Update app.py

app.py

@@ -2,6 +2,85 @@ import gradio as gr
 import os
 from langchain_community.document_loaders import PyPDFDirectoryLoader
 from langchain_text_splitters import RecursiveCharacterTextSplitter
+from langchain_huggingface import HuggingFaceEmbeddings, HuggingFaceEndpoint
+from langchain_community.vectorstores import FAISS
+from langchain.chains import create_retrieval_chain
+from langchain.chains.combine_documents import create_stuff_documents_chain
+from langchain_core.prompts import ChatPromptTemplate
+
+DATA_DIR = "Data" if os.path.exists("Data") else "data"
+if not os.path.exists(DATA_DIR):
+    os.makedirs(DATA_DIR)
+
+print(f"Usando diretorio: {DATA_DIR}")
+print(f"Arquivos: {os.listdir(DATA_DIR) if os.path.exists(DATA_DIR) else 'vazio'}")
+
+loader = PyPDFDirectoryLoader(DATA_DIR)
+documents = loader.load()
+print(f"Documentos carregados: {len(documents)}")
+
+text_splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=150)
+docs = text_splitter.split_documents(documents)
+
+model_name = "sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2"
+embeddings = HuggingFaceEmbeddings(model_name=model_name)
+
+if docs:
+    db = FAISS.from_documents(docs, embeddings)
+else:
+    import faiss
+    dimension = 384
+    index = faiss.IndexFlatL2(dimension)
+    from langchain_community.docstore.in_memory import InMemoryDocstore
+    db = FAISS(embedding_function=embeddings, index=index, docstore=InMemoryDocstore(), index_to_docstore_id={})
+
+retriever = db.as_retriever(search_kwargs={"k": 3})
+
+# Usando HuggingFaceEndpoint (substituto moderno do HuggingFaceHub)
+llm = HuggingFaceEndpoint(
+    repo_id="mistralai/Mistral-7B-Instruct-v0.3",
+    task="text-generation",
+    huggingfacehub_api_token=os.getenv("HUGGINGFACEHUB_API_TOKEN"),
+    max_new_tokens=512,
+    temperature=0.7
+)
+
+system_prompt = (
+    "Você é o Prometheus, um assistente especializado no DETRAN-RJ. "
+    "Use o contexto abaixo para responder às perguntas dos usuários de forma clara e profissional. "
+    "Se não souber a resposta com base no contexto, diga que não encontrou a informação específica, "
+    "mas tente ajudar com o que for possível.\n\n"
+    "{context}"
+)
+
+prompt = ChatPromptTemplate.from_messages([
+    ("system", system_prompt),
+    ("human", "{input}"),
+])
+
+question_answer_chain = create_stuff_documents_chain(llm, prompt)
+rag_chain = create_retrieval_chain(retriever, question_answer_chain)
+
+def respond(message, history):
+    try:
+        response = rag_chain.invoke({"input": message})
+        return response["answer"]
+    except Exception as e:
+        return f"Erro ao processar consulta: {str(e)}"
+
+demo = gr.ChatInterface(
+    respond,
+    title="Prometheus — Agente Especialista DETRAN-RJ",
+    description="IA treinada com regulamentações e manuais do DETRAN-RJ para suporte ao cidadão.",
+    examples=["Como renovar a CNH?", "Quais os documentos para transferência de veículo?", "O que é o GRT?"]
+)
+
+if __name__ == "__main__":
+    demo.launch()
+    import gradio as gr
+import os
+from langchain_community.document_loaders import PyPDFDirectoryLoader
+from langchain_text_splitters import RecursiveCharacterTextSplitter
 from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain_community.vectorstores import FAISS
 from langchain.chains import create_retrieval_chain