Update backend/main.py

backend/main.py

@@ -1,4 +1,4 @@
-#--- START OF FILE main.py ---
+#--- START OF FILE main (1).py ---
 
 import os
 import io
@@ -14,7 +14,7 @@ from fastapi.responses import StreamingResponse
 # RAG Imports
 from langchain_community.document_loaders import PyPDFLoader
 from langchain_community.embeddings import HuggingFaceEmbeddings
-from langchain_text_splitters import RecursiveCharacterTextSplitter # CORRIGIDO: Nova importação
+from langchain_text_splitters import RecursiveCharacterTextSplitter
 from langchain_community.vectorstores import FAISS
 from langchain_core.runnables import RunnablePassthrough, RunnableLambda
 from langchain_core.output_parsers import StrOutputParser
@@ -38,9 +38,10 @@ app.add_middleware(
 # Define o modelo de embedding do Hugging Face (leve para CPU)
 HF_EMBEDDING_MODEL = "sentence-transformers/all-MiniLM-L6-v2"
 
-# Inicializa o modelo Groq e o modelo de embedding
+# Inicializa o modelo Groq
 model = ChatGroq(model=os.getenv("GROQ_MODEL", "mixtral-8x7b-32768"))
-# MUDANÇA: Inicializa o HuggingFaceEmbeddings na CPU
+
+# Inicializa o HuggingFaceEmbeddings na CPU
 embeddings = HuggingFaceEmbeddings(
     model_name=HF_EMBEDDING_MODEL,
     model_kwargs={'device': 'cpu'} 
@@ -108,9 +109,12 @@ async def upload_document(file: UploadFile = File(...)):
         vectorstore = FAISS.from_documents(documents=splits, embedding=embeddings)
         retriever = vectorstore.as_retriever(search_kwargs={"k": 4}) 
         
-        # 5. Criar a nova Chain RAG
+        # 5. Criar a nova Chain RAG (CORRIGIDO)
+        # O lambda extrai apenas o texto da pergunta ("input") do dicionário que chega
         rag_chain = (
-            RunnablePassthrough.assign(context=retriever | format_docs)
+            RunnablePassthrough.assign(
+                context=(lambda x: x["input"]) | retriever | format_docs
+            )
             | rag_prompt
             | model
             | StrOutputParser()
@@ -121,7 +125,7 @@ async def upload_document(file: UploadFile = File(...)):
     except Exception as e:
         print(f"Erro no processamento do arquivo: {e}")
         # Retorna um erro 500 para o frontend
-        raise HTTPException(status_code=500, detail=f"Falha ao processar o arquivo: {e}. Verifique se o modelo HuggingFace foi baixado corretamente.")
+        raise HTTPException(status_code=500, detail=f"Falha ao processar o arquivo: {e}")
     finally:
         # Limpeza: deletar o arquivo temporário
         if 'temp_path' in locals() and os.path.exists(temp_path):
@@ -146,6 +150,7 @@ async def chat(request: ChatRequest):
     async def stream_generator():
         try:
             # 'astream' é o método de streaming assíncrono do LangChain
+            # Passamos {"input": ...} que será interceptado pelo lambda definido acima
             async for chunk in current_chain.astream({"input": request.content}):
                 if chunk:
                     yield chunk
@@ -155,4 +160,5 @@ async def chat(request: ChatRequest):
 
     # Retorna uma resposta de streaming
     return StreamingResponse(stream_generator(), media_type="text/plain")
-#--- END OF FILE main.py ---
+
+#--- END OF FILE main (1).py ---