Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
@@ -1,150 +1,19 @@
|
|
1 |
-
import os
|
2 |
-
import time
|
3 |
-
import uuid
|
4 |
-
from PIL import Image
|
5 |
-
import google.generativeai as genai
|
6 |
import gradio as gr
|
7 |
-
|
8 |
-
from typing import List, Tuple, Optional, Union
|
9 |
-
|
10 |
-
# Cargar las variables de entorno
|
11 |
-
load_dotenv()
|
12 |
-
API_KEY = os.getenv("GOOGLE_API_KEY")
|
13 |
-
|
14 |
-
if not API_KEY:
|
15 |
-
raise ValueError("La clave de API 'GOOGLE_API_KEY' no est谩 configurada en el archivo .env")
|
16 |
-
|
17 |
-
# Configuraci贸n del modelo Gemini
|
18 |
-
generation_config = {
|
19 |
-
"temperature": 1,
|
20 |
-
"top_p": 0.95,
|
21 |
-
"top_k": 40,
|
22 |
-
"max_output_tokens": 8192,
|
23 |
-
"response_mime_type": "text/plain",
|
24 |
-
}
|
25 |
-
|
26 |
-
genai.configure(api_key=API_KEY)
|
27 |
-
|
28 |
-
model = genai.GenerativeModel(
|
29 |
-
model_name="gemini-1.5-flash",
|
30 |
-
generation_config=generation_config,
|
31 |
-
)
|
32 |
-
|
33 |
-
# Inicializar la sesi贸n de chat
|
34 |
-
chat = model.start_chat(history=[])
|
35 |
-
|
36 |
-
# Funci贸n para transformar el historial de Gradio al formato de Gemini
|
37 |
-
def transform_history(history):
|
38 |
-
new_history = []
|
39 |
-
for chat in history:
|
40 |
-
new_history.append({"parts": [{"text": chat[0]}], "role": "user"})
|
41 |
-
new_history.append({"parts": [{"text": chat[1]}], "role": "model"})
|
42 |
-
return new_history
|
43 |
-
|
44 |
-
# Funci贸n de respuesta que maneja el texto y los archivos multimodales
|
45 |
-
def response(message, history):
|
46 |
-
global chat
|
47 |
-
|
48 |
-
# Transformar el historial al formato esperado por Gemini
|
49 |
-
chat.history = transform_history(history)
|
50 |
-
|
51 |
-
# Enviar el mensaje al modelo y obtener la respuesta
|
52 |
-
response = chat.send_message(message["text"])
|
53 |
-
# Eliminar la llamada a `response.resolve()` porque no es necesario.
|
54 |
-
return response.text
|
55 |
-
|
56 |
-
# Constantes y configuraciones
|
57 |
-
IMAGE_CACHE_DIRECTORY = "/tmp"
|
58 |
-
IMAGE_WIDTH = 512
|
59 |
-
CHAT_HISTORY = List[Tuple[Optional[Union[Tuple[str], str]], Optional[str]]]
|
60 |
-
|
61 |
-
# Funci贸n para preprocesar la imagen
|
62 |
-
def preprocess_image(image: Image.Image) -> Optional[Image.Image]:
|
63 |
-
if image:
|
64 |
-
image_height = int(image.height * IMAGE_WIDTH / image.width)
|
65 |
-
return image.resize((IMAGE_WIDTH, image_height))
|
66 |
-
|
67 |
-
# Funci贸n para almacenar la imagen en el cach茅
|
68 |
-
def cache_pil_image(image: Image.Image) -> str:
|
69 |
-
image_filename = f"{uuid.uuid4()}.jpeg"
|
70 |
-
os.makedirs(IMAGE_CACHE_DIRECTORY, exist_ok=True)
|
71 |
-
image_path = os.path.join(IMAGE_CACHE_DIRECTORY, image_filename)
|
72 |
-
image.save(image_path, "JPEG")
|
73 |
-
return image_path
|
74 |
-
|
75 |
-
# Funci贸n para cargar im谩genes
|
76 |
-
def upload(files: Optional[List[str]], chatbot: CHAT_HISTORY) -> CHAT_HISTORY:
|
77 |
-
for file in files:
|
78 |
-
image = Image.open(file).convert('RGB')
|
79 |
-
image_preview = preprocess_image(image)
|
80 |
-
if image_preview:
|
81 |
-
# Enviar la imagen procesada para su visualizaci贸n en Gradio
|
82 |
-
gr.Image(image_preview).render()
|
83 |
-
image_path = cache_pil_image(image)
|
84 |
-
chatbot.append(((image_path,), None))
|
85 |
-
return chatbot
|
86 |
-
|
87 |
-
# Funci贸n para manejar el mensaje del usuario
|
88 |
-
def user(text_prompt: str, chatbot: CHAT_HISTORY):
|
89 |
-
if text_prompt:
|
90 |
-
chatbot.append((text_prompt, None))
|
91 |
-
return "", chatbot
|
92 |
-
|
93 |
-
# Funci贸n para la respuesta del bot
|
94 |
-
def bot(
|
95 |
-
files: Optional[List[str]],
|
96 |
-
model_choice: str,
|
97 |
-
system_instruction: Optional[str], # Sistema de instrucciones opcional
|
98 |
-
chatbot: CHAT_HISTORY
|
99 |
-
):
|
100 |
-
if not API_KEY:
|
101 |
-
raise ValueError("GOOGLE_API_KEY is not set.")
|
102 |
-
|
103 |
-
genai.configure(api_key=API_KEY)
|
104 |
-
generation_config = genai.types.GenerationConfig(
|
105 |
-
temperature=0.7,
|
106 |
-
max_output_tokens=8192,
|
107 |
-
top_k=10,
|
108 |
-
top_p=0.9
|
109 |
-
)
|
110 |
-
|
111 |
-
# Usar el valor por defecto para system_instruction si est谩 vac铆o
|
112 |
-
if not system_instruction:
|
113 |
-
system_instruction = "1" # O puedes poner un valor predeterminado como "No system instruction provided."
|
114 |
-
|
115 |
-
text_prompt = [chatbot[-1][0]] if chatbot and chatbot[-1][0] and isinstance(chatbot[-1][0], str) else []
|
116 |
-
image_prompt = [preprocess_image(Image.open(file).convert('RGB')) for file in files] if files else []
|
117 |
-
|
118 |
-
model = genai.GenerativeModel(
|
119 |
-
model_name=model_choice,
|
120 |
-
generation_config=generation_config,
|
121 |
-
system_instruction=system_instruction # Usar el valor por defecto si est谩 vac铆o
|
122 |
-
)
|
123 |
-
|
124 |
-
# Se debe usar la generaci贸n de contenido multimodal para procesar im谩genes y texto
|
125 |
-
response = model.generate_content(text_prompt + image_prompt, stream=True, generation_config=generation_config)
|
126 |
|
127 |
-
|
128 |
-
|
129 |
-
|
130 |
-
|
131 |
-
|
132 |
-
time.sleep(0.01)
|
133 |
-
yield chatbot
|
134 |
|
135 |
-
# Crear la interfaz de Gradio
|
136 |
demo = gr.ChatInterface(
|
137 |
-
|
138 |
-
|
139 |
-
|
|
|
|
|
140 |
],
|
141 |
-
multimodal=True, # Activar la modalidad multimodal
|
142 |
-
textbox=gr.MultimodalTextbox( # Configuraci贸n del cuadro de texto multimodal
|
143 |
-
file_count="multiple", # Permitir m煤ltiples archivos
|
144 |
-
file_types=["image"], # Aceptar solo im谩genes
|
145 |
-
sources=["upload", "microphone"] # Fuentes de entrada: carga de archivos y micr贸fono
|
146 |
-
)
|
147 |
)
|
148 |
|
149 |
-
# Iniciar la interfaz
|
150 |
demo.launch()
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
2 |
+
import time
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
3 |
|
4 |
+
def echo(message, history, system_prompt, tokens):
|
5 |
+
response = f"System prompt: {system_prompt}\n Message: {message}."
|
6 |
+
for i in range(min(len(response), int(tokens))):
|
7 |
+
time.sleep(0.05)
|
8 |
+
yield response[: i + 1]
|
|
|
|
|
9 |
|
|
|
10 |
demo = gr.ChatInterface(
|
11 |
+
echo,
|
12 |
+
type="messages",
|
13 |
+
additional_inputs=[
|
14 |
+
gr.Textbox("You are helpful AI.", label="System Prompt"),
|
15 |
+
gr.Slider(10, 100),
|
16 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
17 |
)
|
18 |
|
|
|
19 |
demo.launch()
|