Feat: refactor settings.py add LLM_ENGINGES for simple usr in DeepDiv…

…eService

backend/.gitignore

@@ -159,4 +159,6 @@ cython_debug/
 #  option (not recommended) you can uncomment the following to ignore the entire idea folder.
 #.idea/
 
-audio_files/
+audio_files/
+
+.env

backend/src/api.py

@@ -7,8 +7,7 @@
 from .services.exceptions import FileFormatException, NotFoundException
 from .services.models import UploadResponse
 from .services.services import DeepDive
-from .settings.development import CORS_CONFIG, OPEN_AI_TOKEN
-from .summarization.engines import LangChain
+from .settings.development import CORS_CONFIG, LLM_ENGINE
 
 app = FastAPI()
 
@@ -29,8 +28,7 @@ async def healthcheck():
 
 @app.post("/process-audio/", response_model=UploadResponse)
 async def create_upload_file(request: Request, audio_file: UploadFile = File(..., description="arquivo .mp3 ou .wav")):
-    lang_chain = LangChain(api_key=OPEN_AI_TOKEN)
-    service = DeepDive(llm_engine=lang_chain)
+    service = DeepDive(llm_engine=LLM_ENGINE.get('LangChain'))
     service.validate_api_token()
     response = service.upload_audio(audio_file)
 

backend/src/env.exemple

@@ -0,0 +1,3 @@
+# Configurações de API
+OPEN_AI_TOKEN=sk-change-me
+TENSORFLOW_MODEL_NAME="t5-small"

backend/src/settings/development.py

@@ -1,4 +1,10 @@
-# Settings development mode
+import os
+
+from dotenv import load_dotenv
+from src.summarization.engines import LangChain, TensorFlow
+
+# Carregar variáveis de ambiente do arquivo .env
+load_dotenv()
 
 CORS_CONFIG = {
     "allow_origins": ["*"],
@@ -7,5 +13,10 @@
     "allow_headers": ["*"],
 }
 
+OPEN_AI_TOKEN = os.getenv("OPEN_AI_TOKEN")
+TENSORFLOW_MODEL_NAME = os.getenv("TENSORFLOW_MODEL_NAME")
 
-OPEN_AI_TOKEN = "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"
+LLM_ENGINE = {
+    "LangChain": LangChain(api_key=OPEN_AI_TOKEN),
+    "Tensorflow": TensorFlow(model_name=TENSORFLOW_MODEL_NAME),
+}

backend/src/summarization/engines.py

@@ -4,6 +4,7 @@
 from langchain.llms.openai import OpenAI
 from src.services.exceptions import ApiKeyException
 from src.summarization.interfaces import Summarization
+from transformers import AutoTokenizer, TFAutoModelForSeq2SeqLM
 
 
 class LangChain(Summarization):
@@ -45,9 +46,22 @@ def token_is_valid(self):
             raise ApiKeyException()
 
 
-class TensorFlowStrategy(Summarization):
-    def __init__(self):
-        pass
+class TensorFlow(Summarization):
+    def __init__(self, model_name="t5-small"):
+        self.tokenizer = AutoTokenizer.from_pretrained(model_name)
+        self.model = TFAutoModelForSeq2SeqLM.from_pretrained(model_name)
 
-    def summarize(self, text: str) -> str:
-        return "Texto sumarizado com TensorFlow"
+    def summarize(self, text: str, max_length=650):
+        # Preparar a entrada para o modelo
+        inputs = self.tokenizer.encode(
+            f"summarize: {text}", return_tensors="tf", max_length=1500, truncation=True
+        )
+        # Gerar a saída do modelo
+        summary_ids = self.model.generate(
+            inputs,
+            max_length=max_length,
+            length_penalty=4.0,
+            num_beams=4,
+            early_stopping=True,
+        )
+        return self.tokenizer.decode(summary_ids[0], skip_special_tokens=True)