test: current progress with the space

gradio_interface/app.py

@@ -1,7 +1,125 @@
+import os
 import gradio as gr
+from os import getenv
+import base64
+from io import BytesIO
+from dotenv import load_dotenv
+import requests
+import socket
+import logging
 
-def greet(name):
-    return "Hello " + name + "!!"
+from langchain_openai import ChatOpenAI
+from langchain_core.messages import HumanMessage, AIMessage
+from langchain_core.callbacks import StreamingStdOutCallbackHandler
+
+# Configure logging
+logging.basicConfig(level=logging.INFO)
+logger = logging.getLogger(__name__)
+
+# Load environment
+dotenv_path = os.path.join(os.path.dirname(__file__), '.env')
+load_dotenv(dotenv_path=dotenv_path)
+
+# Debug env
+logger.info(f"OPENROUTER_BASE_URL: {getenv('OPENROUTER_BASE_URL')}")
+logger.info(f"OPENROUTER_API_KEY: {'Found' if getenv('OPENROUTER_API_KEY') else 'Missing'}")
+
+# Connectivity test
+def test_connectivity(url="https://openrouter.helicone.ai/api/v1"):
+    try:
+        return requests.get(url, timeout=5).status_code == 200
+    except (requests.RequestException, socket.error) as e:
+        logger.error(f"Connectivity test failed: {e}")
+        return False
+
+if not test_connectivity():
+    logger.warning("No network to OpenRouter; responses may fail.")
+
+# Initialize LLM with streaming and retry logic
+def init_llm():
+    if not test_connectivity():
+        raise RuntimeError("No hay conexión a OpenRouter. Verifica red y claves.")
+    return ChatOpenAI(
+        openai_api_key=getenv("OPENROUTER_API_KEY"),
+        openai_api_base=getenv("OPENROUTER_BASE_URL"),
+        model_name="google/gemini-flash-1.5",
+        streaming=True,
+        callbacks=[StreamingStdOutCallbackHandler()],
+        model_kwargs={
+            "extra_headers": {"Helicone-Auth": f"Bearer {getenv('HELICONE_API_KEY')}"}
+        },
+    )
+
+llm = init_llm()
+
+# Helpers
+def encode_image_to_base64(pil_image):
+    buffer = BytesIO()
+    pil_image.save(buffer, format="PNG")
+    return f"data:image/png;base64,{base64.b64encode(buffer.getvalue()).decode()}"
+
+# Core logic
+def generate_response(message, chat_history, image):
+    messages = [HumanMessage(content="You are an expert image analysis assistant. Answer succinctly.")]
+    for msg in chat_history:
+        role = msg.get('role')
+        content = msg.get('content')
+        if role == 'user':
+            messages.append(HumanMessage(content=content))
+        else:
+            messages.append(AIMessage(content=content))
+    encoded = encode_image_to_base64(image)
+    messages.append(HumanMessage(content={"type":"text","text":message}))
+    messages.append(HumanMessage(content={"type":"image_url","image_url":encoded}))
+
+    try:
+        stream_iter = llm.stream(messages)
+        if stream_iter is None:
+            raise RuntimeError("Received no stream iterator from LLM.")
+        partial = ""
+        for chunk in stream_iter:
+            if chunk is None:
+                logger.warning("Received None chunk from stream, skipping.")
+                continue
+            content = getattr(chunk, 'content', None)
+            if content is None:
+                logger.warning(f"Chunk without content: {chunk}")
+                continue
+            partial += content
+            yield partial
+    except AssertionError as e:
+        logger.error(f"AssertionError in stream: {e}")
+        yield "⚠️ No response del modelo. Por favor reintenta."  
+    except Exception as e:
+        logger.exception("Unexpected error during streaming response.")
+        yield "⚠️ Error al generar respuesta. Intenta más tarde."
+
+# Gradio interface
+def process_message(message, chat_history, image):
+    if chat_history is None:
+        chat_history = []
+    if image is None:
+        chat_history.append({'role':'assistant','content':'Por favor sube una imagen.'})
+        return "", chat_history
+    chat_history.append({'role':'user','content':message})
+    chat_history.append({'role':'assistant','content':'⏳ Procesando...'})
+    yield "", chat_history
+    for chunk in generate_response(message, chat_history, image):
+        chat_history[-1]['content'] = chunk
+        yield "", chat_history
+    return "", chat_history
+
+with gr.Blocks() as demo:
+    with gr.Row():
+        with gr.Column(scale=2):
+            chatbot = gr.Chatbot(type='messages', height=600)
+            msg = gr.Textbox(label="Mensaje", placeholder="Escribe tu pregunta...")
+            clear = gr.ClearButton([msg, chatbot])
+        with gr.Column(scale=1):
+            image_input = gr.Image(type="pil", label="Sube Imagen")
+            info = gr.Textbox(label="Info Imagen", interactive=False)
+
+    msg.submit(process_message, [msg, chatbot, image_input], [msg, chatbot])
+    image_input.change(lambda img: f"Tamaño: {img.size}" if img else "Sin imagen.", [image_input], [info])
 
-demo = gr.Interface(fn=greet, inputs="text", outputs="text")
 demo.launch()