Update app.py

app.py

@@ -3,47 +3,87 @@ from pathlib import Path
 import gradio as gr
 import numpy as np
 import faiss
+from sentence_transformers import SentenceTransformer
 from openai import OpenAI, OpenAIError
-from langchain_community.document_loaders import PyPDFLoader
-from langchain.text_splitter import RecursiveCharacterTextSplitter
 
 # ----------------------------
-# Context retrieval using local SentenceTransformer embeddings
+# NVIDIA API Key (via Space Variables & Secrets)
+# ----------------------------
+NV_API_KEY = os.environ.get("NV_API_KEY")
+if not NV_API_KEY:
+    raise RuntimeError(
+        "🔒 NV_API_KEY not set. "
+        "In your Hugging Face Space, go to Settings → Variables & Secrets and create NV_API_KEY."
+    )
+
+# NVIDIA-compatible OpenAI client for chat
+client = OpenAI(
+    base_url="https://integrate.api.nvidia.com/v1",
+    api_key=NV_API_KEY
+)
+CHAT_MODEL = "meta/llama3-8b-instruct"
+
+# ----------------------------
+# Application configuration
+# ----------------------------
+APP_TITLE = "CVchat – Ronaldo Menezes"
+INTRO = (
+    "👋 Olá! Eu sou o CVchat do Ronaldo Menezes.\n"
+    "Converse sobre minha experiência, projetos, tecnologias, resultados e muito mais.\n\n"
+    "Exemplos de perguntas:\n"
+    "• Quem é o Ronaldo Menezes\n"
+    "• Resuma sua experiência com Process Mining.\n"
+    "• Que linguagens e ferramentas você domina?\n"
+    "• Fale de um projeto com financiamento público que você liderou.\n"
+)
+SUGGESTION_QUESTIONS = [
+    "Links & exemplos de trabalhos",
+    "Quais tecnologias você mais usa?",
+    "Resuma sua experiência com Machine Learning.",
+    "Artigo sobre Landsat ou Sentinel?",
+    "Você já trabalhou com mainframe/COBOL?",
+    "Certificações?",
+]
+
+# Paths for FAISS files
+INDEX_FILE  = "r_docs.index"
+CHUNKS_FILE = "r_chunks.npy"
+PDF_PATH    = "CV-Ronaldo_Menezes_2025_06.pdf"
+
+# verify index files exist
+if not Path(INDEX_FILE).exists() or not Path(CHUNKS_FILE).exists():
+    raise FileNotFoundError(
+        "Index not found. Run build_index.py to generate r_docs.index and r_chunks.npy."
+    )
+
+# load FAISS index and chunks
+index  = faiss.read_index(INDEX_FILE)
+chunks = np.load(CHUNKS_FILE, allow_pickle=True)(CHUNKS_FILE, allow_pickle=True)
+
+# ----------------------------
+# Local embedding model (context retrieval)
 # ----------------------------
-# Embedding model for context retrieval (local)
 embedding_model = SentenceTransformer('all-MiniLM-L6-v2')
 
 def retrieve_context(query: str, k: int = 4) -> str:
-    # encode locally
     q_emb = embedding_model.encode([query], convert_to_numpy=True)
     _, I = index.search(q_emb, k)
-    return "
----
-".join(chunks[i] for i in I[0])(query: str, k: int = 4) -> str:
-    # call NVIDIA embeddings
-    try:
-        resp = client.embeddings.create(
-            model=EMBED_MODEL,
-            input=[query]
-        )
-        q_emb = np.array(resp.data[0].embedding, dtype=np.float32)
-    except OpenAIError as e:
-        raise RuntimeError(f"Embedding error: {e}")
-    # search top k
-    _, I = tmp_index.search(q_emb.reshape(1, -1), k)
     return "\n---\n".join(chunks[i] for i in I[0])
 
 # ----------------------------
-# Chatbot logic
+# Chatbot state
 # ----------------------------
 dialog_history: list[tuple[str, str]] = []
 
-def chatbot(user_input, temperature, top_p, max_tokens):
+# ----------------------------
+# Chat function
+# ----------------------------
+def chatbot(user_input: str, temperature: float, top_p: float, max_tokens: int):
     global dialog_history
     if not user_input:
         return dialog_history, ""
 
-    # retrieve context
+    # build system message with retrieved context
     context = retrieve_context(user_input)
     system_msg = {
         "role": "system",
@@ -53,16 +93,17 @@ def chatbot(user_input, temperature, top_p, max_tokens):
             f"=== Retrieved Context ===\n{context}\n\n"
         )
     }
-    # build messages
+
+    # assemble conversation
     messages = [system_msg]
     for u, a in dialog_history:
-        messages.extend([
+        messages += [
             {"role": "user", "content": u},
             {"role": "assistant", "content": a}
-        ])
+        ]
     messages.append({"role": "user", "content": user_input})
 
-    # call NVIDIA chat API streaming
+    # call NVIDIA chat API
     assistant_reply = ""
     try:
         stream = client.chat.completions.create(
@@ -83,7 +124,9 @@ def chatbot(user_input, temperature, top_p, max_tokens):
     dialog_history.append((user_input, assistant_reply))
     return dialog_history, ""
 
-# clear history
+# ----------------------------
+# Clear chat history
+# ----------------------------
 def clear_history():
     global dialog_history
     dialog_history = []
@@ -101,13 +144,44 @@ custom_css = r"""
   --radius: 8px;
   --spacing: 1rem;
 }
-body { background: var(--bg-light); color: var(--txt-dark); font-family: 'Helvetica Neue', sans-serif; }
-#chat-window { height: 65vh; overflow-y: auto; padding: var(--spacing); border: 1px solid #ddd; border-radius: var(--radius); }
-#input-area { display: flex; margin-top: var(--spacing); }
-#user-input { flex: 1; padding: 0.6rem; border: 1px solid #ccc; border-radius: var(--radius) 0 0 var(--radius); }
-#send-button { padding: 0 1rem; background: var(--primary); color: white; border: none; border-radius: 0 var(--radius) var(--radius) 0; cursor: pointer; }
-.sidebar { background: var(--bg-light); padding: var(--spacing); border-left: 1px solid #eee; }
-.sidebar h3 { margin-top: 0; }
+body {
+  background: var(--bg-light);
+  color: var(--txt-dark);
+  font-family: 'Helvetica Neue', sans-serif;
+}
+#chat-window {
+  height: 65vh;
+  overflow-y: auto;
+  padding: var(--spacing);
+  border: 1px solid #ddd;
+  border-radius: var(--radius);
+}
+#input-area {
+  display: flex;
+  margin-top: var(--spacing);
+}
+#user-input {
+  flex: 1;
+  padding: 0.6rem;
+  border: 1px solid #ccc;
+  border-radius: var(--radius) 0 0 var(--radius);
+}
+#send-button {
+  padding: 0 1rem;
+  background: var(--primary);
+  color: white;
+  border: none;
+  border-radius: 0 var(--radius) var(--radius) 0;
+  cursor: pointer;
+}
+.sidebar {
+  background: var(--bg-light);
+  padding: var(--spacing);
+  border-left: 1px solid #eee;
+}
+.sidebar h3 {
+  margin-top: 0;
+}
 """
 
 with gr.Blocks(title=APP_TITLE, css=custom_css, theme=gr.themes.Base()) as demo:
@@ -115,15 +189,21 @@ with gr.Blocks(title=APP_TITLE, css=custom_css, theme=gr.themes.Base()) as demo:
     gr.Markdown(INTRO)
 
     with gr.Row():
+        # main chat column
         with gr.Column(scale=3):
             chatbot_ui = gr.Chatbot(type="tuples", elem_id="chat-window")
             with gr.Row(elem_id="input-area"):
                 txt = gr.Textbox(placeholder="Digite sua pergunta…", lines=2, elem_id="user-input")
                 btn = gr.Button("Enviar", elem_id="send-button")
-            btn.click(chatbot, [txt, gr.Slider(0,1,0.6), gr.Slider(0,1,0.95), gr.Slider(64,2048,512)], [chatbot_ui, txt])
-            txt.submit(chatbot, [txt, gr.Slider(0,1,0.6), gr.Slider(0,1,0.95), gr.Slider(64,2048,512)], [chatbot_ui, txt])
+            btn.click(chatbot,
+                      [txt, gr.Slider(0,1,0.6), gr.Slider(0,1,0.95), gr.Slider(64,2048,512)],
+                      [chatbot_ui, txt])
+            txt.submit(chatbot,
+                       [txt, gr.Slider(0,1,0.6), gr.Slider(0,1,0.95), gr.Slider(64,2048,512)],
+                       [chatbot_ui, txt])
             gr.Button("Limpar").click(clear_history, [], [chatbot_ui, txt])
 
+        # sidebar with PDF & suggestions
         with gr.Column(scale=1, elem_classes="sidebar"):
             if Path(PDF_PATH).exists():
                 gr.Markdown(f"[📄 Baixar CV em PDF](/file={PDF_PATH})")
@@ -140,3 +220,4 @@ if __name__ == "__main__":
 
 
 
+