Make the schema compatible with azure's gpt

Author: eric-mccann-pro

python/LlmSpeechSummarization/llm_speech_summarization_component/schema.py

@@ -24,40 +24,43 @@
 # limitations under the License.                                            #
 #############################################################################
 
-from pydantic import BaseModel, Field
+from pydantic import BaseModel, Field, ConfigDict
 from typing import List
 
 class EntitiesObject(BaseModel):
     """
     An object containing lists of entities of interest that are present in your input.
 
     COMBINATION_INSTRUCTION: In your output, lists of strings (entities, topics, etc.) should each be the union of the corresponding fields in your input objects.
+    COMBINATION_INSTRUCTION: when combining multiple EntitiesObject objects, treat each field on it as a set of unique strings")
     """
+    model_config = ConfigDict(extra='forbid')
     names_of_people: List[str] = Field(
-        default=[],
         title="Unique names of people referred to in your input",
         description="CLARIFICATION: only include people referred to in the conversation. Unless the speakers use each others' names or refer to each other somehow in an utterance, do not include the speakers."
     )
-    places: List[str] = Field(default=[], title="Unique places referred to in your input")
-    companies: List[str] = Field(default=[], title="Unique names of companies, businesses, and/or institutions in your input")
-    body_parts: List[str] = Field(default=[], title="Unique parts of the human body named in your input")
-    emotions: List[str] = Field(default=[], title="Unique emotions, feelings, and/or sentiments referred to in your input")
+    places: List[str] = Field(title="Unique places referred to in your input")
+    companies: List[str] = Field(title="Unique names of companies, businesses, and/or institutions in your input")
+    body_parts: List[str] = Field(title="Unique parts of the human body named in your input")
+    emotions: List[str] = Field(title="Unique emotions, feelings, and/or sentiments referred to in your input")
 
 class Classifier(BaseModel):
     """
     One classifier object
 
     COMBINATION_INSTRUCTION: You must combine classifiers with the same name, such that classifier names are unique in your output. Combine confidences and reasonings, with higher confidence inputs (and the corresponding reasonings) receiving precedence.
     """
+    model_config = ConfigDict(extra='forbid')
     classifier: str = Field(title='name', description="the name of this classifier")
     confidence: float = Field(title='confidence', description='How confident you are in the presence or absence of this classifier in the input you are summarizing', ge=0, le=1)
     reasoning: str = Field(title='reasoning', description="INSTRUCTION: If the definition of this classifier included a 'Specific Items of Interest' appendage, please make sure to note the presence of any of those specific items of interest in this field, independent of their inclusion or exclusion in any entities category. COMBINATION INSTRUCTION: include the union of your inputs' items of interest in your output's reasoning.")
 
 class StructuredResponse(BaseModel):
+    model_config = ConfigDict(extra='forbid')
     summary: str = Field(title='summary of conversation', description="INSTRUCTION: summarize the conversation with one or more precise, declarative statements about the gestalt of the conversation. COMBINATION_INSTRUCTION: only combine the summaries of your input. Do not cross-contaminate your summary with any other pieces of your input objects.")
     primary_topic: str = Field(title='The primary topic of conversation')
     other_topics: List[str] = Field(title='Other topics of conversation', description="INSTRUCTION: do not include the primary_topic in this list")
     classifiers: List[Classifier] = Field(title='A list of classifier results', description="INSTRUCTION: produce based on the Classifiers between <classifiers></classifiers>. Do not create or infer new classifier categories that are not specified below. Include all classifier categories in your response, even those that have very low confidence.")
-    entities: EntitiesObject = Field(title='Entities identified in your input', description="COMBINATION_INSTRUCTION: when combining multiple json objects, treat each field on the entities object as a set of unique strings")
+    entities: EntitiesObject
 
 response_format_json_schema = StructuredResponse.model_json_schema()