deps: Update google-cloud-aiplatform (#510)

Upgrading the latest google-cloud-aiplatform library made a breaking
change to the evaluation system. This PR addresses the library version
upgrade along with the updates to the evaluation system.

llm_demo/evaluation/eval_golden.py

@@ -39,7 +39,8 @@ class EvalData(BaseModel):
     category: Optional[str] = Field(default=None, description="Evaluation category")
     query: Optional[str] = Field(default=None, description="User query")
     instruction: Optional[str] = Field(
-        default=None, description="Instruction to llm system"
+        default="",
+        description="Part of the input user prompt. It refers to the inference instruction that is sent to you llm",
     )
     content: Optional[str] = Field(
         default=None,
@@ -48,16 +49,20 @@ class EvalData(BaseModel):
     tool_calls: List[ToolCall] = Field(
         default=[], description="Golden tool call for evaluation"
     )
+    prompt: Optional[str] = Field(
+        default="",
+        description="User input for the Gen AI model or application. It's optional in some cases.",
+    )
     context: Optional[List[Dict[str, Any] | List[Dict[str, Any]]]] = Field(
         default=None, description="Context given to llm in order to answer user query"
     )
     output: Optional[str] = Field(
         default=None, description="Golden output for evaluation"
     )
-    prediction_tool_calls: List[ToolCall] = Field(
+    llm_tool_calls: List[ToolCall] = Field(
         default=[], description="Tool call output from LLM"
     )
-    prediction_output: str = Field(default="", description="Final output from LLM")
+    llm_output: str = Field(default="", description="Final output from LLM")
     reset: bool = Field(
         default=True, description="Determine to reset the chat after invoke"
     )

llm_demo/evaluation/evaluation.py

@@ -18,19 +18,21 @@
 
 import pandas as pd
 from pydantic import BaseModel, Field
-from vertexai.preview.evaluation import EvalTask  # type: ignore
-from vertexai.preview.evaluation import _base as evaluation_base
+from vertexai.evaluation import EvalTask
+from vertexai.evaluation import _base as evaluation_base
 
 from orchestrator import BaseOrchestrator
 
 from .eval_golden import EvalData, ToolCall
+from .metrics import response_phase_metrics, retrieval_phase_metrics
 
 
 async def run_llm_for_eval(
     eval_list: List[EvalData], orc: BaseOrchestrator, session: Dict, session_id: str
 ) -> List[EvalData]:
     """
-    Generate prediction_tool_calls and prediction_output for golden dataset query.
+    Generate llm_tool_calls and llm_output for golden dataset query.
+    This function is only compatible with the langchain-tools orchestration.
     """
     agent = orc.get_user_session(session_id)
     for eval_data in eval_list:
@@ -39,23 +41,25 @@ async def run_llm_for_eval(
         except Exception as e:
             print(f"error invoking agent: {e}")
         else:
-            eval_data.prediction_output = query_response.get("output")
+            eval_data.llm_output = query_response.get("output")
 
-            # Retrieve prediction_tool_calls from query response
-            prediction_tool_calls = []
+            # Retrieve llm_tool_calls from query response
+            llm_tool_calls = []
             contexts = []
             for step in query_response.get("intermediate_steps"):
                 called_tool = step[0]
                 tool_call = ToolCall(
                     name=called_tool.tool,
                     arguments=called_tool.tool_input,
                 )
-                prediction_tool_calls.append(tool_call)
+                llm_tool_calls.append(tool_call)
                 context = step[-1]
                 contexts.append(context)
 
-            eval_data.prediction_tool_calls = prediction_tool_calls
+            eval_data.llm_tool_calls = llm_tool_calls
             eval_data.context = contexts
+            eval_data.prompt = PROMPT
+            eval_data.instruction = f"Answer user query based on context given. User query is {eval_data.query}."
 
         if eval_data.reset:
             orc.user_session_reset(session, session_id)
@@ -68,7 +72,6 @@ def evaluate_retrieval_phase(
     """
     Run evaluation for the ability of a model to select the right tool and arguments (retrieval phase).
     """
-    metrics = ["tool_call_quality"]
     # Prepare evaluation task input
     responses = []
     references = []
@@ -85,7 +88,7 @@ def evaluate_retrieval_phase(
             json.dumps(
                 {
                     "content": e.content,
-                    "tool_calls": [t.model_dump() for t in e.prediction_tool_calls],
+                    "tool_calls": [t.model_dump() for t in e.llm_tool_calls],
                 }
             )
         )
@@ -98,7 +101,7 @@ def evaluate_retrieval_phase(
     # Run evaluation
     eval_result = EvalTask(
         dataset=eval_dataset,
-        metrics=metrics,
+        metrics=retrieval_phase_metrics,
         experiment=experiment_name,
     ).evaluate()
     return eval_result
@@ -110,37 +113,32 @@ def evaluate_response_phase(
     """
     Run evaluation for the ability of a model to generate a response based on the context given (response phase).
     """
-    metrics = [
-        "text_generation_quality",
-        "text_generation_factuality",
-        "summarization_pointwise_reference_free",
-        "qa_pointwise_reference_free",
-    ]
     # Prepare evaluation task input
     instructions = []
     contexts = []
     responses = []
+    prompts = []
 
     for e in eval_datas:
-        instructions.append(
-            f"Answer user query based on context given. User query is {e.query}."
-        )
+        instructions.append(e.instruction)
         context_str = (
             [json.dumps(c) for c in e.context] if e.context else ["no data retrieved"]
         )
-        contexts.append(PROMPT + ", " + ", ".join(context_str))
-        responses.append(e.prediction_output or "")
+        prompts.append(e.prompt)
+        contexts.append(", ".join(context_str))
+        responses.append(e.llm_output or "")
     eval_dataset = pd.DataFrame(
         {
             "instruction": instructions,
+            "prompt": prompts,
             "context": contexts,
             "response": responses,
         }
     )
     # Run evaluation
     eval_result = EvalTask(
         dataset=eval_dataset,
-        metrics=metrics,
+        metrics=response_phase_metrics,
         experiment=experiment_name,
     ).evaluate()
     return eval_result

llm_demo/evaluation/metrics.py

@@ -0,0 +1,49 @@
+# Copyright 2024 Google LLC
+#
+# 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
+#
+#     https://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 vertexai.evaluation import MetricPromptTemplateExamples, PointwiseMetric
+
+text_quality_metric = PointwiseMetric(
+    metric="text_quality",
+    metric_prompt_template=MetricPromptTemplateExamples.get_prompt_template(
+        "text_quality"
+    ),
+)
+
+summarization_quality_metric = PointwiseMetric(
+    metric="summarization_quality",
+    metric_prompt_template=MetricPromptTemplateExamples.get_prompt_template(
+        "summarization_quality"
+    ),
+)
+
+question_answering_quality_metric = PointwiseMetric(
+    metric="question_answering_quality",
+    metric_prompt_template=MetricPromptTemplateExamples.get_prompt_template(
+        "question_answering_quality"
+    ),
+)
+
+response_phase_metrics = [
+    text_quality_metric,
+    summarization_quality_metric,
+    question_answering_quality_metric,
+]
+
+retrieval_phase_metrics = [
+    "tool_call_valid",
+    "tool_name_match",
+    "tool_parameter_key_match",
+    "tool_parameter_kv_match",
+]

llm_demo/pyproject.toml

@@ -3,4 +3,8 @@ profile = "black"
 
 [tool.mypy]
 python_version = 3.11
-warn_unused_configs = true
+warn_unused_configs = true
+
+[[tool.mypy.overrides]]
+module = ["vertexai.evaluation"]
+ignore_missing_imports = true

llm_demo/requirements.txt

@@ -1,6 +1,6 @@
 fastapi==0.109.2
 google-auth==2.33.0
-google-cloud-aiplatform[rapid_evaluation]==1.62.0
+google-cloud-aiplatform[evaluation]==1.72.0
 itsdangerous==2.2.0
 jinja2==3.1.4
 langchain-community==0.2.9