedding_features.update({"response": primary_response_column_names})
 
-    if reference_dataset is not None:
+    if reference_inferences is not None:
         reference_embedding_features: Optional[EmbeddingFeatures] = (
-            reference_dataset.schema.embedding_feature_column_names
+            reference_inferences.schema.embedding_feature_column_names
         )
         if reference_embedding_features is not None:
             embedding_features.update(reference_embedding_features)
         reference_prompt_column_names: Optional[EmbeddingColumnNames] = (
-            reference_dataset.schema.prompt_column_names
+            reference_inferences.schema.prompt_column_names
         )
         if reference_prompt_column_names is not None:
             embedding_features.update({"prompt": reference_prompt_column_names})
         reference_response_column_names: Optional[Union[str, EmbeddingColumnNames]] = (
-            reference_dataset.schema.response_column_names
+            reference_inferences.schema.response_column_names
         )
         if isinstance(reference_response_column_names, EmbeddingColumnNames):
             embedding_features.update({"response": reference_response_column_names})
 
     for embedding_feature, embedding_column_names in embedding_features.items():
         embedding_dimensions.append(EmbeddingDimension(name=embedding_feature))
-        if reference_dataset is not None:
-            _check_embedding_vector_lengths_match_across_datasets(
-                embedding_feature, embedding_column_names, primary_dataset, reference_dataset
+        if reference_inferences is not None:
+            _check_embedding_vector_lengths_match_across_inference_sets(
+                embedding_feature, embedding_column_names, primary_inferences, reference_inferences
             )
 
     return embedding_dimensions
 
 
-def _check_embedding_vector_lengths_match_across_datasets(
+def _check_embedding_vector_lengths_match_across_inference_sets(
     embedding_feature_name: str,
     embedding_column_names: EmbeddingColumnNames,
-    primary_dataset: Inferences,
-    reference_dataset: Inferences,
+    primary_inferences: Inferences,
+    reference_inferences: Inferences,
 ) -> None:
     """
     Ensure that for each embedding feature, the vector lengths match across the primary
-    and reference datasets which is required for calculating embedding drift (vector distance)
+    and reference inferences which is required for calculating embedding drift (vector distance)
     """
     primary_vector_length = _get_column_vector_length(
-        primary_dataset, embedding_column_names.vector_column_name
+        primary_inferences, embedding_column_names.vector_column_name
     )
     reference_vector_length = _get_column_vector_length(
-        reference_dataset, embedding_column_names.vector_column_name
+        reference_inferences, embedding_column_names.vector_column_name
     )
 
-    # if one of the datasets doesn't have the embedding column at all, which is fine since we
-    # just consider this as missing from one of the datasets and won't need to worry about
+    # if one of the inferences doesn't have the embedding column at all, which is fine since we
+    # just consider this as missing from one of the inferences and won't need to worry about
     # calculating drift
     if primary_vector_length is None or reference_vector_length is None:
         return
@@ -81,23 +81,23 @@ def _check_embedding_vector_lengths_match_across_datasets(
     if primary_vector_length != reference_vector_length:
         raise ValueError(
             f"Embedding vector length must match for "
-            f"both datasets; embedding_feature={embedding_feature_name} "
+            f"both inference sets; embedding_feature={embedding_feature_name} "
             f"vector_column={embedding_column_names.vector_column_name}"
         )
 
 
 def _get_column_vector_length(
-    dataset: Inferences, embedding_vector_column_name: str
+    inferences: Inferences, embedding_vector_column_name: str
 ) -> Optional[int]:
     """
-    Because a dataset has already been constructed, we can assume that the lengths
-    of the vectors for any given embedding feature in the dataset are the same.
+    Because a inferences has already been constructed, we can assume that the lengths
+    of the vectors for any given embedding feature in the inferences are the same.
     Returns the length a vector by getting the length first non-null vector.
     """
-    if embedding_vector_column_name not in dataset.dataframe:
+    if embedding_vector_column_name not in inferences.dataframe:
         return None
 
-    column = dataset.dataframe[embedding_vector_column_name]
+    column = inferences.dataframe[embedding_vector_column_name]
 
     for row in column:
         # None/NaN is a valid entry for a row and represents the fact that the