Upload 3 files

README.md

@@ -1,13 +1,28 @@
 ---
-title: Chat
-emoji: 👀
-colorFrom: indigo
-colorTo: red
+title: Fenix Brain
+emoji: 🧠
+colorFrom: blue
+colorTo: indigo
 sdk: gradio
-sdk_version: 6.19.0
-python_version: '3.12'
+sdk_version: "5.29.0"
 app_file: app.py
-pinned: false
+pinned: true
+license: mit
 ---
 
-Check out the configuration reference at https://huggingface.co/docs/hub/spaces-config-reference
+# 🧠 Fénix Brain
+
+LLM de código para La Forja de Fénix.  
+Modelo: **Qwen2.5-Coder-14B-Instruct** · GPU: ZeroGPU
+
+## API
+
+```
+POST /ask
+{
+  "prompt": "Tu pregunta de código",
+  "key": "TU_BRAIN_API_KEY",
+  "system": "opcional",
+  "max_tokens": 2048
+}
+```

app.py

@@ -0,0 +1,136 @@
+# -*- coding: utf-8 -*-
+"""
+Fénix Brain — LLM de código para La Forja
+==========================================
+Space con ZeroGPU + Qwen2.5-Coder-14B-Instruct
+Expone:
+  POST /ask  {"prompt": "...", "key": "..."}  → {"ok": True, "response": "..."}
+  GET  /     → Chat UI con Gradio
+"""
+
+import os
+import spaces
+import gradio as gr
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
+
+MODEL_ID  = "Qwen/Qwen2.5-Coder-14B-Instruct"
+API_KEY   = os.environ.get("BRAIN_API_KEY", "")
+
+# ── Cargar modelo ─────────────────────────────────────────────────────────────
+tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True)
+model     = None  # Se carga en primera petición con ZeroGPU
+
+def _cargar_modelo():
+    global model
+    if model is None:
+        model = AutoModelForCausalLM.from_pretrained(
+            MODEL_ID,
+            torch_dtype=torch.bfloat16,
+            device_map="auto",
+            trust_remote_code=True,
+        )
+    return model
+
+# ── Inferencia ────────────────────────────────────────────────────────────────
+@spaces.GPU(duration=120)
+def _inferir(prompt: str, system: str = "", max_tokens: int = 2048) -> str:
+    mdl = _cargar_modelo()
+    messages = []
+    if system:
+        messages.append({"role": "system", "content": system})
+    messages.append({"role": "user", "content": prompt})
+
+    text = tokenizer.apply_chat_template(
+        messages, tokenize=False, add_generation_prompt=True
+    )
+    inputs = tokenizer([text], return_tensors="pt").to(mdl.device)
+    with torch.no_grad():
+        out = mdl.generate(
+            **inputs,
+            max_new_tokens=max_tokens,
+            temperature=0.2,
+            do_sample=True,
+            pad_token_id=tokenizer.eos_token_id,
+        )
+    generated = out[0][inputs["input_ids"].shape[1]:]
+    return tokenizer.decode(generated, skip_special_tokens=True)
+
+# ── FastAPI endpoints ─────────────────────────────────────────────────────────
+from fastapi import FastAPI, Request
+from fastapi.responses import JSONResponse
+
+fapp = FastAPI()
+
+@fapp.post("/ask")
+async def ask(request: Request):
+    try:
+        data   = await request.json()
+        key    = data.get("key", "")
+        if API_KEY and key != API_KEY:
+            return JSONResponse({"ok": False, "error": "Unauthorized"}, status_code=401)
+        prompt = data.get("prompt", "")
+        system = data.get("system", "Eres un experto programador Python y Gradio. Responde siempre en español con código limpio y funcional.")
+        max_t  = int(data.get("max_tokens", 2048))
+        if not prompt:
+            return JSONResponse({"ok": False, "error": "prompt requerido"}, status_code=400)
+        respuesta = _inferir(prompt, system, max_t)
+        return JSONResponse({"ok": True, "response": respuesta})
+    except Exception as e:
+        return JSONResponse({"ok": False, "error": str(e)}, status_code=500)
+
+@fapp.get("/health")
+async def health():
+    return JSONResponse({"ok": True, "model": MODEL_ID})
+
+# ── Gradio UI (chat) ──────────────────────────────────────────────────────────
+SYSTEM_DEFAULT = "Eres Fénix Brain, el asistente de código de La Forja de Fénix. Eres experto en Python, Gradio, FastAPI y HuggingFace Spaces. Responde siempre en español con código limpio."
+
+def chat(mensaje, historial, system_prompt):
+    if not mensaje.strip():
+        return historial, ""
+    messages = [{"role": "system", "content": system_prompt or SYSTEM_DEFAULT}]
+    for h in historial:
+        messages.append({"role": "user",      "content": h[0]})
+        messages.append({"role": "assistant", "content": h[1]})
+    messages.append({"role": "user", "content": mensaje})
+
+    text = tokenizer.apply_chat_template(
+        messages, tokenize=False, add_generation_prompt=True
+    )
+
+    @spaces.GPU(duration=120)
+    def _gen():
+        mdl    = _cargar_modelo()
+        inputs = tokenizer([text], return_tensors="pt").to(mdl.device)
+        with torch.no_grad():
+            out = mdl.generate(
+                **inputs, max_new_tokens=2048, temperature=0.2,
+                do_sample=True, pad_token_id=tokenizer.eos_token_id,
+            )
+        gen = out[0][inputs["input_ids"].shape[1]:]
+        return tokenizer.decode(gen, skip_special_tokens=True)
+
+    respuesta = _gen()
+    historial = historial + [[mensaje, respuesta]]
+    return historial, ""
+
+with gr.Blocks(title="🧠 Fénix Brain", theme=gr.themes.Base()) as demo:
+    gr.Markdown("# 🧠 Fénix Brain\nAsistente de código para La Forja · Qwen2.5-Coder-14B")
+    system_box = gr.Textbox(label="System prompt", value=SYSTEM_DEFAULT, lines=2)
+    chatbot    = gr.Chatbot(height=500, label="Chat")
+    msg_box    = gr.Textbox(placeholder="Escribe tu pregunta de código…", label="Mensaje")
+    with gr.Row():
+        send_btn  = gr.Button("Enviar", variant="primary")
+        clear_btn = gr.Button("Limpiar")
+
+    send_btn.click(chat,  [msg_box, chatbot, system_box], [chatbot, msg_box])
+    msg_box.submit(chat,  [msg_box, chatbot, system_box], [chatbot, msg_box])
+    clear_btn.click(lambda: ([], ""), outputs=[chatbot, msg_box])
+
+# ── Montar Gradio en FastAPI ──────────────────────────────────────────────────
+app = gr.mount_gradio_app(fapp, demo, path="/")
+
+if __name__ == "__main__":
+    import uvicorn
+    uvicorn.run(app, host="0.0.0.0", port=7860)

requirements.txt

@@ -0,0 +1,7 @@
+gradio==5.29.0
+transformers>=4.45.0
+torch>=2.1.0
+accelerate>=0.26.0
+fastapi
+uvicorn
+spaces