save

nexus/actions/models.py

@@ -74,6 +74,7 @@ class Flow(models.Model):
     )
 
     editable = models.BooleanField(default=True)
+    send_llm_response_to_flow = models.BooleanField(default=False)
 
     class Meta:
         constraints = [

router/clients/flows/http/flow_start.py

@@ -19,19 +19,24 @@ def start_flow(
         user: str,
         urns: List,
         user_message: str,
+        llm_message: str = None,
         msg_event: dict = None,
         attachments: list = None
     ) -> None:
 
         url = f"{self.__host}/api/v2/internals/flow_starts/"
+        params = {
+            "message": user_message,
+        }
+
+        if llm_message:
+            params.update({"answer": llm_message})
 
         payload = {
             "user": user,
             "flow": flow.uuid,
             "urns": urns,
-            "params": {
-                "message": user_message,
-            }
+            "params": params
         }
 
         if msg_event:

router/entities/flow.py

@@ -9,3 +9,4 @@ class FlowDTO:
     prompt: str
     fallback: str
     content_base_uuid: str
+    send_llm_response_to_flow: bool = False

router/repositories/orm/__init__.py

@@ -104,7 +104,8 @@ def get_project_flow_by_name(
             name=flow.name,
             prompt=flow.prompt,
             fallback=flow.fallback,
-            content_base_uuid=str(flow.content_base.uuid)
+            content_base_uuid=str(flow.content_base.uuid),
+            send_llm_response_to_flow=flow.send_llm_response_to_flow,
         )
 
     def project_flow_fallback(

router/route.py

@@ -64,54 +64,57 @@ def route(
     log_usecase,
     message_log=None
 ):
-    try:
-        content_base: ContentBaseDTO = content_base_repository.get_content_base_by_project(message.project_uuid)
-
-        if classification == Classifier.CLASSIFICATION_OTHER:
-
-            print("[ + Fallback + ]")
-
-            fallback_flow: FlowDTO = flows_repository.project_flow_fallback(fallback=True)
-
-            if settings.USE_REDIS_CACHE_CONTEXT:
-                last_messages: List[ContactMessageDTO] = message_logs_repository.list_cached_messages(message.project_uuid, message.contact_urn)
-            else:
-                last_messages: List[ContactMessageDTO] = message_logs_repository.list_last_messages(message.project_uuid, message.contact_urn, 5)
-
-            agent: AgentDTO = content_base_repository.get_agent(content_base.uuid)
-            agent = agent.set_default_if_null()
 
-            instructions: List[InstructionDTO] = content_base_repository.list_instructions(content_base.uuid)
-            instructions: List[str] = [instruction.instruction for instruction in instructions]
+    def get_last_messages():
+        if settings.USE_REDIS_CACHE_CONTEXT:
+            last_messages: List[ContactMessageDTO] = message_logs_repository.list_cached_messages(message.project_uuid, message.contact_urn)
+        else:
+            last_messages: List[ContactMessageDTO] = message_logs_repository.list_last_messages(message.project_uuid, message.contact_urn, 5)
+
+        return last_messages
+
+    def get_agent():
+        agent: AgentDTO = content_base_repository.get_agent(content_base.uuid)
+        agent = agent.set_default_if_null()
+        return agent
+
+    def get_instructions():
+        instructions: List[InstructionDTO] = content_base_repository.list_instructions(content_base.uuid)
+        instructions: List[str] = [instruction.instruction for instruction in instructions]
+
+        if instructions == []:
+            instructions += settings.DEFAULT_INSTRUCTIONS
+
+        return instructions
+
+    def get_clean_chunks(full_chunks):
+        chunks: List[str] = []
+        for chunk in full_chunks:
+            full_page = chunk.get("full_page").replace("\x00", "\uFFFD")
+            try:
+                full_page.encode("latin-1")
+                chunks.append(full_page)
+            except UnicodeEncodeError:
+                full_page = fix_encoding(full_page)
+                chunks.append(full_page)
+
+            chunk["full_page"] = full_page
+        return chunks
+
+    def get_llm_response(bad_words_filter: bool = True):
+            last_messages: List[ContactMessageDTO] = get_last_messages()
+            agent: AgentDTO = get_agent()
+            instructions: List[str] = get_instructions()
 
-            if instructions == []:
-                instructions += settings.DEFAULT_INSTRUCTIONS
-
-            # TODO: Implement after changes on create_base_brain_structure usecase.
-            # response_language: str = get_language_codes(llm_config.language)
-
-            # if llm_config.model.lower() != "chatgpt":
-            #     instructions.append(f"Sempre responda em {response_language}")
+            print(f"[+ Instructions: {instructions} +]")
 
             full_chunks: List[Dict] = get_chunks(
                 indexer,
                 text=message.text,
                 content_base_uuid=content_base.uuid
             )
 
-            print(f"[+ Instructions: {instructions} +]")
-
-            chunks: List[str] = []
-            for chunk in full_chunks:
-                full_page = chunk.get("full_page").replace("\x00", "\uFFFD")
-                try:
-                    full_page.encode("latin-1")
-                    chunks.append(full_page)
-                except UnicodeEncodeError:
-                    full_page = fix_encoding(full_page)
-                    chunks.append(full_page)
-
-                chunk["full_page"] = full_page
+            chunks: List[str] = get_clean_chunks(full_chunks)
 
             print(f"[ + Chunks: {chunks} + ]")
 
@@ -125,7 +128,25 @@ def route(
                 last_messages=last_messages,
             )
 
-            llm_response = bad_words_filter(llm_response)
+            if bad_words_filter:
+                llm_response = bad_words_filter(llm_response)
+
+            return llm_response
+
+    try:
+        content_base: ContentBaseDTO = content_base_repository.get_content_base_by_project(message.project_uuid)
+
+        if classification == Classifier.CLASSIFICATION_OTHER:
+            # print("[ + Fallback + ]")
+            # fallback_flow: FlowDTO = flows_repository.project_flow_fallback(fallback=True)
+            llm_response: str = get_llm_response()
+
+
+            # TODO: Implement after changes on create_base_brain_structure usecase.
+            # response_language: str = get_language_codes(llm_config.language)
+
+            # if llm_config.model.lower() != "chatgpt":
+            #     instructions.append(f"Sempre responda em {response_language}")
 
             print(f"[+ LLM Response: {llm_response} +]")
 
@@ -175,6 +196,9 @@ def route(
 
         flow: FlowDTO = flows_repository.get_project_flow_by_name(name=classification)
 
+        if flow.send_llm_response_to_flow:
+            pass
+
         log_usecase.update_log_field(
             project_id=message.project_uuid,
             content_base_id=content_base.uuid,