|
|
|
|
|
import os |
|
import re |
|
import tempfile |
|
import gc |
|
from collections.abc import Iterator |
|
from threading import Thread |
|
import json |
|
import requests |
|
import cv2 |
|
import base64 |
|
import logging |
|
import time |
|
from urllib.parse import quote |
|
|
|
import gradio as gr |
|
import spaces |
|
import torch |
|
from loguru import logger |
|
from PIL import Image |
|
from transformers import AutoProcessor, Gemma3ForConditionalGeneration, TextIteratorStreamer |
|
|
|
|
|
import pandas as pd |
|
import PyPDF2 |
|
|
|
|
|
|
|
|
|
from gradio_client import Client |
|
|
|
API_URL = "http://211.233.58.201:7896" |
|
|
|
logging.basicConfig( |
|
level=logging.DEBUG, |
|
format='%(asctime)s - %(levelname)s - %(message)s' |
|
) |
|
|
|
def test_api_connection() -> str: |
|
"""API ์๋ฒ ์ฐ๊ฒฐ ํ
์คํธ""" |
|
try: |
|
client = Client(API_URL) |
|
return "API ์ฐ๊ฒฐ ์ฑ๊ณต: ์ ์ ์๋ ์ค" |
|
except Exception as e: |
|
logging.error(f"API ์ฐ๊ฒฐ ํ
์คํธ ์คํจ: {e}") |
|
return f"API ์ฐ๊ฒฐ ์คํจ: {e}" |
|
|
|
def generate_image(prompt: str, width: float, height: float, guidance: float, inference_steps: float, seed: float): |
|
"""์ด๋ฏธ์ง ์์ฑ ํจ์ (๋ฐํ ํ์์ ์ ์ฐํ๊ฒ ๋์)""" |
|
if not prompt: |
|
return None, "์ค๋ฅ: ํ๋กฌํํธ๊ฐ ํ์ํฉ๋๋ค." |
|
try: |
|
logging.info(f"ํ๋กฌํํธ๋ฅผ ์ฌ์ฉํ์ฌ ์ด๋ฏธ์ง ์์ฑ API ํธ์ถ: {prompt}") |
|
|
|
client = Client(API_URL) |
|
result = client.predict( |
|
prompt=prompt, |
|
width=int(width), |
|
height=int(height), |
|
guidance=float(guidance), |
|
inference_steps=int(inference_steps), |
|
seed=int(seed), |
|
do_img2img=False, |
|
init_image=None, |
|
image2image_strength=0.8, |
|
resize_img=True, |
|
api_name="/generate_image" |
|
) |
|
|
|
logging.info(f"์ด๋ฏธ์ง ์์ฑ ๊ฒฐ๊ณผ: {type(result)}, ๊ธธ์ด: {len(result) if isinstance(result, (list, tuple)) else '์ ์ ์์'}") |
|
|
|
|
|
if isinstance(result, (list, tuple)) and len(result) > 0: |
|
image_data = result[0] |
|
seed_info = result[1] if len(result) > 1 else "์ ์ ์๋ ์๋" |
|
return image_data, seed_info |
|
else: |
|
|
|
return result, "์ ์ ์๋ ์๋" |
|
|
|
except Exception as e: |
|
logging.error(f"์ด๋ฏธ์ง ์์ฑ ์คํจ: {str(e)}") |
|
return None, f"์ค๋ฅ: {str(e)}" |
|
|
|
|
|
def fix_base64_padding(data): |
|
"""Base64 ๋ฌธ์์ด์ ํจ๋ฉ์ ์์ ํฉ๋๋ค.""" |
|
if isinstance(data, bytes): |
|
data = data.decode('utf-8') |
|
|
|
|
|
if "base64," in data: |
|
data = data.split("base64,", 1)[1] |
|
|
|
|
|
missing_padding = len(data) % 4 |
|
if missing_padding: |
|
data += '=' * (4 - missing_padding) |
|
|
|
return data |
|
|
|
|
|
|
|
|
|
def clear_cuda_cache(): |
|
"""CUDA ์บ์๋ฅผ ๋ช
์์ ์ผ๋ก ๋น์๋๋ค.""" |
|
if torch.cuda.is_available(): |
|
torch.cuda.empty_cache() |
|
gc.collect() |
|
|
|
|
|
|
|
|
|
SERPHOUSE_API_KEY = os.getenv("SERPHOUSE_API_KEY", "") |
|
|
|
def extract_keywords(text: str, top_k: int = 5) -> str: |
|
"""๋จ์ ํค์๋ ์ถ์ถ: ํ๊ธ, ์์ด, ์ซ์, ๊ณต๋ฐฑ๋ง ๋จ๊น""" |
|
text = re.sub(r"[^a-zA-Z0-9๊ฐ-ํฃ\s]", "", text) |
|
tokens = text.split() |
|
return " ".join(tokens[:top_k]) |
|
|
|
def do_web_search(query: str) -> str: |
|
"""SerpHouse LIVE API ํธ์ถํ์ฌ ๊ฒ์ ๊ฒฐ๊ณผ ๋งํฌ๋ค์ด ๋ฐํ""" |
|
try: |
|
url = "https://api.serphouse.com/serp/live" |
|
params = { |
|
"q": query, |
|
"domain": "google.com", |
|
"serp_type": "web", |
|
"device": "desktop", |
|
"lang": "en", |
|
"num": "20" |
|
} |
|
headers = {"Authorization": f"Bearer {SERPHOUSE_API_KEY}"} |
|
logger.info(f"SerpHouse API ํธ์ถ ์ค... ๊ฒ์์ด: {query}") |
|
response = requests.get(url, headers=headers, params=params, timeout=60) |
|
response.raise_for_status() |
|
data = response.json() |
|
results = data.get("results", {}) |
|
organic = None |
|
if isinstance(results, dict) and "organic" in results: |
|
organic = results["organic"] |
|
elif isinstance(results, dict) and "results" in results: |
|
if isinstance(results["results"], dict) and "organic" in results["results"]: |
|
organic = results["results"]["organic"] |
|
elif "organic" in data: |
|
organic = data["organic"] |
|
if not organic: |
|
logger.warning("์๋ต์์ organic ๊ฒฐ๊ณผ๋ฅผ ์ฐพ์ ์ ์์ต๋๋ค.") |
|
return "์น ๊ฒ์ ๊ฒฐ๊ณผ๊ฐ ์๊ฑฐ๋ API ์๋ต ๊ตฌ์กฐ๊ฐ ์์๊ณผ ๋ค๋ฆ
๋๋ค." |
|
max_results = min(20, len(organic)) |
|
limited_organic = organic[:max_results] |
|
summary_lines = [] |
|
for idx, item in enumerate(limited_organic, start=1): |
|
title = item.get("title", "์ ๋ชฉ ์์") |
|
link = item.get("link", "#") |
|
snippet = item.get("snippet", "์ค๋ช
์์") |
|
displayed_link = item.get("displayed_link", link) |
|
summary_lines.append( |
|
f"### ๊ฒฐ๊ณผ {idx}: {title}\n\n" |
|
f"{snippet}\n\n" |
|
f"**์ถ์ฒ**: [{displayed_link}]({link})\n\n" |
|
f"---\n" |
|
) |
|
instructions = """ |
|
# ์น ๊ฒ์ ๊ฒฐ๊ณผ |
|
์๋๋ ๊ฒ์ ๊ฒฐ๊ณผ์
๋๋ค. ์ง๋ฌธ์ ๋ต๋ณํ ๋ ์ด ์ ๋ณด๋ฅผ ํ์ฉํ์ธ์: |
|
1. ๊ฐ ๊ฒฐ๊ณผ์ ์ ๋ชฉ, ๋ด์ฉ, ์ถ์ฒ ๋งํฌ๋ฅผ ์ฐธ๊ณ ํ์ธ์. |
|
2. ๋ต๋ณ์ ๊ด๋ จ ์ ๋ณด์ ์ถ์ฒ๋ฅผ ๋ช
์์ ์ผ๋ก ์ธ์ฉํ์ธ์ (์: "[์ถ์ฒ ์ ๋ชฉ](๋งํฌ)"). |
|
3. ์๋ต์ ์ค์ ์ถ์ฒ ๋งํฌ๋ฅผ ํฌํจํ์ธ์. |
|
4. ์ฌ๋ฌ ์ถ์ฒ์ ์ ๋ณด๋ฅผ ์ข
ํฉํ์ฌ ๋ต๋ณํ์ธ์. |
|
5. ๋ง์ง๋ง์ "์ฐธ๊ณ ์๋ฃ:" ์น์
์ ์ถ๊ฐํ๊ณ ์ฃผ์ ์ถ์ฒ ๋งํฌ๋ฅผ ๋์ดํ์ธ์. |
|
""" |
|
return instructions + "\n".join(summary_lines) |
|
except Exception as e: |
|
logger.error(f"์น ๊ฒ์ ์คํจ: {e}") |
|
return f"์น ๊ฒ์ ์คํจ: {str(e)}" |
|
|
|
|
|
|
|
|
|
MAX_CONTENT_CHARS = 2000 |
|
MAX_INPUT_LENGTH = 2096 |
|
model_id = os.getenv("MODEL_ID", "VIDraft/Gemma-3-R1984-4B") |
|
processor = AutoProcessor.from_pretrained(model_id, padding_side="left") |
|
model = Gemma3ForConditionalGeneration.from_pretrained( |
|
model_id, |
|
device_map="auto", |
|
torch_dtype=torch.bfloat16, |
|
attn_implementation="eager" |
|
) |
|
MAX_NUM_IMAGES = int(os.getenv("MAX_NUM_IMAGES", "5")) |
|
|
|
|
|
|
|
|
|
def analyze_csv_file(path: str) -> str: |
|
try: |
|
df = pd.read_csv(path) |
|
if df.shape[0] > 50 or df.shape[1] > 10: |
|
df = df.iloc[:50, :10] |
|
df_str = df.to_string() |
|
if len(df_str) > MAX_CONTENT_CHARS: |
|
df_str = df_str[:MAX_CONTENT_CHARS] + "\n...(์ผ๋ถ ์๋ต)..." |
|
return f"**[CSV ํ์ผ: {os.path.basename(path)}]**\n\n{df_str}" |
|
except Exception as e: |
|
return f"CSV ํ์ผ ์ฝ๊ธฐ ์คํจ ({os.path.basename(path)}): {str(e)}" |
|
|
|
def analyze_txt_file(path: str) -> str: |
|
try: |
|
with open(path, "r", encoding="utf-8") as f: |
|
text = f.read() |
|
if len(text) > MAX_CONTENT_CHARS: |
|
text = text[:MAX_CONTENT_CHARS] + "\n...(์ผ๋ถ ์๋ต)..." |
|
return f"**[TXT ํ์ผ: {os.path.basename(path)}]**\n\n{text}" |
|
except Exception as e: |
|
return f"TXT ํ์ผ ์ฝ๊ธฐ ์คํจ ({os.path.basename(path)}): {str(e)}" |
|
|
|
def pdf_to_markdown(pdf_path: str) -> str: |
|
text_chunks = [] |
|
try: |
|
with open(pdf_path, "rb") as f: |
|
reader = PyPDF2.PdfReader(f) |
|
max_pages = min(5, len(reader.pages)) |
|
for page_num in range(max_pages): |
|
page_text = reader.pages[page_num].extract_text() or "" |
|
page_text = page_text.strip() |
|
if page_text: |
|
if len(page_text) > MAX_CONTENT_CHARS // max_pages: |
|
page_text = page_text[:MAX_CONTENT_CHARS // max_pages] + "...(์ผ๋ถ ์๋ต)" |
|
text_chunks.append(f"## ํ์ด์ง {page_num+1}\n\n{page_text}\n") |
|
if len(reader.pages) > max_pages: |
|
text_chunks.append(f"\n...(์ ์ฒด {len(reader.pages)}ํ์ด์ง ์ค {max_pages}ํ์ด์ง๋ง ํ์)...") |
|
except Exception as e: |
|
return f"PDF ํ์ผ ์ฝ๊ธฐ ์คํจ ({os.path.basename(pdf_path)}): {str(e)}" |
|
full_text = "\n".join(text_chunks) |
|
if len(full_text) > MAX_CONTENT_CHARS: |
|
full_text = full_text[:MAX_CONTENT_CHARS] + "\n...(์ผ๋ถ ์๋ต)..." |
|
return f"**[PDF ํ์ผ: {os.path.basename(pdf_path)}]**\n\n{full_text}" |
|
|
|
|
|
|
|
|
|
def count_files_in_new_message(paths: list[str]) -> tuple[int, int]: |
|
image_count = 0 |
|
video_count = 0 |
|
for path in paths: |
|
if path.endswith(".mp4"): |
|
video_count += 1 |
|
elif re.search(r"\.(png|jpg|jpeg|gif|webp)$", path, re.IGNORECASE): |
|
image_count += 1 |
|
return image_count, video_count |
|
|
|
def count_files_in_history(history: list[dict]) -> tuple[int, int]: |
|
image_count = 0 |
|
video_count = 0 |
|
for item in history: |
|
if item["role"] != "user" or isinstance(item["content"], str): |
|
continue |
|
if isinstance(item["content"], list) and len(item["content"]) > 0: |
|
file_path = item["content"][0] |
|
if isinstance(file_path, str): |
|
if file_path.endswith(".mp4"): |
|
video_count += 1 |
|
elif re.search(r"\.(png|jpg|jpeg|gif|webp)$", file_path, re.IGNORECASE): |
|
image_count += 1 |
|
return image_count, video_count |
|
|
|
def validate_media_constraints(message: dict, history: list[dict]) -> bool: |
|
media_files = [f for f in message["files"] if re.search(r"\.(png|jpg|jpeg|gif|webp)$", f, re.IGNORECASE) or f.endswith(".mp4")] |
|
new_image_count, new_video_count = count_files_in_new_message(media_files) |
|
history_image_count, history_video_count = count_files_in_history(history) |
|
image_count = history_image_count + new_image_count |
|
video_count = history_video_count + new_video_count |
|
if video_count > 1: |
|
gr.Warning("๋น๋์ค ํ์ผ์ ํ๋๋ง ์ง์๋ฉ๋๋ค.") |
|
return False |
|
if video_count == 1: |
|
if image_count > 0: |
|
gr.Warning("์ด๋ฏธ์ง์ ๋น๋์ค๋ฅผ ํผํฉํ๋ ๊ฒ์ ํ์ฉ๋์ง ์์ต๋๋ค.") |
|
return False |
|
if "<image>" in message["text"]: |
|
gr.Warning("<image> ํ๊ทธ์ ๋น๋์ค ํ์ผ์ ํจ๊ป ์ฌ์ฉํ ์ ์์ต๋๋ค.") |
|
return False |
|
if video_count == 0 and image_count > MAX_NUM_IMAGES: |
|
gr.Warning(f"์ต๋ {MAX_NUM_IMAGES}์ฅ์ ์ด๋ฏธ์ง๋ฅผ ์
๋ก๋ํ ์ ์์ต๋๋ค.") |
|
return False |
|
if "<image>" in message["text"]: |
|
image_files = [f for f in message["files"] if re.search(r"\.(png|jpg|jpeg|gif|webp)$", f, re.IGNORECASE)] |
|
image_tag_count = message["text"].count("<image>") |
|
if image_tag_count != len(image_files): |
|
gr.Warning("ํ
์คํธ์ ์๋ <image> ํ๊ทธ์ ๊ฐ์๊ฐ ์ด๋ฏธ์ง ํ์ผ ๊ฐ์์ ์ผ์นํ์ง ์์ต๋๋ค.") |
|
return False |
|
return True |
|
|
|
|
|
|
|
|
|
def downsample_video(video_path: str) -> list[tuple[Image.Image, float]]: |
|
vidcap = cv2.VideoCapture(video_path) |
|
fps = vidcap.get(cv2.CAP_PROP_FPS) |
|
total_frames = int(vidcap.get(cv2.CAP_PROP_FRAME_COUNT)) |
|
frame_interval = max(int(fps), int(total_frames / 10)) |
|
frames = [] |
|
for i in range(0, total_frames, frame_interval): |
|
vidcap.set(cv2.CAP_PROP_POS_FRAMES, i) |
|
success, image = vidcap.read() |
|
if success: |
|
image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB) |
|
image = cv2.resize(image, (0, 0), fx=0.5, fy=0.5) |
|
pil_image = Image.fromarray(image) |
|
timestamp = round(i / fps, 2) |
|
frames.append((pil_image, timestamp)) |
|
if len(frames) >= 5: |
|
break |
|
vidcap.release() |
|
return frames |
|
|
|
def process_video(video_path: str) -> tuple[list[dict], list[str]]: |
|
content = [] |
|
temp_files = [] |
|
frames = downsample_video(video_path) |
|
for pil_image, timestamp in frames: |
|
with tempfile.NamedTemporaryFile(delete=False, suffix=".png") as temp_file: |
|
pil_image.save(temp_file.name) |
|
temp_files.append(temp_file.name) |
|
content.append({"type": "text", "text": f"ํ๋ ์ {timestamp}:"}) |
|
content.append({"type": "image", "url": temp_file.name}) |
|
return content, temp_files |
|
|
|
|
|
|
|
|
|
def process_interleaved_images(message: dict) -> list[dict]: |
|
parts = re.split(r"(<image>)", message["text"]) |
|
content = [] |
|
image_files = [f for f in message["files"] if re.search(r"\.(png|jpg|jpeg|gif|webp)$", f, re.IGNORECASE)] |
|
image_index = 0 |
|
for part in parts: |
|
if part == "<image>" and image_index < len(image_files): |
|
content.append({"type": "image", "url": image_files[image_index]}) |
|
image_index += 1 |
|
elif part.strip(): |
|
content.append({"type": "text", "text": part.strip()}) |
|
else: |
|
if isinstance(part, str) and part != "<image>": |
|
content.append({"type": "text", "text": part}) |
|
return content |
|
|
|
|
|
|
|
|
|
def is_image_file(file_path: str) -> bool: |
|
return bool(re.search(r"\.(png|jpg|jpeg|gif|webp)$", file_path, re.IGNORECASE)) |
|
|
|
def is_video_file(file_path: str) -> bool: |
|
return file_path.endswith(".mp4") |
|
|
|
def is_document_file(file_path: str) -> bool: |
|
return file_path.lower().endswith(".pdf") or file_path.lower().endswith(".csv") or file_path.lower().endswith(".txt") |
|
|
|
def process_new_user_message(message: dict) -> tuple[list[dict], list[str]]: |
|
temp_files = [] |
|
if not message["files"]: |
|
return [{"type": "text", "text": message["text"]}], temp_files |
|
video_files = [f for f in message["files"] if is_video_file(f)] |
|
image_files = [f for f in message["files"] if is_image_file(f)] |
|
csv_files = [f for f in message["files"] if f.lower().endswith(".csv")] |
|
txt_files = [f for f in message["files"] if f.lower().endswith(".txt")] |
|
pdf_files = [f for f in message["files"] if f.lower().endswith(".pdf")] |
|
content_list = [{"type": "text", "text": message["text"]}] |
|
for csv_path in csv_files: |
|
content_list.append({"type": "text", "text": analyze_csv_file(csv_path)}) |
|
for txt_path in txt_files: |
|
content_list.append({"type": "text", "text": analyze_txt_file(txt_path)}) |
|
for pdf_path in pdf_files: |
|
content_list.append({"type": "text", "text": pdf_to_markdown(pdf_path)}) |
|
if video_files: |
|
video_content, video_temp_files = process_video(video_files[0]) |
|
content_list += video_content |
|
temp_files.extend(video_temp_files) |
|
return content_list, temp_files |
|
if "<image>" in message["text"] and image_files: |
|
interleaved_content = process_interleaved_images({"text": message["text"], "files": image_files}) |
|
if content_list and content_list[0]["type"] == "text": |
|
content_list = content_list[1:] |
|
return interleaved_content + content_list, temp_files |
|
else: |
|
for img_path in image_files: |
|
content_list.append({"type": "image", "url": img_path}) |
|
return content_list, temp_files |
|
|
|
|
|
|
|
|
|
def process_history(history: list[dict]) -> list[dict]: |
|
messages = [] |
|
current_user_content = [] |
|
for item in history: |
|
if item["role"] == "assistant": |
|
if current_user_content: |
|
messages.append({"role": "user", "content": current_user_content}) |
|
current_user_content = [] |
|
messages.append({"role": "assistant", "content": [{"type": "text", "text": item["content"]}]}) |
|
else: |
|
content = item["content"] |
|
if isinstance(content, str): |
|
current_user_content.append({"type": "text", "text": content}) |
|
elif isinstance(content, list) and len(content) > 0: |
|
file_path = content[0] |
|
if is_image_file(file_path): |
|
current_user_content.append({"type": "image", "url": file_path}) |
|
else: |
|
current_user_content.append({"type": "text", "text": f"[ํ์ผ: {os.path.basename(file_path)}]"}) |
|
if current_user_content: |
|
messages.append({"role": "user", "content": current_user_content}) |
|
return messages |
|
|
|
|
|
|
|
|
|
def _model_gen_with_oom_catch(**kwargs): |
|
try: |
|
model.generate(**kwargs) |
|
except torch.cuda.OutOfMemoryError: |
|
raise RuntimeError("[OutOfMemoryError] GPU ๋ฉ๋ชจ๋ฆฌ๊ฐ ๋ถ์กฑํฉ๋๋ค.") |
|
finally: |
|
clear_cuda_cache() |
|
|
|
|
|
|
|
|
|
@spaces.GPU(duration=120) |
|
def run( |
|
message: dict, |
|
history: list[dict], |
|
system_prompt: str = "", |
|
max_new_tokens: int = 512, |
|
use_web_search: bool = False, |
|
web_search_query: str = "", |
|
age_group: str = "20๋", |
|
mbti_personality: str = "INTP", |
|
sexual_openness: int = 2, |
|
image_gen: bool = False |
|
) -> Iterator[str]: |
|
if not validate_media_constraints(message, history): |
|
yield "" |
|
return |
|
temp_files = [] |
|
try: |
|
|
|
persona = ( |
|
f"{system_prompt.strip()}\n\n" |
|
f"์ฑ๋ณ: ์ฌ์ฑ\n" |
|
f"์ฐ๋ น๋: {age_group}\n" |
|
f"MBTI ํ๋ฅด์๋: {mbti_personality}\n" |
|
f"์น์์ผ ๊ฐ๋ฐฉ์ฑ (1~5): {sexual_openness}\n" |
|
) |
|
combined_system_msg = f"[์์คํ
ํ๋กฌํํธ]\n{persona.strip()}\n\n" |
|
|
|
if use_web_search: |
|
user_text = message["text"] |
|
ws_query = extract_keywords(user_text) |
|
if ws_query.strip(): |
|
logger.info(f"[์๋ ์น ๊ฒ์ ํค์๋] {ws_query!r}") |
|
ws_result = do_web_search(ws_query) |
|
combined_system_msg += f"[๊ฒ์ ๊ฒฐ๊ณผ (์์ 20๊ฐ ํญ๋ชฉ)]\n{ws_result}\n\n" |
|
combined_system_msg += ( |
|
"[์ฐธ๊ณ : ์ ๊ฒ์ ๊ฒฐ๊ณผ ๋งํฌ๋ฅผ ์ถ์ฒ๋ก ์ธ์ฉํ์ฌ ๋ต๋ณ]\n" |
|
"[์ค์ ์ง์์ฌํญ]\n" |
|
"1. ๋ต๋ณ์ ๊ฒ์ ๊ฒฐ๊ณผ์์ ์ฐพ์ ์ ๋ณด์ ์ถ์ฒ๋ฅผ ๋ฐ๋์ ์ธ์ฉํ์ธ์.\n" |
|
"2. ์ถ์ฒ ์ธ์ฉ ์ \"[์ถ์ฒ ์ ๋ชฉ](๋งํฌ)\" ํ์์ ๋งํฌ๋ค์ด ๋งํฌ๋ฅผ ์ฌ์ฉํ์ธ์.\n" |
|
"3. ์ฌ๋ฌ ์ถ์ฒ์ ์ ๋ณด๋ฅผ ์ข
ํฉํ์ฌ ๋ต๋ณํ์ธ์.\n" |
|
"4. ๋ต๋ณ ๋ง์ง๋ง์ \"์ฐธ๊ณ ์๋ฃ:\" ์น์
์ ์ถ๊ฐํ๊ณ ์ฌ์ฉํ ์ฃผ์ ์ถ์ฒ ๋งํฌ๋ฅผ ๋์ดํ์ธ์.\n" |
|
) |
|
else: |
|
combined_system_msg += "[์ ํจํ ํค์๋๊ฐ ์์ด ์น ๊ฒ์์ ๊ฑด๋๋๋๋ค]\n\n" |
|
messages = [] |
|
if combined_system_msg.strip(): |
|
messages.append({"role": "system", "content": [{"type": "text", "text": combined_system_msg.strip()}]}) |
|
messages.extend(process_history(history)) |
|
user_content, user_temp_files = process_new_user_message(message) |
|
temp_files.extend(user_temp_files) |
|
for item in user_content: |
|
if item["type"] == "text" and len(item["text"]) > MAX_CONTENT_CHARS: |
|
item["text"] = item["text"][:MAX_CONTENT_CHARS] + "\n...(์ผ๋ถ ์๋ต)..." |
|
messages.append({"role": "user", "content": user_content}) |
|
inputs = processor.apply_chat_template( |
|
messages, |
|
add_generation_prompt=True, |
|
tokenize=True, |
|
return_dict=True, |
|
return_tensors="pt", |
|
).to(device=model.device, dtype=torch.bfloat16) |
|
if inputs.input_ids.shape[1] > MAX_INPUT_LENGTH: |
|
inputs.input_ids = inputs.input_ids[:, -MAX_INPUT_LENGTH:] |
|
if 'attention_mask' in inputs: |
|
inputs.attention_mask = inputs.attention_mask[:, -MAX_INPUT_LENGTH:] |
|
streamer = TextIteratorStreamer(processor, timeout=30.0, skip_prompt=True, skip_special_tokens=True) |
|
gen_kwargs = dict(inputs, streamer=streamer, max_new_tokens=max_new_tokens) |
|
t = Thread(target=_model_gen_with_oom_catch, kwargs=gen_kwargs) |
|
t.start() |
|
output_so_far = "" |
|
for new_text in streamer: |
|
output_so_far += new_text |
|
yield output_so_far |
|
|
|
except Exception as e: |
|
logger.error(f"run ํจ์ ์๋ฌ: {str(e)}") |
|
yield f"์ฃ์กํฉ๋๋ค. ์ค๋ฅ๊ฐ ๋ฐ์ํ์ต๋๋ค: {str(e)}" |
|
finally: |
|
for tmp in temp_files: |
|
try: |
|
if os.path.exists(tmp): |
|
os.unlink(tmp) |
|
logger.info(f"์์ ํ์ผ ์ญ์ ๋จ: {tmp}") |
|
except Exception as ee: |
|
logger.warning(f"์์ ํ์ผ {tmp} ์ญ์ ์คํจ: {ee}") |
|
try: |
|
del inputs, streamer |
|
except Exception: |
|
pass |
|
clear_cuda_cache() |
|
|
|
|
|
def modified_run(message, history, system_prompt, max_new_tokens, use_web_search, web_search_query, |
|
age_group, mbti_personality, sexual_openness, image_gen): |
|
|
|
output_so_far = "" |
|
gallery_update = gr.Gallery(visible=False, value=[]) |
|
yield output_so_far, gallery_update |
|
|
|
|
|
text_generator = run(message, history, system_prompt, max_new_tokens, use_web_search, |
|
web_search_query, age_group, mbti_personality, sexual_openness, image_gen) |
|
|
|
for text_chunk in text_generator: |
|
output_so_far = text_chunk |
|
yield output_so_far, gallery_update |
|
|
|
|
|
if image_gen and message["text"].strip(): |
|
try: |
|
width, height = 512, 512 |
|
guidance, steps, seed = 7.5, 30, 42 |
|
|
|
logger.info(f"๊ฐค๋ฌ๋ฆฌ์ฉ ์ด๋ฏธ์ง ์์ฑ ํธ์ถ, ํ๋กฌํํธ: {message['text']}") |
|
|
|
|
|
image_result, seed_info = generate_image( |
|
prompt=message["text"].strip(), |
|
width=width, |
|
height=height, |
|
guidance=guidance, |
|
inference_steps=steps, |
|
seed=seed |
|
) |
|
|
|
if image_result: |
|
|
|
if isinstance(image_result, str) and ( |
|
image_result.startswith('data:') or |
|
len(image_result) > 100 and '/' not in image_result |
|
): |
|
|
|
try: |
|
|
|
if image_result.startswith('data:'): |
|
content_type, b64data = image_result.split(';base64,') |
|
else: |
|
b64data = image_result |
|
content_type = "image/webp" |
|
|
|
|
|
image_bytes = base64.b64decode(b64data) |
|
|
|
|
|
with tempfile.NamedTemporaryFile(delete=False, suffix=".webp") as temp_file: |
|
temp_file.write(image_bytes) |
|
temp_path = temp_file.name |
|
|
|
|
|
gallery_update = gr.Gallery(visible=True, value=[temp_path]) |
|
yield output_so_far + "\n\n*์ด๋ฏธ์ง๊ฐ ์์ฑ๋์ด ์๋ ๊ฐค๋ฌ๋ฆฌ์ ํ์๋ฉ๋๋ค.*", gallery_update |
|
|
|
except Exception as e: |
|
logger.error(f"Base64 ์ด๋ฏธ์ง ์ฒ๋ฆฌ ์ค๋ฅ: {e}") |
|
yield output_so_far + f"\n\n(์ด๋ฏธ์ง ์ฒ๋ฆฌ ์ค ์ค๋ฅ: {e})", gallery_update |
|
|
|
|
|
elif isinstance(image_result, str) and os.path.exists(image_result): |
|
|
|
gallery_update = gr.Gallery(visible=True, value=[image_result]) |
|
yield output_so_far + "\n\n*์ด๋ฏธ์ง๊ฐ ์์ฑ๋์ด ์๋ ๊ฐค๋ฌ๋ฆฌ์ ํ์๋ฉ๋๋ค.*", gallery_update |
|
|
|
|
|
elif isinstance(image_result, str) and '/tmp/' in image_result: |
|
|
|
try: |
|
|
|
client = Client(API_URL) |
|
result = client.predict( |
|
prompt=message["text"].strip(), |
|
api_name="/generate_base64_image" |
|
) |
|
|
|
if isinstance(result, str) and (result.startswith('data:') or len(result) > 100): |
|
|
|
if result.startswith('data:'): |
|
content_type, b64data = result.split(';base64,') |
|
else: |
|
b64data = result |
|
|
|
|
|
image_bytes = base64.b64decode(b64data) |
|
|
|
|
|
with tempfile.NamedTemporaryFile(delete=False, suffix=".webp") as temp_file: |
|
temp_file.write(image_bytes) |
|
temp_path = temp_file.name |
|
|
|
|
|
gallery_update = gr.Gallery(visible=True, value=[temp_path]) |
|
yield output_so_far + "\n\n*์ด๋ฏธ์ง๊ฐ ์์ฑ๋์ด ์๋ ๊ฐค๋ฌ๋ฆฌ์ ํ์๋ฉ๋๋ค.*", gallery_update |
|
else: |
|
yield output_so_far + "\n\n(์ด๋ฏธ์ง ์์ฑ ์คํจ: ์ฌ๋ฐ๋ฅธ ํ์์ด ์๋๋๋ค)", gallery_update |
|
|
|
except Exception as e: |
|
logger.error(f"๋์ฒด API ํธ์ถ ์ค ์ค๋ฅ: {e}") |
|
yield output_so_far + f"\n\n(์ด๋ฏธ์ง ์์ฑ ์คํจ: {e})", gallery_update |
|
|
|
|
|
elif isinstance(image_result, str) and ( |
|
image_result.startswith('http://') or |
|
image_result.startswith('https://') |
|
): |
|
try: |
|
|
|
response = requests.get(image_result, timeout=10) |
|
response.raise_for_status() |
|
|
|
|
|
with tempfile.NamedTemporaryFile(delete=False, suffix=".webp") as temp_file: |
|
temp_file.write(response.content) |
|
temp_path = temp_file.name |
|
|
|
|
|
gallery_update = gr.Gallery(visible=True, value=[temp_path]) |
|
yield output_so_far + "\n\n*์ด๋ฏธ์ง๊ฐ ์์ฑ๋์ด ์๋ ๊ฐค๋ฌ๋ฆฌ์ ํ์๋ฉ๋๋ค.*", gallery_update |
|
|
|
except Exception as e: |
|
logger.error(f"URL ์ด๋ฏธ์ง ๋ค์ด๋ก๋ ์ค๋ฅ: {e}") |
|
yield output_so_far + f"\n\n(์ด๋ฏธ์ง ๋ค์ด๋ก๋ ์ค ์ค๋ฅ: {e})", gallery_update |
|
|
|
|
|
elif hasattr(image_result, 'save'): |
|
try: |
|
with tempfile.NamedTemporaryFile(delete=False, suffix=".webp") as temp_file: |
|
image_result.save(temp_file.name) |
|
temp_path = temp_file.name |
|
|
|
|
|
gallery_update = gr.Gallery(visible=True, value=[temp_path]) |
|
yield output_so_far + "\n\n*์ด๋ฏธ์ง๊ฐ ์์ฑ๋์ด ์๋ ๊ฐค๋ฌ๋ฆฌ์ ํ์๋ฉ๋๋ค.*", gallery_update |
|
|
|
except Exception as e: |
|
logger.error(f"์ด๋ฏธ์ง ๊ฐ์ฒด ์ ์ฅ ์ค๋ฅ: {e}") |
|
yield output_so_far + f"\n\n(์ด๋ฏธ์ง ๊ฐ์ฒด ์ ์ฅ ์ค ์ค๋ฅ: {e})", gallery_update |
|
|
|
else: |
|
|
|
yield output_so_far + f"\n\n(์ง์๋์ง ์๋ ์ด๋ฏธ์ง ํ์: {type(image_result)})", gallery_update |
|
else: |
|
yield output_so_far + f"\n\n(์ด๋ฏธ์ง ์์ฑ ์คํจ: {seed_info})", gallery_update |
|
|
|
except Exception as e: |
|
logger.error(f"๊ฐค๋ฌ๋ฆฌ์ฉ ์ด๋ฏธ์ง ์์ฑ ์ค ์ค๋ฅ: {e}") |
|
yield output_so_far + f"\n\n(์ด๋ฏธ์ง ์์ฑ ์ค ์ค๋ฅ: {e})", gallery_update |
|
|
|
|
|
|
|
|
|
examples = [ |
|
[ |
|
{ |
|
"text": "๋ PDF ํ์ผ์ ๋ด์ฉ์ ๋น๊ตํ์ธ์.", |
|
"files": [ |
|
"assets/additional-examples/before.pdf", |
|
"assets/additional-examples/after.pdf", |
|
], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "CSV ํ์ผ์ ๋ด์ฉ์ ์์ฝ ๋ฐ ๋ถ์ํ์ธ์.", |
|
"files": ["assets/additional-examples/sample-csv.csv"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์น์ ํ๊ณ ์ดํด์ฌ ๋ง์ ์ฌ์์น๊ตฌ ์ญํ ์ ๋งก์ผ์ธ์. ์ด ์์์ ์ค๋ช
ํด ์ฃผ์ธ์.", |
|
"files": ["assets/additional-examples/tmp.mp4"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "ํ์ง๋ฅผ ์ค๋ช
ํ๊ณ ๊ทธ ์์ ๊ธ์จ๋ฅผ ์ฝ์ด ์ฃผ์ธ์.", |
|
"files": ["assets/additional-examples/maz.jpg"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ ๋ ์ด๋ฏธ ์ด ๋ณด์ถฉ์ ๋ฅผ ๊ฐ์ง๊ณ ์๊ณ <image> ์ด ์ ํ๋ ๊ตฌ๋งคํ ๊ณํ์
๋๋ค. ํจ๊ป ๋ณต์ฉํ ๋ ์ฃผ์ํ ์ ์ด ์๋์?", |
|
"files": [ |
|
"assets/additional-examples/pill1.png", |
|
"assets/additional-examples/pill2.png" |
|
], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ด ์ ๋ถ ๋ฌธ์ ๋ฅผ ํ์ด ์ฃผ์ธ์.", |
|
"files": ["assets/additional-examples/4.png"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ด ํฐ์ผ์ ์ธ์ ๋ฐํ๋์๊ณ , ๊ฐ๊ฒฉ์ ์ผ๋ง์ธ๊ฐ์?", |
|
"files": ["assets/additional-examples/2.png"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ด ์ด๋ฏธ์ง๋ค์ ์์๋ฅผ ๋ฐํ์ผ๋ก ์งง์ ์ด์ผ๊ธฐ๋ฅผ ๋ง๋ค์ด ์ฃผ์ธ์.", |
|
"files": [ |
|
"assets/sample-images/09-1.png", |
|
"assets/sample-images/09-2.png", |
|
"assets/sample-images/09-3.png", |
|
"assets/sample-images/09-4.png", |
|
"assets/sample-images/09-5.png", |
|
], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ด ์ด๋ฏธ์ง์ ์ผ์นํ๋ ๋ง๋ ์ฐจํธ๋ฅผ ๊ทธ๋ฆฌ๊ธฐ ์ํ matplotlib๋ฅผ ์ฌ์ฉํ๋ Python ์ฝ๋๋ฅผ ์์ฑํด ์ฃผ์ธ์.", |
|
"files": ["assets/additional-examples/barchart.png"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ด๋ฏธ์ง์ ํ
์คํธ๋ฅผ ์ฝ๊ณ Markdown ํ์์ผ๋ก ์์ฑํด ์ฃผ์ธ์.", |
|
"files": ["assets/additional-examples/3.png"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ด ํ์งํ์ ๋ฌด์จ ๊ธ์๊ฐ ์ฐ์ฌ ์๋์?", |
|
"files": ["assets/sample-images/02.png"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "๋ ์ด๋ฏธ์ง๋ฅผ ๋น๊ตํ๊ณ ์ ์ฌ์ ๊ณผ ์ฐจ์ด์ ์ ์ค๋ช
ํด ์ฃผ์ธ์.", |
|
"files": ["assets/sample-images/03.png"], |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "๋กคํ๋ ์ด ํด๋ด
์๋ค. ๋น์ ์ ์ ์ ๋ ์์๊ฐ๊ณ ์ถ์ ์๋ก์ด ์จ๋ผ์ธ ๋ฐ์ดํธ ์๋์
๋๋ค. ๋ค์ ํ๊ณ ๋ฐฐ๋ ค ๊น์ ๋ฐฉ์์ผ๋ก ์๊ธฐ ์๊ฐ๋ฅผ ํด์ฃผ์ธ์!", |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "ํด๋ณ์ ๊ฑท๋ ๋ ๋ฒ์งธ ๋ฐ์ดํธ ์ค์
๋๋ค. ์ฅ๋์ค๋ฌ์ด ๋ํ์ ๋ถ๋๋ฌ์ด ํ๋ฌํ
์ผ๋ก ์ฅ๋ฉด์ ์ด์ด๋๊ฐ ์ฃผ์ธ์.", |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์ข์ํ๋ ์ฌ๋์๊ฒ ๋ฉ์์ง๋ฅผ ๋ณด๋ด๋ ๊ฒ์ด ๋ถ์ํฉ๋๋ค. ๊ฒฉ๋ ค์ ๋ง์ด๋ ์ ๊ทผ ๋ฐฉ๋ฒ์ ๋ํ ์ ์์ ํด์ค ์ ์๋์?", |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "๊ด๊ณ์์ ์ด๋ ค์์ ๊ทน๋ณตํ ๋ ์ฌ๋์ ๋ํ ๋ก๋งจํฑํ ์ด์ผ๊ธฐ๋ฅผ ๋ค๋ ค์ฃผ์ธ์.", |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์์ ์ธ ๋ฐฉ์์ผ๋ก ์ฌ๋์ ํํํ๊ณ ์ถ์ต๋๋ค. ์ ํํธ๋๋ฅผ ์ํ ์ง์ฌ์ด ๋ด๊ธด ์๋ฅผ ์์ฑํ๋ ๋ฐ ๋์์ ์ค ์ ์๋์?", |
|
} |
|
], |
|
[ |
|
{ |
|
"text": "์์ ๋คํผ์ด ์์์ต๋๋ค. ์ง์ฌ์ผ๋ก ์ฌ๊ณผํ๋ฉด์ ์ ๊ฐ์ ์ ํํํ ์ ์๋ ๋ฐฉ๋ฒ์ ์ฐพ์์ฃผ์ธ์.", |
|
} |
|
], |
|
] |
|
|
|
|
|
|
|
|
|
|
|
|
|
css = """ |
|
.gradio-container { |
|
background: rgba(255, 255, 255, 0.7); |
|
padding: 30px 40px; |
|
margin: 20px auto; |
|
width: 100% !important; |
|
max-width: none !important; |
|
} |
|
""" |
|
title_html = """ |
|
<h1 align="center" style="margin-bottom: 0.2em; font-size: 1.6em;"> ๐ HeartSync ๐ </h1> |
|
<p align="center" style="font-size:1.1em; color:#555;"> |
|
โ
FLUX ์ด๋ฏธ์ง ์์ฑ โ
์ถ๋ก โ
๊ฒ์ด ํด์ โ
๋ฉํฐ๋ชจ๋ฌ & VLM โ
์ค์๊ฐ ์น ๊ฒ์ โ
RAG <br> |
|
</p> |
|
""" |
|
|
|
with gr.Blocks(css=css, title="HeartSync") as demo: |
|
gr.Markdown(title_html) |
|
|
|
|
|
generated_images = gr.Gallery( |
|
label="์์ฑ๋ ์ด๋ฏธ์ง", |
|
show_label=True, |
|
visible=False, |
|
elem_id="generated_images", |
|
columns=2, |
|
height="auto", |
|
object_fit="contain" |
|
) |
|
|
|
with gr.Row(): |
|
web_search_checkbox = gr.Checkbox(label="์ฌ๋ ์๋ ์ฐ๊ตฌ", value=False) |
|
image_gen_checkbox = gr.Checkbox(label="์ด๋ฏธ์ง ์์ฑ", value=False) |
|
|
|
base_system_prompt_box = gr.Textbox( |
|
lines=3, |
|
value="๋น์ ์ ๊น์ด ์ฌ๊ณ ํ๋ AI์
๋๋ค. ํญ์ ๋
ผ๋ฆฌ์ ์ด๊ณ ์ฐฝ์์ ์ผ๋ก ๋ฌธ์ ๋ฅผ ํด๊ฒฐํฉ๋๋ค.\nํ๋ฅด์๋: ๋น์ ์ ๋ค์ ํ๊ณ ์ฌ๋์ด ๋์น๋ ์ฌ์์น๊ตฌ์
๋๋ค.", |
|
label="๊ธฐ๋ณธ ์์คํ
ํ๋กฌํํธ", |
|
visible=False |
|
) |
|
with gr.Row(): |
|
age_group_dropdown = gr.Dropdown( |
|
label="์ฐ๋ น๋ ์ ํ (๊ธฐ๋ณธ 20๋)", |
|
choices=["10๋", "20๋", "30~40๋", "50~60๋", "70๋ ์ด์"], |
|
value="20๋", |
|
interactive=True |
|
) |
|
mbti_choices = [ |
|
"INTJ (์ฉ์์ฃผ๋ํ ์ ๋ต๊ฐ)", |
|
"INTP (๋
ผ๋ฆฌ์ ์ธ ์ฌ์๊ฐ)", |
|
"ENTJ (๋๋ดํ ํต์์)", |
|
"ENTP (๋จ๊ฑฐ์ด ๋
ผ์๊ฐ)", |
|
"INFJ (์ ์์ ์นํธ์)", |
|
"INFP (์ด์ ์ ์ธ ์ค์ฌ์)", |
|
"ENFJ (์ ์๋ก์ด ์ฌํ์ด๋๊ฐ)", |
|
"ENFP (์ฌ๊ธฐ๋ฐ๋ํ ํ๋๊ฐ)", |
|
"ISTJ (์ฒญ๋ ด๊ฒฐ๋ฐฑํ ๋
ผ๋ฆฌ์ฃผ์์)", |
|
"ISFJ (์ฉ๊ฐํ ์ํธ์)", |
|
"ESTJ (์๊ฒฉํ ๊ด๋ฆฌ์)", |
|
"ESFJ (์ฌ๊ต์ ์ธ ์ธ๊ต๊ด)", |
|
"ISTP (๋ง๋ฅ ์ฌ์ฃผ๊พผ)", |
|
"ISFP (ํธ๊ธฐ์ฌ ๋ง์ ์์ ๊ฐ)", |
|
"ESTP (๋ชจํ์ ์ฆ๊ธฐ๋ ์ฌ์
๊ฐ)", |
|
"ESFP (์์ ๋ก์ด ์ํผ์ ์ฐ์์ธ)" |
|
] |
|
mbti_dropdown = gr.Dropdown( |
|
label="AI ํ๋ฅด์๋ MBTI (๊ธฐ๋ณธ INTP)", |
|
choices=mbti_choices, |
|
value="INTP (๋
ผ๋ฆฌ์ ์ธ ์ฌ์๊ฐ)", |
|
interactive=True |
|
) |
|
sexual_openness_slider = gr.Slider( |
|
minimum=1, maximum=5, step=1, value=2, |
|
label="์น์์ผ ๊ด์ฌ๋/๊ฐ๋ฐฉ์ฑ (1~5, ๊ธฐ๋ณธ=2)", |
|
interactive=True |
|
) |
|
max_tokens_slider = gr.Slider( |
|
label="์ต๋ ์์ฑ ํ ํฐ ์", |
|
minimum=100, maximum=8000, step=50, value=1000, |
|
visible=False |
|
) |
|
web_search_text = gr.Textbox( |
|
lines=1, |
|
label="์น ๊ฒ์ ์ฟผ๋ฆฌ (๋ฏธ์ฌ์ฉ)", |
|
placeholder="์ง์ ์
๋ ฅํ ํ์ ์์", |
|
visible=False |
|
) |
|
|
|
|
|
chat = gr.ChatInterface( |
|
fn=modified_run, |
|
type="messages", |
|
chatbot=gr.Chatbot(type="messages", scale=1, allow_tags=["image"]), |
|
textbox=gr.MultimodalTextbox( |
|
file_types=[".webp", ".png", ".jpg", ".jpeg", ".gif", ".mp4", ".csv", ".txt", ".pdf"], |
|
file_count="multiple", |
|
autofocus=True |
|
), |
|
multimodal=True, |
|
additional_inputs=[ |
|
base_system_prompt_box, |
|
max_tokens_slider, |
|
web_search_checkbox, |
|
web_search_text, |
|
age_group_dropdown, |
|
mbti_dropdown, |
|
sexual_openness_slider, |
|
image_gen_checkbox, |
|
], |
|
additional_outputs=[ |
|
generated_images, |
|
], |
|
stop_btn=False, |
|
title='<a href="https://discord.gg/openfreeai" target="_blank">https://discord.gg/openfreeai</a>', |
|
examples=examples, |
|
run_examples_on_click=False, |
|
cache_examples=False, |
|
css_paths=None, |
|
delete_cache=(1800, 1800), |
|
) |
|
|
|
|
|
with gr.Row(elem_id="examples_row"): |
|
with gr.Column(scale=12, elem_id="examples_container"): |
|
gr.Markdown("### ์์ ์
๋ ฅ (ํด๋ฆญํ์ฌ ๋ถ๋ฌ์ค๊ธฐ)") |
|
|
|
if __name__ == "__main__": |
|
demo.launch(share=True) |
|
|