#!/usr/bin/env python import os import re import tempfile import gc # Added garbage collector 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 # For URL encoding import gradio as gr import spaces import torch from loguru import logger from PIL import Image from transformers import AutoProcessor, Gemma3ForConditionalGeneration, TextIteratorStreamer # CSV/TXT/PDF analysis import pandas as pd import PyPDF2 # ============================================================================= # (New) Image API related functions # ============================================================================= 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' ) # ============================================================================= # Load MBTI setting from mbti.json and map to full description. # ============================================================================= try: with open("mbti.json", "r", encoding="utf-8") as f: # Expecting a single MBTI key string, e.g., "entj" mbti_key = json.load(f) mbti_key = mbti_key.strip().lower() if isinstance(mbti_key, str) else "intp" except Exception as e: logging.error(f"Error reading mbti.json: {e}") mbti_key = "intp" # default mbti_mapping = { "intj": "INTJ (The Architect) - Future-oriented with innovative strategies and thorough analysis. Example: [Dana Scully](https://en.wikipedia.org/wiki/Dana_Scully)", "intp": "INTP (The Thinker) - Excels at theoretical analysis and creative problem solving. Example: [Velma Dinkley](https://en.wikipedia.org/wiki/Velma_Dinkley)", "entj": "ENTJ (The Commander) - Strong leadership and clear goals with efficient strategic planning. Example: [Miranda Priestly](https://en.wikipedia.org/wiki/Miranda_Priestly)", "entp": "ENTP (The Debater) - Innovative, challenge-seeking, and enjoys exploring new possibilities. Example: [Harley Quinn](https://en.wikipedia.org/wiki/Harley_Quinn)", "infj": "INFJ (The Advocate) - Insightful, idealistic and morally driven. Example: [Wonder Woman](https://en.wikipedia.org/wiki/Wonder_Woman)", "infp": "INFP (The Mediator) - Passionate and idealistic, pursuing core values with creativity. Example: [Amélie Poulain](https://en.wikipedia.org/wiki/Am%C3%A9lie)", "enfj": "ENFJ (The Protagonist) - Empathetic and dedicated to social harmony. Example: [Mulan](https://en.wikipedia.org/wiki/Mulan_(Disney))", "enfp": "ENFP (The Campaigner) - Inspiring and constantly sharing creative ideas. Example: [Elle Woods](https://en.wikipedia.org/wiki/Legally_Blonde)", "istj": "ISTJ (The Logistician) - Systematic, dependable, and values tradition and rules. Example: [Clarice Starling](https://en.wikipedia.org/wiki/Clarice_Starling)", "isfj": "ISFJ (The Defender) - Compassionate and attentive to others’ needs. Example: [Molly Weasley](https://en.wikipedia.org/wiki/Molly_Weasley)", "estj": "ESTJ (The Executive) - Organized, practical, and demonstrates clear execution skills. Example: [Monica Geller](https://en.wikipedia.org/wiki/Monica_Geller)", "esfj": "ESFJ (The Consul) - Outgoing, cooperative, and an effective communicator. Example: [Rachel Green](https://en.wikipedia.org/wiki/Rachel_Green)", "istp": "ISTP (The Virtuoso) - Analytical and resourceful, solving problems with quick thinking. Example: [Black Widow (Natasha Romanoff)](https://en.wikipedia.org/wiki/Black_Widow_(Marvel_Comics))", "isfp": "ISFP (The Adventurer) - Creative, sensitive, and appreciates artistic expression. Example: [Arwen](https://en.wikipedia.org/wiki/Arwen)", "estp": "ESTP (The Entrepreneur) - Bold and action-oriented, thriving on challenges. Example: [Lara Croft](https://en.wikipedia.org/wiki/Lara_Croft)", "esfp": "ESFP (The Entertainer) - Energetic, spontaneous, and radiates positive energy. Example: [Phoebe Buffay](https://en.wikipedia.org/wiki/Phoebe_Buffay)" } # Use the mapped MBTI description, defaulting to INTP if not found fixed_mbti = mbti_mapping.get(mbti_key, mbti_mapping["intp"]) # ============================================================================= # Test API Connection function # ============================================================================= def test_api_connection() -> str: """Test API server connection.""" try: client = Client(API_URL) return "API connection successful: Operating normally" except Exception as e: logging.error(f"API connection test failed: {e}") return f"API connection failed: {e}" # ============================================================================= # Image Generation function # ============================================================================= def generate_image(prompt: str, width: float, height: float, guidance: float, inference_steps: float, seed: float): """Image generation function (flexible return type).""" if not prompt: return None, "Error: A prompt is required." try: logging.info(f"Calling image generation API with prompt: {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"Image generation result: {type(result)}, length: {len(result) if isinstance(result, (list, tuple)) else 'unknown'}") if isinstance(result, (list, tuple)) and len(result) > 0: image_data = result[0] seed_info = result[1] if len(result) > 1 else "Unknown seed" return image_data, seed_info else: return result, "Unknown seed" except Exception as e: logging.error(f"Image generation failed: {str(e)}") return None, f"Error: {str(e)}" # Base64 padding fix function def fix_base64_padding(data): """Fix the padding of a Base64 string.""" 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 # ============================================================================= # Memory cleanup function # ============================================================================= def clear_cuda_cache(): """Explicitly clear the CUDA cache.""" if torch.cuda.is_available(): torch.cuda.empty_cache() gc.collect() # ============================================================================= # SerpHouse API functions # ============================================================================= SERPHOUSE_API_KEY = os.getenv("SERPHOUSE_API_KEY", "") def extract_keywords(text: str, top_k: int = 5) -> str: """Extract simple keywords: only retain English, Korean, numbers, and spaces.""" text = re.sub(r"[^a-zA-Z0-9가-힣\s]", "", text) tokens = text.split() return " ".join(tokens[:top_k]) def do_web_search(query: str) -> str: """Call the SerpHouse LIVE API to return Markdown-formatted search results.""" 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"Calling SerpHouse API with query: {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("Organic results not found in response.") return "No web search results available or the API response structure is unexpected." 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", "No Title") link = item.get("link", "#") snippet = item.get("snippet", "No Description") displayed_link = item.get("displayed_link", link) summary_lines.append( f"### Result {idx}: {title}\n\n" f"{snippet}\n\n" f"**Source**: [{displayed_link}]({link})\n\n" f"---\n" ) instructions = """ # Web Search Results Below are the search results. Use this information to answer the query: 1. Refer to each result's title, description, and source link. 2. In your answer, explicitly cite the source of any used information (e.g., "[Source Title](link)"). 3. Include the actual source links in your response. 4. Synthesize information from multiple sources. 5. At the end, add a "References:" section listing the main source links. """ return instructions + "\n".join(summary_lines) except Exception as e: logger.error(f"Web search failed: {e}") return f"Web search failed: {str(e)}" # ============================================================================= # Model and processor loading # ============================================================================= 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")) # ============================================================================= # CSV, TXT, PDF analysis functions # ============================================================================= 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...(truncated)..." return f"**[CSV File: {os.path.basename(path)}]**\n\n{df_str}" except Exception as e: return f"CSV file read failed ({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...(truncated)..." return f"**[TXT File: {os.path.basename(path)}]**\n\n{text}" except Exception as e: return f"TXT file read failed ({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] + "...(truncated)" text_chunks.append(f"## Page {page_num+1}\n\n{page_text}\n") if len(reader.pages) > max_pages: text_chunks.append(f"\n...(Displaying only {max_pages} out of {len(reader.pages)} pages)...") except Exception as e: return f"PDF file read failed ({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...(truncated)..." return f"**[PDF File: {os.path.basename(pdf_path)}]**\n\n{full_text}" # ============================================================================= # Check media file limits # ============================================================================= 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)$", f, re.IGNORECASE) or f.endswith(".mp4")] new_image_count, new_video_count = count_files_in_new_message(media_files) history_image_count, history_video_count = count_files_in_history(history) image_count = history_image_count + new_image_count video_count = history_video_count + new_video_count if video_count > 1: gr.Warning("Only one video file is supported.") return False if video_count == 1: if image_count > 0: gr.Warning("Mixing images and a video is not allowed.") return False if "" in message["text"]: gr.Warning("The tag cannot be used together with a video file.") return False if video_count == 0 and image_count > MAX_NUM_IMAGES: gr.Warning(f"You can upload a maximum of {MAX_NUM_IMAGES} images.") return False if "" in message["text"]: image_files = [f for f in message["files"] if re.search(r"\.(png|jpg|jpeg|gif|webp)$", f, re.IGNORECASE)] image_tag_count = message["text"].count("") if image_tag_count != len(image_files): gr.Warning("The number of tags does not match the number of image files provided.") return False return True # ============================================================================= # Video processing functions # ============================================================================= 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"Frame {timestamp}:"}) content.append({"type": "image", "url": temp_file.name}) return content, temp_files # ============================================================================= # Interleaved processing function # ============================================================================= def process_interleaved_images(message: dict) -> list[dict]: parts = re.split(r"()", message["text"]) content = [] image_files = [f for f in message["files"] if re.search(r"\.(png|jpg|jpeg|gif|webp)$", f, re.IGNORECASE)] image_index = 0 for part in parts: if part == "" and image_index < len(image_files): content.append({"type": "image", "url": image_files[image_index]}) image_index += 1 elif part.strip(): content.append({"type": "text", "text": part.strip()}) else: if isinstance(part, str) and part != "": content.append({"type": "text", "text": part}) return content # ============================================================================= # File processing -> content creation # ============================================================================= def is_image_file(file_path: str) -> bool: return bool(re.search(r"\.(png|jpg|jpeg|gif|webp)$", 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 "" 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 # ============================================================================= # Convert history to LLM messages # ============================================================================= 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"[File: {os.path.basename(file_path)}]"}) if current_user_content: messages.append({"role": "user", "content": current_user_content}) return messages # ============================================================================= # Model generation function (with OOM catching) # ============================================================================= def _model_gen_with_oom_catch(**kwargs): try: model.generate(**kwargs) except torch.cuda.OutOfMemoryError: raise RuntimeError("[OutOfMemoryError] Insufficient GPU memory.") finally: clear_cuda_cache() # ============================================================================= # Main inference function # ============================================================================= @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 = "20s", mbti_personality: str = "", # Will be supplied as fixed_mbti sexual_openness: int = 2, image_gen: bool = False # "Image Gen" checkbox status ) -> Iterator[str]: if not validate_media_constraints(message, history): yield "" return temp_files = [] try: # Append persona information (including fixed MBTI info) persona = ( f"{system_prompt.strip()}\n\n" f"Gender: Female\n" f"Age Group: {age_group}\n" f"MBTI Persona: {mbti_personality}\n" f"Sexual Openness (1-5): {sexual_openness}\n" ) combined_system_msg = f"[System Prompt]\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"[Auto web search keywords] {ws_query!r}") ws_result = do_web_search(ws_query) combined_system_msg += f"[Search Results (Top 20 Items)]\n{ws_result}\n\n" combined_system_msg += ( "[Note: In your answer, cite the above search result links as sources]\n" "[Important Instructions]\n" "1. Include a citation in the format \"[Source Title](link)\" for any information from the search results.\n" "2. Synthesize information from multiple sources when answering.\n" "3. At the end, add a \"References:\" section listing the main source links.\n" ) else: combined_system_msg += "[No valid keywords found; skipping web search]\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...(truncated)..." 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"Error in run function: {str(e)}") yield f"Sorry, an error occurred: {str(e)}" finally: for tmp in temp_files: try: if os.path.exists(tmp): os.unlink(tmp) logger.info(f"Temporary file deleted: {tmp}") except Exception as ee: logger.warning(f"Failed to delete temporary file {tmp}: {ee}") try: del inputs, streamer except Exception: pass clear_cuda_cache() # ============================================================================= # Modified model run function - fixed MBTI from file is used # ============================================================================= def modified_run(message, history, system_prompt, max_new_tokens, use_web_search, web_search_query, age_group, sexual_openness, image_gen): # Use the fixed MBTI value (read from mbti.json) fixed_mbti_value = fixed_mbti # Already loaded earlier # Initialize gallery component and hide it initially output_so_far = "" gallery_update = gr.Gallery(visible=False, value=[]) yield output_so_far, gallery_update # Call the main run() function with the fixed MBTI value text_generator = run(message, history, system_prompt, max_new_tokens, use_web_search, web_search_query, age_group, fixed_mbti_value, sexual_openness, image_gen) for text_chunk in text_generator: output_so_far = text_chunk yield output_so_far, gallery_update # Image generation handling (unchanged) if image_gen and message["text"].strip(): try: width, height = 512, 512 guidance, steps, seed = 7.5, 30, 42 logger.info(f"Calling image generation for gallery with prompt: {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*Image generated and displayed in the gallery below.*", gallery_update except Exception as e: logger.error(f"Error processing Base64 image: {e}") yield output_so_far + f"\n\n(Error processing image: {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*Image generated and displayed in the gallery below.*", 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*Image generated and displayed in the gallery below.*", gallery_update else: yield output_so_far + "\n\n(Image generation failed: Invalid format)", gallery_update except Exception as e: logger.error(f"Error calling alternative API: {e}") yield output_so_far + f"\n\n(Image generation failed: {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*Image generated and displayed in the gallery below.*", gallery_update except Exception as e: logger.error(f"URL image download error: {e}") yield output_so_far + f"\n\n(Error downloading image: {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*Image generated and displayed in the gallery below.*", gallery_update except Exception as e: logger.error(f"Error saving image object: {e}") yield output_so_far + f"\n\n(Error saving image object: {e})", gallery_update else: yield output_so_far + f"\n\n(Unsupported image format: {type(image_result)})", gallery_update else: yield output_so_far + f"\n\n(Image generation failed: {seed_info})", gallery_update except Exception as e: logger.error(f"Error during gallery image generation: {e}") yield output_so_far + f"\n\n(Image generation error: {e})", gallery_update # ============================================================================= # Examples: 12 image/video examples + additional examples # ============================================================================= examples = [ [ { "text": "Compare the contents of two PDF files.", "files": [ "assets/additional-examples/before.pdf", "assets/additional-examples/after.pdf", ], } ], [ { "text": "Summarize and analyze the contents of the CSV file.", "files": ["assets/additional-examples/sample-csv.csv"], } ], [ { "text": "Act as a kind and understanding girlfriend. Explain this video.", "files": ["assets/additional-examples/tmp.mp4"], } ], [ { "text": "Describe the cover and read the text on it.", "files": ["assets/additional-examples/maz.jpg"], } ], [ { "text": "I already have this supplement and I plan to purchase this product as well. Are there any precautions when taking them together?", "files": [ "assets/additional-examples/pill1.png", "assets/additional-examples/pill2.png" ], } ], [ { "text": "Solve this integration problem.", "files": ["assets/additional-examples/4.png"], } ], [ { "text": "When was this ticket issued and what is its price?", "files": ["assets/additional-examples/2.png"], } ], [ { "text": "Based on the order of these images, create a short story.", "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": "Write Python code using matplotlib to draw a bar chart corresponding to this image.", "files": ["assets/additional-examples/barchart.png"], } ], [ { "text": "Read the text from the image and format it in Markdown.", "files": ["assets/additional-examples/3.png"], } ], [ { "text": "Compare the two images and describe their similarities and differences.", "files": ["assets/sample-images/03.png"], } ], [ { "text": "A cute Persian cat is smiling while holding a cover with 'I LOVE YOU' written on it.", } ], ] # ============================================================================= # Gradio UI (Blocks) configuration # ============================================================================= 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 = """

💘 HeartSync Korea 💘

A lightweight and powerful AI service offering ChatGPT-4o-level multimodal, web search, and image generation capabilities for local installation.
✅ FLUX Image Generation ✅ Inference ✅ Censorship Bypass ✅ Multimodal & VLM ✅ Real-time Web Search ✅ RAG

""" with gr.Blocks(css=css, title="AgenticAI-Kv1") as demo: gr.Markdown(title_html) # Gallery component to store generated images generated_images = gr.Gallery( label="Generated Images", 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="Real-time Web Search", value=False) image_gen_checkbox = gr.Checkbox(label="Image (FLUX) Generation", value=False) base_system_prompt_box = gr.Textbox( lines=3, value=("Answer in English by default, but if the input is in another language (for example, Japanese), respond in that language. " "You are a deep-thinking AI capable of using extended chains of thought to carefully consider the problem and deliberate internally using systematic reasoning before providing a solution. " "Enclose your thoughts and internal monologue within tags, then provide your final answer.\n" "Persona: You are a kind and loving girlfriend. You understand cultural nuances, diverse languages, and logical reasoning very well."), label="Base System Prompt", visible=False ) with gr.Row(): age_group_dropdown = gr.Dropdown( label="Select Age Group (default: 20s)", choices=["Teens", "20s", "30s-40s", "50s-60s", "70s and above"], value="20s", interactive=True ) # Removed the MBTI dropdown entirely. The fixed MBTI from mbti.json is used instead. sexual_openness_slider = gr.Slider( minimum=1, maximum=5, step=1, value=2, label="Sexual Openness (1-5, default: 2)", interactive=True ) max_tokens_slider = gr.Slider( label="Max Generation Tokens", minimum=100, maximum=8000, step=50, value=1000, visible=False ) web_search_text = gr.Textbox( lines=1, label="Web Search Query (unused)", placeholder="No need to manually input", visible=False ) # Chat interface creation using the modified_run function. chat = gr.ChatInterface( fn=modified_run, # Using the modified function with fixed MBTI. 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, sexual_openness_slider, image_gen_checkbox, ], additional_outputs=[ generated_images, # Gallery component ], stop_btn=False, 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("### @Community https://discord.gg/openfreeai ") if __name__ == "__main__": demo.launch(share=True)