remove references to tool types

Author: keith-decker

util/opentelemetry-util-genai/src/opentelemetry/util/genai/api.py

@@ -23,9 +23,9 @@
 from opentelemetry.semconv.schemas import Schemas
 from opentelemetry.trace import get_tracer
 
-from .data import ChatGeneration, Error, Message, ToolFunction, ToolOutput
+from .data import ChatGeneration, Error, Message
 from .exporters import SpanMetricEventExporter, SpanMetricExporter
-from .types import LLMInvocation, ToolInvocation
+from .types import LLMInvocation
 from .version import __version__
 
 
@@ -80,20 +80,17 @@ def __init__(self, exporter_type_full: bool = True, **kwargs):
         )
 
         self._llm_registry: dict[UUID, LLMInvocation] = {}
-        self._tool_registry: dict[UUID, ToolInvocation] = {}
         self._lock = Lock()
 
     def start_llm(
         self,
         prompts: List[Message],
-        tool_functions: List[ToolFunction],
         run_id: UUID,
         parent_run_id: Optional[UUID] = None,
         **attributes,
     ):
         invocation = LLMInvocation(
             messages=prompts,
-            tool_functions=tool_functions,
             run_id=run_id,
             parent_run_id=parent_run_id,
             attributes=attributes,
@@ -122,47 +119,10 @@ def fail_llm(
         with self._lock:
             invocation = self._llm_registry.pop(run_id)
         invocation.end_time = time.time()
-        invocation.attributes.update(**attributes)
+        invocation.attributes.update(attributes)
         self._exporter.error_llm(error, invocation)
         return invocation
 
-    def start_tool(
-        self,
-        input_str: str,
-        run_id: UUID,
-        parent_run_id: Optional[UUID] = None,
-        **attributes,
-    ):
-        invocation = ToolInvocation(
-            input_str=input_str,
-            run_id=run_id,
-            parent_run_id=parent_run_id,
-            attributes=attributes,
-        )
-        with self._lock:
-            self._tool_registry[invocation.run_id] = invocation
-        self._exporter.init_tool(invocation)
-
-    def stop_tool(
-        self, run_id: UUID, output: ToolOutput, **attributes
-    ) -> ToolInvocation:
-        with self._lock:
-            invocation = self._tool_registry.pop(run_id)
-        invocation.end_time = time.time()
-        invocation.output = output
-        self._exporter.export_tool(invocation)
-        return invocation
-
-    def fail_tool(
-        self, run_id: UUID, error: Error, **attributes
-    ) -> ToolInvocation:
-        with self._lock:
-            invocation = self._tool_registry.pop(run_id)
-        invocation.end_time = time.time()
-        invocation.attributes.update(**attributes)
-        self._exporter.error_tool(error, invocation)
-        return invocation
-
 
 # Singleton accessor
 _default_client: TelemetryClient | None = None

util/opentelemetry-util-genai/src/opentelemetry/util/genai/data.py

@@ -1,26 +1,4 @@
-from dataclasses import dataclass, field
-from typing import List
-
-
-@dataclass
-class ToolOutput:
-    tool_call_id: str
-    content: str
-
-
-@dataclass
-class ToolFunction:
-    name: str
-    description: str
-    parameters: str
-
-
-@dataclass
-class ToolFunctionCall:
-    id: str
-    name: str
-    arguments: str
-    type: str
+from dataclasses import dataclass
 
 
 @dataclass
@@ -29,15 +7,13 @@ class Message:
     type: str
     name: str
     tool_call_id: str
-    tool_function_calls: List[ToolFunctionCall] = field(default_factory=list)
 
 
 @dataclass
 class ChatGeneration:
     content: str
     type: str
     finish_reason: str = None
-    tool_function_calls: List[ToolFunctionCall] = field(default_factory=list)
 
 
 @dataclass

util/opentelemetry-util-genai/src/opentelemetry/util/genai/evals.py

@@ -0,0 +1,89 @@
+# Copyright The OpenTelemetry Authors
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from abc import ABC, abstractmethod
+
+from .types import LLMInvocation
+
+
+class EvaluationResult:
+    """
+    Standardized result for any GenAI evaluation.
+    """
+
+    def __init__(self, score: float, details: dict = None):
+        self.score = score
+        self.details = details or {}
+
+
+class Evaluator(ABC):
+    """
+    Abstract base: any evaluation backend must implement.
+    """
+
+    @abstractmethod
+    def evaluate(self, invocation: LLMInvocation) -> EvaluationResult:
+        """
+        Evaluate a completed LLMInvocation and return a result.
+        """
+        pass
+
+
+class DeepEvalsEvaluator(Evaluator):
+    """
+    Uses DeepEvals library for LLM-as-judge evaluations.
+    """
+
+    def __init__(self, config: dict = None):
+        # e.g. load models, setup API keys
+        self.config = config or {}
+
+    def evaluate(self, invocation: LLMInvocation) -> EvaluationResult:
+        # stub: integrate with deepevals SDK
+        # result = deepevals.judge(invocation.prompt, invocation.response, **self.config)
+        score = 0.0  # placeholder
+        details = {"method": "deepevals"}
+        return EvaluationResult(score=score, details=details)
+
+
+class OpenLitEvaluator(Evaluator):
+    """
+    Uses OpenLit or similar OSS evaluation library.
+    """
+
+    def __init__(self, config: dict = None):
+        self.config = config or {}
+
+    def evaluate(self, invocation: LLMInvocation) -> EvaluationResult:
+        # stub: integrate with openlit SDK
+        score = 0.0  # placeholder
+        details = {"method": "openlit"}
+        return EvaluationResult(score=score, details=details)
+
+
+# Registry for easy lookup
+EVALUATORS = {
+    "deepevals": DeepEvalsEvaluator,
+    "openlit": OpenLitEvaluator,
+}
+
+
+def get_evaluator(name: str, config: dict = None) -> Evaluator:
+    """
+    Factory: return an evaluator by name.
+    """
+    cls = EVALUATORS.get(name.lower())
+    if not cls:
+        raise ValueError(f"Unknown evaluator: {name}")
+    return cls(config)