Create app.py

app.py

@@ -0,0 +1,65 @@
+import os
+from dotenv import load_dotenv
+import gradio as gr
+from langchain.chat_models import ChatOpenAI
+from langchain.document_loaders import PyPDFLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.embeddings import HuggingFaceEmbeddings
+from langchain.vectorstores import FAISS
+from langchain.chains import RetrievalQA
+
+# Carrega variáveis de ambiente
+load_dotenv()
+api_key = os.getenv("OPENROUTER_API_KEY")
+
+# Inicializa LLM
+llm = ChatOpenAI(
+    openai_api_base="https://openrouter.ai/api/v1",
+    openai_api_key=api_key,
+    model="deepseek/deepseek-r1-zero:free"
+)
+
+def processar_pdf(pdf_file, pergunta):
+    # Usa diretamente o caminho fornecido pelo Gradio
+    pdf_path = pdf_file.name
+
+    # Carrega e divide o PDF
+    loader = PyPDFLoader(pdf_path)
+    documents = loader.load()
+    splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=200)
+    docs = splitter.split_documents(documents)
+
+    # Embeddings e índice
+    embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")
+    vectorstore = FAISS.from_documents(docs, embeddings)
+
+    # Cadeia QA
+    qa_chain = RetrievalQA.from_chain_type(
+        llm=llm,
+        retriever=vectorstore.as_retriever(),
+        return_source_documents=True
+    )
+
+    # Resposta
+    resposta = qa_chain.invoke({"query": pergunta})
+    result = resposta["result"]
+    fontes = "\n\n".join([f"Fonte {i+1}: {doc.page_content[:300]}..." for i, doc in enumerate(resposta["source_documents"])])
+    return result, fontes
+
+# Interface Gradio
+interface = gr.Interface(
+    fn=processar_pdf,
+    inputs=[
+        gr.File(label="Envie um PDF"),
+        gr.Textbox(label="Sua pergunta", placeholder="Ex: Qual a duração do curso?")
+    ],
+    outputs=[
+        gr.Textbox(label="Resposta"),
+        gr.Textbox(label="Fontes utilizadas")
+    ],
+    title="Chat com PDF (LangChain)",
+    description="Carregue um PDF e faça perguntas sobre ele. Powered by LangChain + Hugging Face Embeddings"
+)
+
+if __name__ == "__main__":
+    interface.launch()