Create app.py

app.py

@@ -0,0 +1,54 @@
+import gradio as gr
+from langchain_community.chat_models import HuggingFaceHub
+from langchain_community.vectorstores import Chroma
+from langchain.document_loaders import PyPDFLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_core.output_parsers import StrOutputParser
+from langchain_huggingface import HuggingFaceEmbeddings
+from langchain import hub
+from rerankers import Reranker
+import os
+
+# Configuración del token de acceso a Hugging Face (si usas modelo privado)
+os.environ["HUGGINGFACEHUB_API_TOKEN"] = os.getenv("HUGGINGFACEHUB_API_TOKEN", "")
+
+# Cargar PDF
+loader = PyPDFLoader("80dias.pdf")
+documents = loader.load()
+splitter = RecursiveCharacterTextSplitter(chunk_size=1000, chunk_overlap=20)
+splits = splitter.split_documents(documents)
+
+# Crear embeddings
+embedding_model = "sentence-transformers/paraphrase-multilingual-mpnet-base-v2"
+embeddings = HuggingFaceEmbeddings(model_name=embedding_model)
+vectordb = Chroma.from_documents(splits, embedding=embeddings)
+
+# Modelo LLM desde HuggingFace (usa uno disponible en Spaces)
+llm = HuggingFaceHub(repo_id="mistralai/Mistral-7B-Instruct-v0.1", model_kwargs={"temperature": 0.5, "max_new_tokens": 500})
+chain = llm | StrOutputParser()
+
+# Reranker
+ranker = Reranker("answerdotai/answerai-colbert-small-v1", model_type="colbert")
+
+# Función RAG
+def rag_chat(query):
+    results = vectordb.similarity_search_with_score(query)
+    context = []
+    for doc, score in results:
+        if score < 7:
+            context.append(doc.page_content)
+    if not context:
+        return "No tengo información para responder a esa pregunta."
+
+    ranking = ranker.rank(query=query, docs=context)
+    best_context = ranking[0].text
+
+    prompt = hub.pull("rlm/rag-prompt")
+    rag_chain = prompt | llm | StrOutputParser()
+
+    result = rag_chain.invoke({"context": best_context, "question": query})
+    return result
+
+# Interfaz Gradio
+iface = gr.ChatInterface(fn=rag_chat, title="Chat Julio Verne - RAG", description="Pregunta lo que quieras sobre *La vuelta al mundo en 80 días* de Julio Verne.")
+iface.launch()