arize-phoenix 5.3.0__tar.gz → 5.4.0__tar.gz

{arize_phoenix-5.3.0 → arize_phoenix-5.4.0}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.3
 Name: arize-phoenix
-Version: 5.3.0
+Version: 5.4.0
 Summary: AI Observability and Evaluation
 Project-URL: Documentation, https://docs.arize.com/phoenix/
 Project-URL: Issues, https://github.com/Arize-ai/phoenix/issues
@@ -58,6 +58,7 @@ Requires-Dist: wrapt
 Provides-Extra: container
 Requires-Dist: fast-hdbscan>=0.2.0; extra == 'container'
 Requires-Dist: numba>=0.60.0; extra == 'container'
+Requires-Dist: openai>=1.0.0; extra == 'container'
 Requires-Dist: opentelemetry-exporter-otlp; extra == 'container'
 Requires-Dist: opentelemetry-instrumentation-fastapi; extra == 'container'
 Requires-Dist: opentelemetry-instrumentation-grpc; extra == 'container'

{arize_phoenix-5.3.0 → arize_phoenix-5.4.0}/pyproject.toml

@@ -124,6 +124,7 @@ pg = [
 ]
 container = [
   "prometheus-client",
+  "openai>=1.0.0",
   "opentelemetry-sdk",
   "opentelemetry-proto>=1.12.0",
   "opentelemetry-exporter-otlp",

{arize_phoenix-5.3.0 → arize_phoenix-5.4.0}/src/phoenix/server/api/subscriptions.py

@@ -45,6 +45,7 @@ from phoenix.server.api.input_types.ChatCompletionMessageInput import ChatComple
 from phoenix.server.api.input_types.InvocationParameters import InvocationParameters
 from phoenix.server.api.types.ChatCompletionMessageRole import ChatCompletionMessageRole
 from phoenix.server.api.types.GenerativeProvider import GenerativeProviderKey
+from phoenix.server.api.types.Span import Span, to_gql_span
 from phoenix.server.dml_event import SpanInsertEvent
 from phoenix.trace.attributes import unflatten
 from phoenix.utilities.json import jsonify
@@ -55,6 +56,7 @@ from phoenix.utilities.template_formatters import (
 )
 
 if TYPE_CHECKING:
+    from openai.types import CompletionUsage
     from openai.types.chat import (
         ChatCompletionMessageParam,
     )
@@ -93,8 +95,14 @@ class ToolCallChunk:
     function: FunctionCallChunk
 
 
-ChatCompletionChunk: TypeAlias = Annotated[
-    Union[TextChunk, ToolCallChunk], strawberry.union("ChatCompletionChunk")
+@strawberry.type
+class FinishedChatCompletion:
+    span: Span
+
+
+ChatCompletionSubscriptionPayload: TypeAlias = Annotated[
+    Union[TextChunk, ToolCallChunk, FinishedChatCompletion],
+    strawberry.union("ChatCompletionSubscriptionPayload"),
 ]
 
 
@@ -159,8 +167,9 @@ class Subscription:
     @strawberry.subscription
     async def chat_completion(
         self, info: Info[Context, None], input: ChatCompletionInput
-    ) -> AsyncIterator[ChatCompletionChunk]:
+    ) -> AsyncIterator[ChatCompletionSubscriptionPayload]:
         from openai import NOT_GIVEN, AsyncAzureOpenAI, AsyncOpenAI
+        from openai.types.chat import ChatCompletionStreamOptionsParam
 
         client: Union[AsyncAzureOpenAI, AsyncOpenAI]
 
@@ -208,14 +217,19 @@ class Subscription:
             text_chunks: List[TextChunk] = []
             tool_call_chunks: DefaultDict[ToolCallIndex, List[ToolCallChunk]] = defaultdict(list)
             role: Optional[str] = None
+            token_usage: Optional[CompletionUsage] = None
             async for chunk in await client.chat.completions.create(
                 messages=openai_messages,
                 model=input.model.name,
                 stream=True,
                 tools=input.tools or NOT_GIVEN,
+                stream_options=ChatCompletionStreamOptionsParam(include_usage=True),
                 **invocation_parameters,
             ):
                 response_chunks.append(chunk)
+                if (usage := chunk.usage) is not None:
+                    token_usage = usage
+                    continue
                 choice = chunk.choices[0]
                 delta = choice.delta
                 if role is None:
@@ -246,6 +260,7 @@ class Subscription:
                 dict(
                     chain(
                         _output_value_and_mime_type(response_chunks),
+                        _llm_token_counts(token_usage) if token_usage is not None else [],
                         _llm_output_messages(text_chunks, tool_call_chunks),
                     )
                 )
@@ -257,6 +272,8 @@ class Subscription:
         assert (attributes := finished_span.attributes) is not None
         start_time = _datetime(epoch_nanoseconds=finished_span.start_time)
         end_time = _datetime(epoch_nanoseconds=finished_span.end_time)
+        prompt_tokens = token_usage.prompt_tokens if token_usage is not None else 0
+        completion_tokens = token_usage.completion_tokens if token_usage is not None else 0
         trace_id = _hex(finished_span.context.trace_id)
         span_id = _hex(finished_span.context.span_id)
         status = finished_span.status
@@ -274,36 +291,35 @@ class Subscription:
                         description="Traces from prompt playground",
                     )
                 )
-            trace_rowid = await session.scalar(
-                insert(models.Trace)
-                .returning(models.Trace.id)
-                .values(
-                    project_rowid=playground_project_id,
-                    trace_id=trace_id,
-                    start_time=start_time,
-                    end_time=end_time,
-                )
+            playground_trace = models.Trace(
+                project_rowid=playground_project_id,
+                trace_id=trace_id,
+                start_time=start_time,
+                end_time=end_time,
             )
-            await session.execute(
-                insert(models.Span).values(
-                    trace_rowid=trace_rowid,
-                    span_id=span_id,
-                    parent_id=None,
-                    name=span_name,
-                    span_kind=LLM,
-                    start_time=start_time,
-                    end_time=end_time,
-                    attributes=unflatten(attributes.items()),
-                    events=finished_span.events,
-                    status_code=status.status_code.name,
-                    status_message=status.description or "",
-                    cumulative_error_count=int(not status.is_ok),
-                    cumulative_llm_token_count_prompt=0,
-                    cumulative_llm_token_count_completion=0,
-                    llm_token_count_prompt=0,
-                    llm_token_count_completion=0,
-                )
+            playground_span = models.Span(
+                trace_rowid=playground_trace.id,
+                span_id=span_id,
+                parent_id=None,
+                name=span_name,
+                span_kind=LLM,
+                start_time=start_time,
+                end_time=end_time,
+                attributes=unflatten(attributes.items()),
+                events=finished_span.events,
+                status_code=status.status_code.name,
+                status_message=status.description or "",
+                cumulative_error_count=int(not status.is_ok),
+                cumulative_llm_token_count_prompt=prompt_tokens,
+                cumulative_llm_token_count_completion=completion_tokens,
+                llm_token_count_prompt=prompt_tokens,
+                llm_token_count_completion=completion_tokens,
+                trace=playground_trace,
             )
+            session.add(playground_trace)
+            session.add(playground_span)
+            await session.flush()
+            yield FinishedChatCompletion(span=to_gql_span(playground_span))
         info.context.event_queue.put(SpanInsertEvent(ids=(playground_project_id,)))
 
 
@@ -324,6 +340,12 @@ def _llm_tools(tools: List[JSONScalarType]) -> Iterator[Tuple[str, Any]]:
         yield f"{LLM_TOOLS}.{tool_index}.{TOOL_JSON_SCHEMA}", json.dumps(tool)
 
 
+def _llm_token_counts(usage: "CompletionUsage") -> Iterator[Tuple[str, Any]]:
+    yield LLM_TOKEN_COUNT_PROMPT, usage.prompt_tokens
+    yield LLM_TOKEN_COUNT_COMPLETION, usage.completion_tokens
+    yield LLM_TOKEN_COUNT_TOTAL, usage.total_tokens
+
+
 def _input_value_and_mime_type(input: ChatCompletionInput) -> Iterator[Tuple[str, Any]]:
     assert any(field.name == (api_key := "api_key") for field in fields(ChatCompletionInput))
     yield INPUT_MIME_TYPE, JSON
@@ -419,6 +441,9 @@ LLM_OUTPUT_MESSAGES = SpanAttributes.LLM_OUTPUT_MESSAGES
 LLM_MODEL_NAME = SpanAttributes.LLM_MODEL_NAME
 LLM_INVOCATION_PARAMETERS = SpanAttributes.LLM_INVOCATION_PARAMETERS
 LLM_TOOLS = SpanAttributes.LLM_TOOLS
+LLM_TOKEN_COUNT_PROMPT = SpanAttributes.LLM_TOKEN_COUNT_PROMPT
+LLM_TOKEN_COUNT_COMPLETION = SpanAttributes.LLM_TOKEN_COUNT_COMPLETION
+LLM_TOKEN_COUNT_TOTAL = SpanAttributes.LLM_TOKEN_COUNT_TOTAL
 
 MESSAGE_CONTENT = MessageAttributes.MESSAGE_CONTENT
 MESSAGE_ROLE = MessageAttributes.MESSAGE_ROLE

arize_phoenix-5.4.0/src/phoenix/server/static/.vite/manifest.json

@@ -0,0 +1,78 @@
+{
+  "_components-8zh9kCOG.js": {
+    "file": "assets/components-8zh9kCOG.js",
+    "name": "components",
+    "imports": [
+      "_vendor-6IcPAw_j.js",
+      "_vendor-arizeai-uC2sozJ6.js",
+      "_pages-CFNU-U_5.js",
+      "_vendor-three-DwGkEfCM.js",
+      "_vendor-codemirror-DVE2_WBr.js"
+    ]
+  },
+  "_pages-CFNU-U_5.js": {
+    "file": "assets/pages-CFNU-U_5.js",
+    "name": "pages",
+    "imports": [
+      "_vendor-6IcPAw_j.js",
+      "_vendor-arizeai-uC2sozJ6.js",
+      "_components-8zh9kCOG.js",
+      "_vendor-recharts-DwrexFA4.js",
+      "_vendor-codemirror-DVE2_WBr.js"
+    ]
+  },
+  "_vendor-!~{003}~.js": {
+    "file": "assets/vendor-DxkFTwjz.css",
+    "src": "_vendor-!~{003}~.js"
+  },
+  "_vendor-6IcPAw_j.js": {
+    "file": "assets/vendor-6IcPAw_j.js",
+    "name": "vendor",
+    "imports": [
+      "_vendor-three-DwGkEfCM.js"
+    ],
+    "css": [
+      "assets/vendor-DxkFTwjz.css"
+    ]
+  },
+  "_vendor-arizeai-uC2sozJ6.js": {
+    "file": "assets/vendor-arizeai-uC2sozJ6.js",
+    "name": "vendor-arizeai",
+    "imports": [
+      "_vendor-6IcPAw_j.js"
+    ]
+  },
+  "_vendor-codemirror-DVE2_WBr.js": {
+    "file": "assets/vendor-codemirror-DVE2_WBr.js",
+    "name": "vendor-codemirror",
+    "imports": [
+      "_vendor-6IcPAw_j.js"
+    ]
+  },
+  "_vendor-recharts-DwrexFA4.js": {
+    "file": "assets/vendor-recharts-DwrexFA4.js",
+    "name": "vendor-recharts",
+    "imports": [
+      "_vendor-6IcPAw_j.js"
+    ]
+  },
+  "_vendor-three-DwGkEfCM.js": {
+    "file": "assets/vendor-three-DwGkEfCM.js",
+    "name": "vendor-three"
+  },
+  "index.tsx": {
+    "file": "assets/index-Ci93KI-L.js",
+    "name": "index",
+    "src": "index.tsx",
+    "isEntry": true,
+    "imports": [
+      "_vendor-6IcPAw_j.js",
+      "_vendor-arizeai-uC2sozJ6.js",
+      "_pages-CFNU-U_5.js",
+      "_components-8zh9kCOG.js",
+      "_vendor-three-DwGkEfCM.js",
+      "_vendor-recharts-DwrexFA4.js",
+      "_vendor-codemirror-DVE2_WBr.js"
+    ]
+  }
+}