feat: adicionando temperatura enviada no front para as chamadas do gemini

_utils/gerar_documento_utils/GerarDocumento.py

@@ -166,11 +166,17 @@ class GerarDocumentoUtils:
         elif llm_ultimas_requests == "deepseek-chat":
             llm = llm_instance.deepseek()
         elif llm_ultimas_requests == "gemini-2.0-flash":
-            llm = llm_instance.google_gemini("gemini-2.0-flash")
+            llm = llm_instance.google_gemini(
+                "gemini-2.0-flash", temperature=self.temperature
+            )
         elif llm_ultimas_requests == "gemini-2.5-pro":
-            llm = llm_instance.google_gemini("gemini-2.5-pro-preview-05-06")
+            llm = llm_instance.google_gemini(
+                "gemini-2.5-pro-preview-05-06", temperature=self.temperature
+            )
         elif llm_ultimas_requests == "gemini-2.5-flash":
-            llm = llm_instance.google_gemini("gemini-2.5-flash-preview-04-17")
+            llm = llm_instance.google_gemini(
+                "gemini-2.5-flash-preview-04-17", temperature=self.temperature
+            )
         return llm
 
 
@@ -204,12 +210,13 @@ class GerarDocumento:
         axiom_instance: Axiom,
     ):
         self.gerar_documento_utils = GerarDocumentoUtils(axiom_instance)
+        self.gerar_documento_utils.temperature = serializer.gpt_temperature
         self.config = self.gerar_documento_utils.create_retrieval_config(serializer)
         self.serializer = serializer
         self.logger = logging.getLogger(__name__)
         # self.prompt_auxiliar = prompt_auxiliar
         self.gpt_model = serializer.model
-        self.gpt_temperature = serializer.gpt_temperature
+        self.llm_temperature = serializer.gpt_temperature
         self.prompt_gerar_documento = serializer.prompt_gerar_documento
         self.should_use_llama_parse = serializer.should_use_llama_parse
         self.isBubble = isBubble
@@ -274,7 +281,9 @@ class GerarDocumento:
             "COMEÇANDO REQUISIÇÃO PARA GERAR O QUERY DINAMICAMENTE DO VECTOR STORE"
         )
         response = await self.llm.google_gemini_ainvoke(
-            prompt_para_gerar_query_dinamico, "gemini-2.0-flash"
+            prompt_para_gerar_query_dinamico,
+            "gemini-2.0-flash",
+            temperature=self.llm_temperature,
         )
 
         self.query_gerado_dinamicamente_para_o_vector_store = cast(
@@ -410,7 +419,7 @@ class GerarDocumento:
             )
 
             self.gerar_documento_utils.model = self.gpt_model
-            self.gerar_documento_utils.temperature = self.gpt_temperature
+            self.gerar_documento_utils.temperature = self.llm_temperature
             documento_gerado = await self.gerar_documento_utils.checar_se_resposta_vazia_do_documento_final(
                 llm_ultimas_requests, prompt_primeira_etapa
             )
@@ -493,7 +502,9 @@ class GerarDocumento:
             resumo_para_gerar_titulo = self.texto_completo_como_html
 
         prompt = prompt_para_gerar_titulo(resumo_para_gerar_titulo)
-        response = await agemini_answer(prompt, "gemini-2.0-flash-lite")
+        response = await agemini_answer(
+            prompt, "gemini-2.0-flash-lite", temperature=self.llm_temperature
+        )
         self.titulo_do_documento = response
         return self.titulo_do_documento
 
@@ -578,9 +589,9 @@ class GerarDocumento:
                 )
 
                 prompt = create_prompt_auxiliar_do_contextual_prompt(chunk_text)
-                response = await llms.google_gemini().ainvoke(
-                    [HumanMessage(content=prompt)]
-                )
+                response = await llms.google_gemini(
+                    temperature=self.llm_temperature
+                ).ainvoke([HumanMessage(content=prompt)])
                 responses.append(response.content)
 
                 # Start new chunk with current part
@@ -599,9 +610,9 @@ class GerarDocumento:
                 f"\nProcessing final chunk {chunk_counter} with {current_token_count} tokens"
             )
             prompt = create_prompt_auxiliar_do_contextual_prompt(chunk_text)
-            response = await llms.google_gemini().ainvoke(
-                [HumanMessage(content=prompt)]
-            )
+            response = await llms.google_gemini(
+                temperature=self.llm_temperature
+            ).ainvoke([HumanMessage(content=prompt)])
             responses.append(response.content)
 
         self.resumo_auxiliar = "".join(responses)

_utils/gerar_documento_utils/llm_calls.py

@@ -62,8 +62,9 @@ async def agemini_answer(
     model: Literal[
         "gemini-2.5-pro-preview-05-06", "gemini-2.0-flash", "gemini-2.0-flash-lite"
     ] = "gemini-2.0-flash",
+    temperature=0.4,
 ) -> str:
-    gemini = llm.google_gemini(model)
+    gemini = llm.google_gemini(model, temperature)
     resposta = await gemini.ainvoke([HumanMessage(content=prompt)])
 
     if isinstance(resposta.content, list):

_utils/langchain_utils/LLM_class.py

@@ -32,14 +32,11 @@ class LLM:
             model=model,
         )
 
-    def google_gemini(
-        self,
-        model: Google_llms = "gemini-2.0-flash",
-    ):
+    def google_gemini(self, model: Google_llms = "gemini-2.0-flash", temperature=0.4):
         return ChatGoogleGenerativeAI(
             api_key=SecretStr(google_api_key),
             model=model,
-            temperature=0,
+            temperature=temperature,
             max_tokens=None,
             timeout=None,
             max_retries=2,
@@ -50,10 +47,11 @@ class LLM:
         prompt: str,
         model: Google_llms = "gemini-2.0-flash",
         max_retries: int = 3,
+        temperature=0.4,
     ):
         for attempt in range(max_retries):
             try:
-                response = await self.google_gemini(model).ainvoke(
+                response = await self.google_gemini(model, temperature).ainvoke(
                     [HumanMessage(content=prompt)]
                 )