Create app.py

app.py

@@ -0,0 +1,184 @@
+import os
+import warnings
+import torch
+import gradio as gr
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+# ==========================
+# CONFIG
+# ==========================
+
+MODEL_NAME = "sshleifer/tiny-gpt2"
+MAX_NEW_TOKENS = 300
+MAX_INPUT_CHARS = 4000
+MAX_HISTORY = 5
+
+# ==========================
+# SETTINGS
+# ==========================
+
+warnings.filterwarnings("ignore")
+os.environ["TOKENIZERS_PARALLELISM"] = "false"
+
+torch.set_grad_enabled(False)
+torch.set_num_threads(max(1, os.cpu_count() // 2))
+
+device = "cuda" if torch.cuda.is_available() else "cpu"
+
+print("🔄 Loading tokenizer...")
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+
+print("🔄 Loading model...")
+model = AutoModelForCausalLM.from_pretrained(
+    MODEL_NAME,
+    low_cpu_mem_usage=True
+).to(device)
+
+model.eval()
+
+if tokenizer.pad_token is None:
+    tokenizer.pad_token = tokenizer.eos_token
+
+print("✅ Model loaded successfully.")
+
+# ==========================
+# PROMPTS
+# ==========================
+
+GENERATOR_PROMPT = """
+You are a Senior Software Engineer.
+Generate clean, secure, optimized production-ready code.
+Explain briefly what the code does.
+"""
+
+ULTRA_REVIEW_PROMPT = """
+You are a Principal Software Architect and Security Engineer.
+Perform a strict professional review.
+
+Return:
+
+## Executive Summary
+## Critical Bugs
+## Security Issues
+## Architecture Problems
+## Performance Issues
+## Code Quality Problems
+## Refactored Version
+## Final Verdict
+"""
+
+ARCHITECTURE_ANALYSIS_PROMPT = """
+You are a Senior Software Architect.
+Perform deep architectural analysis.
+
+Return:
+
+## Architecture Overview
+## Structural Problems
+## Scalability Evaluation
+## Resilience Evaluation
+## Improvements
+## Final Verdict
+"""
+
+# ==========================
+# PROMPT BUILDER
+# ==========================
+
+def build_prompt(mode, user_input, history):
+
+    if mode == "Generar Código":
+        system_prompt = GENERATOR_PROMPT
+    elif mode == "Revisión Ultra Crítica":
+        system_prompt = ULTRA_REVIEW_PROMPT
+        user_input = f"Deep review:\n\n{user_input}"
+    else:
+        system_prompt = ARCHITECTURE_ANALYSIS_PROMPT
+        user_input = f"Deep architectural analysis:\n\n{user_input}"
+
+    conversation = system_prompt + "\n\n"
+
+    history = history[-MAX_HISTORY:]
+
+    for user, assistant in history:
+        conversation += f"User: {user}\nAssistant: {assistant}\n"
+
+    conversation += f"User: {user_input}\nAssistant:"
+
+    return conversation
+
+# ==========================
+# GENERATION
+# ==========================
+
+def generate_text(prompt):
+
+    inputs = tokenizer(
+        prompt,
+        return_tensors="pt",
+        truncation=True,
+        max_length=1024
+    ).to(device)
+
+    with torch.inference_mode():
+        outputs = model.generate(
+            **inputs,
+            max_new_tokens=MAX_NEW_TOKENS,
+            temperature=0.3,
+            top_p=0.9,
+            do_sample=False,
+            pad_token_id=tokenizer.eos_token_id
+        )
+
+    generated_tokens = outputs[0][inputs["input_ids"].shape[-1]:]
+
+    return tokenizer.decode(
+        generated_tokens,
+        skip_special_tokens=True
+    ).strip()
+
+# ==========================
+# CHAT
+# ==========================
+
+def chat(user_input, history, mode):
+
+    if not user_input:
+        return "⚠️ Empty input."
+
+    if len(user_input) > MAX_INPUT_CHARS:
+        return "⚠️ Input too large."
+
+    try:
+        prompt = build_prompt(mode, user_input, history)
+        return generate_text(prompt)
+    except RuntimeError as e:
+        return f"⚠️ Generation error: {str(e)}"
+    except Exception:
+        return "⚠️ Unexpected error occurred."
+
+# ==========================
+# UI
+# ==========================
+
+with gr.Blocks(theme=gr.themes.Soft()) as demo:
+
+    gr.Markdown("# 🧠 Federico - Lightweight CTO Assistant")
+
+    mode = gr.Radio(
+        [
+            "Generar Código",
+            "Revisión Ultra Crítica",
+            "Análisis Arquitectura"
+        ],
+        value="Revisión Ultra Crítica",
+        label="Modo"
+    )
+
+    gr.ChatInterface(
+        fn=chat,
+        additional_inputs=[mode],
+        chatbot=gr.Chatbot(height=500),
+    )
+
+demo.launch(server_name="0.0.0.0", server_port=7860)