arize-phoenix 3.0.3__tar.gz → 3.1.0__tar.gz

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: arize-phoenix
-Version: 3.0.3
+Version: 3.1.0
 Summary: ML Observability in your notebook
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues
@@ -51,7 +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<0.10.0; extra == 'dev'
+Requires-Dist: llama-index>=0.10.3; extra == 'dev'
 Requires-Dist: nbqa; extra == 'dev'
 Requires-Dist: pandas-stubs<=2.0.2.230605; extra == 'dev'
 Requires-Dist: pre-commit; extra == 'dev'
@@ -64,8 +64,8 @@ Requires-Dist: strawberry-graphql[debug-server]==0.208.2; extra == 'dev'
 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'
+Requires-Dist: llama-index==0.10.3; extra == 'llama-index'
+Requires-Dist: openinference-instrumentation-llama-index>=1.0.0; extra == 'llama-index'
 Description-Content-Type: text/markdown
 
 <p align="center">

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/pyproject.toml

@@ -65,7 +65,7 @@ dev = [
   "strawberry-graphql[debug-server]==0.208.2",
   "pre-commit",
   "arize[AutoEmbeddings, LLM_Evaluation]",
-  "llama-index<0.10.0",
+  "llama-index>=0.10.3",
   "langchain>=0.0.334",
   "litellm>=1.0.3",
   "google-cloud-aiplatform>=1.3",
@@ -75,8 +75,8 @@ experimental = [
   "tenacity",
 ]
 llama-index = [
-  "llama-index==0.9.45",  # always pin to a version that keeps our notebooks working
-  "openinference-instrumentation-llama-index==0.1.3",
+  "llama-index==0.10.3",  # always pin to a version that keeps our notebooks working
+  "openinference-instrumentation-llama-index>=1.0.0",
 ]
 
 [project.urls]
@@ -111,7 +111,7 @@ dependencies = [
   "arize",
   "langchain>=0.0.334",
   "litellm>=1.0.3",
-  "llama-index<0.10.0",
+  "llama-index>=0.10.3",
   "openai>=1.0.0",
   "tenacity",
   "nltk==3.8.1",

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/src/phoenix/experimental/evals/models/litellm.py

@@ -1,4 +1,5 @@
 import logging
+import warnings
 from dataclasses import dataclass, field
 from typing import TYPE_CHECKING, Any, Dict, List, Optional
 
@@ -12,7 +13,7 @@ logger = logging.getLogger(__name__)
 
 @dataclass
 class LiteLLMModel(BaseEvalModel):
-    model_name: str = "gpt-3.5-turbo"
+    model: str = "gpt-3.5-turbo"
     """The model name to use."""
     temperature: float = 0.0
     """What sampling temperature to use."""
@@ -34,22 +35,42 @@ class LiteLLMModel(BaseEvalModel):
     max_content_size: Optional[int] = None
     """If you're using a fine-tuned model, set this to the maximum content size"""
 
+    # Deprecated fields
+    model_name: Optional[str] = None
+    """
+    .. deprecated:: 3.0.0
+       use `model` instead. This will be removed in a future release.
+    """
+
     def __post_init__(self) -> None:
+        self._migrate_model_name()
         self._init_environment()
         self._init_model_encoding()
 
+    def _migrate_model_name(self) -> None:
+        if self.model_name is not None:
+            warning_message = "The `model_name` field is deprecated. Use `model` instead. \
+                This will be removed in a future release."
+            warnings.warn(
+                warning_message,
+                DeprecationWarning,
+            )
+            print(warning_message)
+            self.model = self.model_name
+            self.model_name = None
+
     def _init_environment(self) -> None:
         try:
             import litellm
             from litellm import validate_environment
 
             self._litellm = litellm
-            env_info = validate_environment(self._litellm.utils.get_llm_provider(self.model_name))
+            env_info = validate_environment(self._litellm.utils.get_llm_provider(self.model))
 
             if not env_info["keys_in_environment"]:
                 raise RuntimeError(
                     f"Missing environment variable(s): '{str(env_info['missing_keys'])}', for "
-                    f"model: {self.model_name}. \nFor additional information about the right "
+                    f"model: {self.model}. \nFor additional information about the right "
                     "environment variables for specific model providers:\n"
                     "https://docs.litellm.ai/docs/completion/input#provider-specific-params."
                 )
@@ -67,14 +88,14 @@ class LiteLLMModel(BaseEvalModel):
 
     @property
     def max_context_size(self) -> int:
-        context_size = self.max_content_size or self._litellm.get_max_tokens(self.model_name).get(
+        context_size = self.max_content_size or self._litellm.get_max_tokens(self.model).get(
             "max_tokens", None
         )
 
         if context_size is None:
             raise ValueError(
-                "Can't determine maximum context size. An unknown model name was "
-                + f"used: {self.model_name}."
+                "Can't determine maximum context size. An unknown model was "
+                + f"used: {self.model}."
             )
 
         return context_size
@@ -84,11 +105,11 @@ class LiteLLMModel(BaseEvalModel):
         raise NotImplementedError
 
     def get_tokens_from_text(self, text: str) -> List[int]:
-        result: List[int] = self._encoding(model=self.model_name, text=text)
+        result: List[int] = self._encoding(model=self.model, text=text)
         return result
 
     def get_text_from_tokens(self, tokens: List[int]) -> str:
-        return str(self._decoding(model=self.model_name, tokens=tokens))
+        return str(self._decoding(model=self.model, tokens=tokens))
 
     async def _async_generate(self, prompt: str, **kwargs: Dict[str, Any]) -> str:
         return self._generate(prompt, **kwargs)
@@ -96,7 +117,7 @@ class LiteLLMModel(BaseEvalModel):
     def _generate(self, prompt: str, **kwargs: Dict[str, Any]) -> str:
         messages = self._get_messages_from_prompt(prompt)
         response = self._litellm.completion(
-            model=self.model_name,
+            model=self.model,
             messages=messages,
             temperature=self.temperature,
             max_tokens=self.max_tokens,

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/src/phoenix/experimental/evals/models/openai.py

@@ -1,5 +1,6 @@
 import logging
 import os
+import warnings
 from dataclasses import dataclass, field, fields
 from typing import (
     TYPE_CHECKING,
@@ -64,8 +65,10 @@ class OpenAIModel(BaseEvalModel):
     An optional base URL to use for the OpenAI API. If not provided, will default
     to what's configured in OpenAI
     """
-    model_name: str = "gpt-4"
-    """Model name to use. In of azure, this is the deployment name such as gpt-35-instant"""
+    model: str = "gpt-4"
+    """
+    Model name to use. In of azure, this is the deployment name such as gpt-35-instant
+    """
     temperature: float = 0.0
     """What sampling temperature to use."""
     max_tokens: int = 256
@@ -106,7 +109,15 @@ class OpenAIModel(BaseEvalModel):
     azure_ad_token: Optional[str] = field(default=None)
     azure_ad_token_provider: Optional[Callable[[], str]] = field(default=None)
 
+    # Deprecated fields
+    model_name: Optional[str] = field(default=None)
+    """
+    .. deprecated:: 3.0.0
+       use `model` instead. This will be removed
+    """
+
     def __post_init__(self) -> None:
+        self._migrate_model_name()
         self._init_environment()
         self._init_open_ai()
         self._init_tiktoken()
@@ -115,6 +126,17 @@ class OpenAIModel(BaseEvalModel):
     def reload_client(self) -> None:
         self._init_open_ai()
 
+    def _migrate_model_name(self) -> None:
+        if self.model_name:
+            warning_message = "The `model_name` field is deprecated. Use `model` instead. \
+                This will be removed in a future release."
+            print(
+                warning_message,
+            )
+            warnings.warn(warning_message, DeprecationWarning)
+            self.model = self.model_name
+            self.model_name = None
+
     def _init_environment(self) -> None:
         try:
             import openai
@@ -141,9 +163,7 @@ class OpenAIModel(BaseEvalModel):
         # For Azure, you need to provide the endpoint and the endpoint
         self._is_azure = bool(self.azure_endpoint)
 
-        self._model_uses_legacy_completion_api = self.model_name.startswith(
-            LEGACY_COMPLETION_API_MODELS
-        )
+        self._model_uses_legacy_completion_api = self.model.startswith(LEGACY_COMPLETION_API_MODELS)
         if self.api_key is None:
             api_key = os.getenv(OPENAI_API_KEY_ENVVAR_NAME)
             if api_key is None:
@@ -203,7 +223,7 @@ class OpenAIModel(BaseEvalModel):
 
     def _init_tiktoken(self) -> None:
         try:
-            encoding = self._tiktoken.encoding_for_model(self.model_name)
+            encoding = self._tiktoken.encoding_for_model(self.model)
         except KeyError:
             encoding = self._tiktoken.get_encoding("cl100k_base")
         self._tiktoken_encoding = encoding
@@ -333,20 +353,20 @@ class OpenAIModel(BaseEvalModel):
 
     @property
     def max_context_size(self) -> int:
-        model_name = self.model_name
+        model = self.model
         # handling finetuned models
-        if "ft-" in model_name:
-            model_name = self.model_name.split(":")[0]
-        if model_name == "gpt-4":
+        if "ft-" in model:
+            model = self.model.split(":")[0]
+        if model == "gpt-4":
             # Map gpt-4 to the current default
-            model_name = "gpt-4-0613"
+            model = "gpt-4-0613"
 
-        context_size = MODEL_TOKEN_LIMIT_MAPPING.get(model_name, None)
+        context_size = MODEL_TOKEN_LIMIT_MAPPING.get(model, None)
 
         if context_size is None:
             raise ValueError(
                 "Can't determine maximum context size. An unknown model name was "
-                f"used: {model_name}. Please provide a valid OpenAI model name. "
+                f"used: {model}. Please provide a valid OpenAI model name. "
                 "Known models are: " + ", ".join(MODEL_TOKEN_LIMIT_MAPPING.keys())
             )
 
@@ -355,7 +375,7 @@ class OpenAIModel(BaseEvalModel):
     @property
     def public_invocation_params(self) -> Dict[str, Any]:
         return {
-            **({"model": self.model_name}),
+            **({"model": self.model}),
             **self._default_params,
             **self.model_kwargs,
         }
@@ -388,8 +408,8 @@ class OpenAIModel(BaseEvalModel):
 
         Official documentation: https://github.com/openai/openai-cookbook/blob/main/examples/How_to_format_inputs_to_ChatGPT_models.ipynb
         """  # noqa
-        model_name = self.model_name
-        if model_name == "gpt-3.5-turbo-0301":
+        model = self.model
+        if model == "gpt-3.5-turbo-0301":
             tokens_per_message = 4  # every message follows <|start|>{role/name}\n{content}<|end|>\n
             tokens_per_name = -1  # if there's a name, the role is omitted
         else:

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/src/phoenix/experimental/evals/models/vertexai.py

@@ -1,3 +1,5 @@
+import logging
+import warnings
 from dataclasses import dataclass
 from typing import TYPE_CHECKING, Any, Dict, List, Optional
 
@@ -6,6 +8,7 @@ from phoenix.experimental.evals.models.base import BaseEvalModel
 if TYPE_CHECKING:
     from google.auth.credentials import Credentials  # type:ignore
 
+logger = logging.getLogger(__name__)
 
 MINIMUM_VERTEX_AI_VERSION = "1.33.0"
 
@@ -18,9 +21,9 @@ class VertexAIModel(BaseEvalModel):
     "location (str): The default location to use when making API calls. If not "
     "set defaults to us-central-1."
     credentials: Optional["Credentials"] = None
-    model_name: str = "text-bison"
-    tuned_model_name: Optional[str] = None
-    "The name of a tuned model. If provided, model_name is ignored."
+    model: str = "text-bison"
+    tuned_model: Optional[str] = None
+    "The name of a tuned model. If provided, model is ignored."
     max_retries: int = 6
     """Maximum number of retries to make when generating."""
     retry_min_seconds: int = 10
@@ -40,11 +43,50 @@ class VertexAIModel(BaseEvalModel):
     "How the model selects tokens for output, the next token is selected from "
     "among the top-k most probable tokens. Top-k is ignored for Codey models."
 
+    # Deprecated fields
+    model_name: Optional[str] = None
+    """
+    .. deprecated:: 3.0.0
+       use `model` instead. This will be removed in a future release.
+    """
+    tuned_model_name: Optional[str] = None
+    """
+    .. deprecated:: 3.0.0
+       use `tuned_model` instead. This will be removed in a future release.
+    """
+
     def __post_init__(self) -> None:
+        self._migrate_model_name()
         self._init_environment()
         self._init_vertex_ai()
         self._instantiate_model()
 
+    def _migrate_model_name(self) -> None:
+        if self.model_name is not None:
+            warning_message = (
+                "The `model_name` field is deprecated. Use `model` instead. "
+                + "This will be removed in a future release."
+            )
+            warnings.warn(
+                warning_message,
+                DeprecationWarning,
+            )
+            print(warning_message)
+            self.model = self.model_name
+            self.model_name = None
+        if self.tuned_model_name is not None:
+            warning_message = (
+                "`tuned_model_name` field is deprecated. Use `tuned_model` instead. "
+                + "This will be removed in a future release."
+            )
+            warnings.warn(
+                warning_message,
+                DeprecationWarning,
+            )
+            print(warning_message)
+            self.tuned_model = self.tuned_model_name
+            self.tuned_model_name = None
+
     def _init_environment(self) -> None:
         try:
             import google.api_core.exceptions as google_exceptions  # type:ignore
@@ -72,10 +114,10 @@ class VertexAIModel(BaseEvalModel):
 
             model = TextGenerationModel
 
-        if self.tuned_model_name:
-            self._model = model.get_tuned_model(self.tuned_model_name)
+        if self.tuned_model:
+            self._model = model.get_tuned_model(self.tuned_model)
         else:
-            self._model = model.from_pretrained(self.model_name)
+            self._model = model.from_pretrained(self.model)
 
     def verbose_generation_info(self) -> str:
         return f"VertexAI invocation parameters: {self.invocation_params}"
@@ -93,7 +135,7 @@ class VertexAIModel(BaseEvalModel):
 
     @property
     def is_codey_model(self) -> bool:
-        return is_codey_model(self.tuned_model_name or self.model_name)
+        return is_codey_model(self.tuned_model or self.model)
 
     @property
     def _init_params(self) -> Dict[str, Any]:

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/src/phoenix/session/session.py

@@ -477,6 +477,7 @@ def launch_app(
             f"port {port} is not occupied by another process) or file an issue "
             f"with us at https://github.com/Arize-ai/phoenix"
         )
+        _session = None
         return None
 
     print(f"🌍 To view the Phoenix app in your browser, visit {_session.url}")
@@ -489,7 +490,9 @@ def active_session() -> Optional[Session]:
     """
     Returns the active session if one exists, otherwise returns None
     """
-    return _session
+    if _session and _session.active:
+        return _session
+    return None
 
 
 def close_app() -> None:

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/src/phoenix/trace/dsl/filter.py

@@ -166,8 +166,14 @@ class _Translator(ast.NodeTransformer):
         # In Python 3.9+, we can use `ast.unparse(node)` (no need for `source`).
         self._source = source
 
+    def visit_Subscript(self, node: ast.Subscript) -> Any:
+        if _is_metadata(node) and (key := _get_subscript_key(node)):
+            return _ast_metadata_subscript(key)
+        source_segment: str = cast(str, ast.get_source_segment(self._source, node))
+        raise SyntaxError(f"invalid expression: {source_segment}")  # TODO: add details
+
     def visit_Attribute(self, node: ast.Attribute) -> Any:
-        if _is_eval(node.value) and (eval_name := _get_eval_name(node.value)):
+        if _is_eval(node.value) and (eval_name := _get_subscript_key(node.value)):
             # e.g. `evals["name"].score`
             return _ast_evaluation_result_value(eval_name, node.attr)
         source_segment: str = cast(str, ast.get_source_segment(self._source, node))
@@ -209,9 +215,11 @@ def _validate_expression(
         if i == 0:
             if isinstance(node, (ast.BoolOp, ast.Compare)):
                 continue
+        elif _is_metadata(node):
+            continue
         elif _is_eval(node):
             # e.g. `evals["name"]`
-            if not (eval_name := _get_eval_name(node)) or (
+            if not (eval_name := _get_subscript_key(node)) or (
                 valid_eval_names is not None and eval_name not in valid_eval_names
             ):
                 source_segment = cast(str, ast.get_source_segment(source, node))
@@ -296,6 +304,19 @@ def _ast_evaluation_result_value(name: str, attr: str) -> ast.expr:
     return ast.parse(source, mode="eval").body
 
 
+def _ast_metadata_subscript(key: str) -> ast.expr:
+    source = (
+        f"_MISSING if ("
+        f"    _MD := span.attributes.get('metadata')"
+        f") is None else ("
+        f"    _MISSING if not hasattr(_MD, 'get') or ("
+        f"        _VALUE := _MD.get('{key}')"
+        f"    ) is None else _VALUE"
+        f")"
+    )
+    return ast.parse(source, mode="eval").body
+
+
 def _is_eval(node: Any) -> TypeGuard[ast.Subscript]:
     # e.g. `evals["name"]`
     return (
@@ -305,7 +326,16 @@ def _is_eval(node: Any) -> TypeGuard[ast.Subscript]:
     )
 
 
-def _get_eval_name(node: ast.Subscript) -> Optional[str]:
+def _is_metadata(node: Any) -> TypeGuard[ast.Subscript]:
+    # e.g. `metadata["name"]`
+    return (
+        isinstance(node, ast.Subscript)
+        and isinstance(value := node.value, ast.Name)
+        and value.id == "metadata"
+    )
+
+
+def _get_subscript_key(node: ast.Subscript) -> Optional[str]:
     if sys.version_info < (3, 9):
         # Note that `ast.Index` is deprecated in Python 3.9+, but is necessary
         # for Python 3.8 as part of `ast.Subscript`.

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/src/phoenix/trace/fixtures.py

@@ -137,8 +137,6 @@ def _download_traces_fixture(
 def load_example_traces(use_case: str) -> TraceDataset:
     """
     Loads a trace dataframe by name.
-
-    NB: this functionality is under active construction.
     """
     fixture = _get_trace_fixture_by_name(use_case)
     return TraceDataset(json_lines_to_df(_download_traces_fixture(fixture)))

{arize_phoenix-3.0.3 → arize_phoenix-3.1.0}/src/phoenix/trace/trace_dataset.py

@@ -101,9 +101,12 @@ class TraceDataset:
     """
 
     name: str
+    """
+    A human readable name for the dataset.
+    """
     dataframe: pd.DataFrame
     evaluations: List[Evaluations] = []
-    _id: UUID = uuid4()
+    _id: UUID
     _data_file_name: str = "data.parquet"
 
     def __init__(
@@ -131,8 +134,10 @@ class TraceDataset:
             raise ValueError(
                 f"The dataframe is missing some required columns: {', '.join(missing_columns)}"
             )
+        self._id = uuid4()
         self.dataframe = normalize_dataframe(dataframe)
-        self.name = name or f"{GENERATED_DATASET_NAME_PREFIX}{str(uuid4())}"
+        # TODO: This is not used in any meaningful way. Should remove
+        self.name = name or f"{GENERATED_DATASET_NAME_PREFIX}{str(self._id)}"
         self.evaluations = list(evaluations)
 
     @classmethod
@@ -249,6 +254,8 @@ class TraceDataset:
             }
         )
         parquet.write_table(table, path)
+        print(f"💾 Trace dataset saved to under ID: {self._id}")
+        print(f"📂 Trace dataset path: {path}")
         return self._id
 
     @classmethod
@@ -288,7 +295,7 @@ class TraceDataset:
                 warn(f'Failed to load evaluations with id: "{eval_id}"')
         table = parquet.read_table(path)
         dataframe = table.to_pandas()
-        ds = cls(dataframe, dataset_name, evaluations)
+        ds = cls(dataframe=dataframe, name=dataset_name, evaluations=evaluations)
         ds._id = dataset_id
         return ds
 

arize_phoenix-3.1.0/src/phoenix/version.py

@@ -0,0 +1 @@
+__version__ = "3.1.0"

arize_phoenix-3.0.3/src/phoenix/version.py

@@ -1 +0,0 @@
-__version__ = "3.0.3"