Spaces:
Running
Running
import os | |
import gradio as gr | |
from gradio import ChatMessage | |
from typing import Iterator, List, Dict, Tuple, Any | |
import google.generativeai as genai | |
from huggingface_hub import HfApi | |
import requests | |
import re | |
import traceback | |
# HuggingFace ๊ด๋ จ API ํค (์คํ์ด์ค ๋ถ์ ์ฉ) | |
HF_TOKEN = os.getenv("HF_TOKEN") | |
hf_api = HfApi(token=HF_TOKEN) | |
# Gemini 2.0 Flash Thinking ๋ชจ๋ธ ๊ด๋ จ API ํค ๋ฐ ํด๋ผ์ด์ธํธ (LLM ์ฉ) | |
GEMINI_API_KEY = os.getenv("GEMINI_API_KEY") | |
genai.configure(api_key=GEMINI_API_KEY) | |
model = genai.GenerativeModel("gemini-2.0-flash-thinking-exp-01-21") | |
def get_headers(): | |
if not HF_TOKEN: | |
raise ValueError("Hugging Face token not found in environment variables") | |
return {"Authorization": f"Bearer {HF_TOKEN}"} | |
def get_file_content(space_id: str, file_path: str) -> str: | |
file_url = f"https://huggingface.co/spaces/{space_id}/raw/main/{file_path}" | |
try: | |
response = requests.get(file_url, headers=get_headers()) | |
if response.status_code == 200: | |
return response.text | |
else: | |
return f"File not found or inaccessible: {file_path}" | |
except requests.RequestException: | |
return f"Error fetching content for file: {file_path}" | |
def get_space_structure(space_id: str) -> Dict: | |
try: | |
files = hf_api.list_repo_files(repo_id=space_id, repo_type="space") | |
tree = {"type": "directory", "path": "", "name": space_id, "children": []} | |
for file in files: | |
path_parts = file.split('/') | |
current = tree | |
for i, part in enumerate(path_parts): | |
if i == len(path_parts) - 1: # ํ์ผ | |
current["children"].append({"type": "file", "path": file, "name": part}) | |
else: | |
found = False | |
for child in current["children"]: | |
if child["type"] == "directory" and child["name"] == part: | |
current = child | |
found = True | |
break | |
if not found: | |
new_dir = {"type": "directory", "path": '/'.join(path_parts[:i+1]), "name": part, "children": []} | |
current["children"].append(new_dir) | |
current = new_dir | |
return tree | |
except Exception as e: | |
print(f"Error in get_space_structure: {str(e)}") | |
return {"error": f"API request error: {str(e)}"} | |
def format_tree_structure(tree_data: Dict, indent: str = "") -> str: | |
if "error" in tree_data: | |
return tree_data["error"] | |
formatted = f"{indent}{'๐' if tree_data.get('type') == 'directory' else '๐'} {tree_data.get('name', 'Unknown')}\n" | |
if tree_data.get("type") == "directory": | |
# ๋๋ ํ ๋ฆฌ๋ฅผ ๋จผ์ , ํ์ผ์ ๋์ค์ ํ์ํ๊ธฐ ์ํด ์ ๋ ฌ ์กฐ๊ฑด ์ฌ์ฉ | |
for child in sorted(tree_data.get("children", []), key=lambda x: (x.get("type", "") != "directory", x.get("name", ""))): | |
formatted += format_tree_structure(child, indent + " ") | |
return formatted | |
def analyze_space(url: str, progress=gr.Progress()): | |
""" | |
HuggingFace Space์ app.py์ ํ์ผ๊ตฌ์กฐ ๋ฑ์ ๋ถ๋ฌ์์: | |
1) ์ฝ๋ ์์ฝ | |
2) ์ฝ๋ ๋ถ์ | |
3) ์ฌ์ฉ๋ฒ | |
๋ฑ์ ๋ฐํํฉ๋๋ค. | |
""" | |
try: | |
space_id = url.split('spaces/')[-1] | |
if not re.match(r'^[\w.-]+/[\w.-]+$', space_id): | |
raise ValueError(f"Invalid Space ID format: {space_id}") | |
progress(0.1, desc="ํ์ผ ๊ตฌ์กฐ ๋ถ์ ์ค...") | |
tree_structure = get_space_structure(space_id) | |
if "error" in tree_structure: | |
raise ValueError(tree_structure["error"]) | |
tree_view = format_tree_structure(tree_structure) | |
progress(0.3, desc="app.py ๋ด์ฉ ๊ฐ์ ธ์ค๋ ์ค...") | |
app_content = get_file_content(space_id, "app.py") | |
progress(0.5, desc="์ฝ๋ ์์ฝ ์ค...") | |
summary = summarize_code(app_content) | |
progress(0.7, desc="์ฝ๋ ๋ถ์ ์ค...") | |
analysis = analyze_code(app_content) | |
progress(0.9, desc="์ฌ์ฉ๋ฒ ์ค๋ช ์์ฑ ์ค...") | |
usage = explain_usage(app_content) | |
# lines ์ ์กฐ์ | |
lines_for_app_py = adjust_lines_for_code(app_content) | |
progress(1.0, desc="์๋ฃ") | |
return app_content, tree_view, tree_structure, space_id, summary, analysis, usage, lines_for_app_py | |
except Exception as e: | |
print(f"Error in analyze_space: {str(e)}") | |
print(traceback.format_exc()) | |
return f"์ค๋ฅ๊ฐ ๋ฐ์ํ์ต๋๋ค: {str(e)}", "", None, "", "", "", "", 10 | |
def adjust_lines_for_code(code_content: str, min_lines: int = 10, max_lines: int = 100) -> int: | |
""" | |
์ฝ๋์ ์ค ์์ ๋ง์ถฐ ํ์ํ lines ์๋ฅผ ๋์ ์ผ๋ก ์กฐ์ ํฉ๋๋ค. | |
""" | |
num_lines = len(code_content.split('\n')) | |
return min(max(num_lines, min_lines), max_lines) | |
# -------------------------------------------------- | |
# Gemini 2.0 Flash Thinking ๋ชจ๋ธ ๊ด๋ จ (LLM) ํจ์๋ค | |
# -------------------------------------------------- | |
from gradio import ChatMessage | |
def format_chat_history(messages: List[ChatMessage]) -> List[Dict]: | |
""" | |
ChatMessage ๋ชฉ๋ก์ Gemini ๋ชจ๋ธ์ด ์ดํดํ ์ ์๋ ํ์์ผ๋ก ๋ณํ | |
(Thinking ๋ฉํ๋ฐ์ดํฐ ํฌํจ ๋ฉ์์ง๋ ๋ฌด์) | |
""" | |
formatted = [] | |
for m in messages: | |
# 'Thinking' metadata๊ฐ ์์ผ๋ฉด ๋ฌด์ | |
if hasattr(m, "metadata") and m.metadata: | |
continue | |
role = "assistant" if m.role == "assistant" else "user" | |
formatted.append({"role": role, "parts": [m.content or ""]}) | |
return formatted | |
def gemini_chat_completion(system_message: str, user_message: str, max_tokens: int = 200, temperature: float = 0.7) -> str: | |
""" | |
์์คํ ๋ฉ์์ง์ ์ ์ ๋ฉ์์ง๋ฅผ ๋ฐ์ Gemini์ ์คํธ๋ฆฌ๋ฐ ์์ฒญ, | |
์ต์ข ์๋ต ํ ์คํธ๋ฅผ ๋ฐํํฉ๋๋ค. | |
""" | |
init_msgs = [ | |
ChatMessage(role="system", content=system_message), | |
ChatMessage(role="user", content=user_message) | |
] | |
chat_history = format_chat_history(init_msgs) | |
chat = model.start_chat(history=chat_history) | |
final = "" | |
try: | |
for chunk in chat.send_message(user_message, stream=True): | |
parts = chunk.candidates[0].content.parts | |
if len(parts) == 2: | |
# Thinking + ์ต์ข ์๋ต | |
final += parts[1].text | |
else: | |
final += parts[0].text | |
return final.strip() | |
except Exception as e: | |
return f"LLM ํธ์ถ ์ค ์ค๋ฅ ๋ฐ์: {str(e)}" | |
def summarize_code(app_content: str): | |
system_msg = "๋น์ ์ Python ์ฝ๋๋ฅผ ๋ถ์ํ๊ณ ์์ฝํ๋ AI ์กฐ์์ ๋๋ค. ์ฃผ์ด์ง ์ฝ๋๋ฅผ 3์ค ์ด๋ด๋ก ๊ฐ๊ฒฐํ๊ฒ ์์ฝํด์ฃผ์ธ์." | |
user_msg = f"๋ค์ Python ์ฝ๋๋ฅผ 3์ค ์ด๋ด๋ก ์์ฝํด์ฃผ์ธ์:\n\n{app_content}" | |
try: | |
return gemini_chat_completion(system_msg, user_msg, max_tokens=200, temperature=0.7) | |
except Exception as e: | |
return f"์์ฝ ์์ฑ ์ค ์ค๋ฅ ๋ฐ์: {str(e)}" | |
def analyze_code(app_content: str): | |
# ์์คํ ํ๋กฌํํธ์ '๋ฅ์ฝํน' ์๋ด๋ฌธ ์ถ๊ฐ | |
system_msg = ( | |
"You are a deep thinking AI. You may use extremely long chains of thought to deeply consider the problem " | |
"and deliberate with yourself via systematic reasoning processes to help come to a correct solution prior to answering. " | |
"You should enclose your thoughts and internal monologue inside tags, and then provide your solution or response to the problem. " | |
"๋น์ ์ Python ์ฝ๋๋ฅผ ๋ถ์ํ๋ AI ์กฐ์์ ๋๋ค. ์ฃผ์ด์ง ์ฝ๋๋ฅผ ๋ถ์ํ์ฌ ์๋น์ค์ ํจ์ฉ์ฑ๊ณผ ํ์ฉ ์ธก๋ฉด์์ ๋ค์ ํญ๋ชฉ์ ๋ํด ์ค๋ช ํด์ฃผ์ธ์:\n" | |
"A. ๋ฐฐ๊ฒฝ ๋ฐ ํ์์ฑ\n" | |
"B. ๊ธฐ๋ฅ์ ํจ์ฉ์ฑ ๋ฐ ๊ฐ์น\n" | |
"C. ํน์ฅ์ \n" | |
"D. ์ ์ฉ ๋์ ๋ฐ ํ๊ฒ\n" | |
"E. ๊ธฐ๋ํจ๊ณผ\n" | |
"๊ธฐ์กด ๋ฐ ์ ์ฌ ํ๋ก์ ํธ์ ๋น๊ตํ์ฌ ๋ถ์ํด์ฃผ์ธ์. Markdown ํ์์ผ๋ก ์ถ๋ ฅํ์ธ์." | |
) | |
user_msg = f"๋ค์ Python ์ฝ๋๋ฅผ ๋ถ์ํด์ฃผ์ธ์:\n\n{app_content}" | |
try: | |
return gemini_chat_completion(system_msg, user_msg, max_tokens=1000, temperature=0.7) | |
except Exception as e: | |
return f"๋ถ์ ์์ฑ ์ค ์ค๋ฅ ๋ฐ์: {str(e)}" | |
def explain_usage(app_content: str): | |
# ์์คํ ํ๋กฌํํธ์ '๋ฅ์ฝํน' ์๋ด๋ฌธ ์ถ๊ฐ | |
system_msg = ( | |
"You are a deep thinking AI. You may use extremely long chains of thought to deeply consider the problem " | |
"and deliberate with yourself via systematic reasoning processes to help come to a correct solution prior to answering. " | |
"You should enclose your thoughts and internal monologue inside tags, and then provide your solution or response to the problem. " | |
"๋น์ ์ Python ์ฝ๋๋ฅผ ๋ถ์ํ์ฌ ์ฌ์ฉ๋ฒ์ ์ค๋ช ํ๋ AI ์กฐ์์ ๋๋ค. ์ฃผ์ด์ง ์ฝ๋๋ฅผ ๋ฐํ์ผ๋ก ๋ง์น ํ๋ฉด์ ๋ณด๋ ๊ฒ์ฒ๋ผ ์ฌ์ฉ๋ฒ์ ์์ธํ ์ค๋ช ํด์ฃผ์ธ์. Markdown ํ์์ผ๋ก ์ถ๋ ฅํ์ธ์." | |
) | |
user_msg = f"๋ค์ Python ์ฝ๋์ ์ฌ์ฉ๋ฒ์ ์ค๋ช ํด์ฃผ์ธ์:\n\n{app_content}" | |
try: | |
return gemini_chat_completion(system_msg, user_msg, max_tokens=800, temperature=0.7) | |
except Exception as e: | |
return f"์ฌ์ฉ๋ฒ ์ค๋ช ์์ฑ ์ค ์ค๋ฅ ๋ฐ์: {str(e)}" | |
def stream_gemini_response(user_message: str, conversation_state: List[ChatMessage]) -> Iterator[List[ChatMessage]]: | |
""" | |
conversation_state: ChatMessage ๊ฐ์ฒด๋ก๋ง ์ด๋ฃจ์ด์ง '๋ํ ์ด๋ ฅ' (Gradio State). | |
(์์ ) ๋น ๋ฌธ์์ด์ด์ด๋ ์ฒ๋ฆฌํ๋๋ก ๋ณ๊ฒฝ. ์๋ฌ๋ฅผ ๋์ฐ์ง ์์. | |
""" | |
# ๊ธฐ์กด์๋ if not user_message.strip(): ... return ํ์ผ๋, | |
# "Please provide a non-empty text message..." ์ค๋ฅ๊ฐ ๋ถํธํ๋ค๋ ์์ฒญ์ผ๋ก ์ ๊ฑฐ/์ํํจ. | |
# ํ์ํ๋ค๋ฉด user_message๊ฐ ์ ๋ง ์๋ฌด๊ฒ๋ ์์ ๋ ์ฒ๋ฆฌ ๋ก์ง์ ์ถ๊ฐํ์ธ์. | |
print(f"\n=== New Request ===\nUser message: {user_message if user_message.strip() else '(Empty)'}") | |
# ๊ธฐ์กด ๋ํ๋ฅผ Gemini ํ์์ผ๋ก ๋ณํ | |
chat_history = format_chat_history(conversation_state) | |
chat = model.start_chat(history=chat_history) | |
response = chat.send_message(user_message, stream=True) | |
thought_buffer = "" | |
response_buffer = "" | |
thinking_complete = False | |
# 'Thinking' ํ์์ฉ | |
conversation_state.append( | |
ChatMessage( | |
role="assistant", | |
content="", | |
metadata={"title": "โ๏ธ Thinking: *The thoughts produced by the model are experimental"} | |
) | |
) | |
try: | |
for chunk in response: | |
parts = chunk.candidates[0].content.parts | |
current_chunk = parts[0].text | |
if len(parts) == 2 and not thinking_complete: | |
thought_buffer += current_chunk | |
print(f"\n=== Complete Thought ===\n{thought_buffer}") | |
conversation_state[-1] = ChatMessage( | |
role="assistant", | |
content=thought_buffer, | |
metadata={"title": "โ๏ธ Thinking: *The thoughts produced by the model are experimental"} | |
) | |
yield conversation_state | |
response_buffer = parts[1].text | |
print(f"\n=== Starting Response ===\n{response_buffer}") | |
conversation_state.append( | |
ChatMessage(role="assistant", content=response_buffer) | |
) | |
thinking_complete = True | |
elif thinking_complete: | |
response_buffer += current_chunk | |
print(f"\n=== Response Chunk ===\n{current_chunk}") | |
conversation_state[-1] = ChatMessage( | |
role="assistant", | |
content=response_buffer | |
) | |
else: | |
thought_buffer += current_chunk | |
print(f"\n=== Thinking Chunk ===\n{current_chunk}") | |
conversation_state[-1] = ChatMessage( | |
role="assistant", | |
content=thought_buffer, | |
metadata={"title": "โ๏ธ Thinking: *The thoughts produced by the model are experimental"} | |
) | |
yield conversation_state | |
print(f"\n=== Final Response ===\n{response_buffer}") | |
except Exception as e: | |
print(f"\n=== Error ===\n{str(e)}") | |
conversation_state.append( | |
ChatMessage( | |
role="assistant", | |
content=f"I apologize, but I encountered an error: {str(e)}" | |
) | |
) | |
yield conversation_state | |
def convert_to_display_tuples(messages: List[ChatMessage]) -> List[Tuple[str, str]]: | |
""" | |
ํ๋ฉด์ ํ์ํ๊ธฐ ์ํด (user, assistant) ํํ ๋ชฉ๋ก์ผ๋ก ๋ณํ | |
""" | |
result = [] | |
i = 0 | |
while i < len(messages): | |
if messages[i].role == "user": | |
user_text = messages[i].content | |
assistant_text = "" | |
if i + 1 < len(messages) and messages[i+1].role == "assistant": | |
assistant_text = messages[i+1].content | |
i += 2 | |
else: | |
i += 1 | |
result.append((user_text, assistant_text)) | |
else: | |
# assistant ๋จ๋ | |
result.append(("", messages[i].content)) | |
i += 1 | |
return result | |
def user_submit_message(msg: str, conversation_state: List[ChatMessage]): | |
""" | |
์ฌ์ฉ์๊ฐ ๋ฉ์์ง๋ฅผ ์ ๋ ฅํ ๋ ํธ์ถ. | |
ChatMessage ๋ฆฌ์คํธ(conversation_state)์ user ๋ฉ์์ง๋ฅผ ์ถ๊ฐํ ๋ค ๋ฐํ. | |
""" | |
conversation_state.append(ChatMessage(role="user", content=msg)) | |
# ์ ๋ ฅ์ฐฝ์ ๋น์์ค | |
return "", conversation_state | |
def respond_wrapper(message: str, conversation_state: List[ChatMessage], max_tokens, temperature, top_p): | |
""" | |
์ ์ ๋ฉ์์ง๋ฅผ ๋ฐ์ Gemini์๊ฒ ์์ฒญ(์คํธ๋ฆฌ๋ฐ)ํ๊ณ , ๋ํ ์ด๋ ฅ์ ์ ๋ฐ์ดํธ ํ | |
ํ๋ฉด์๋ (user, assistant) ํํ์ ๋ฐํํ๋ค. | |
""" | |
for updated_messages in stream_gemini_response(message, conversation_state): | |
# ํ๋ฉด ํ์์ฉ (user, assistant) ํํ๋ก ๋ณํ | |
yield "", convert_to_display_tuples(updated_messages) | |
def create_ui(): | |
""" | |
Gradio UI๋ฅผ ๊ตฌ์ฑํ๋ ํจ์ | |
""" | |
try: | |
css = """ | |
footer {visibility: hidden;} | |
""" | |
with gr.Blocks(css=css) as demo: | |
gr.Markdown("# MOUSE: Space Research Thinking") | |
with gr.Tabs(): | |
with gr.TabItem("๋ถ์"): | |
with gr.Row(): | |
with gr.Column(): | |
url_input = gr.Textbox(label="HuggingFace Space URL") | |
analyze_button = gr.Button("๋ถ์") | |
summary_output = gr.Markdown(label="์์ฝ") | |
analysis_output = gr.Markdown(label="๋ถ์") | |
usage_output = gr.Markdown(label="์ฌ์ฉ๋ฒ") | |
tree_view_output = gr.Textbox(label="ํ์ผ ๊ตฌ์กฐ", lines=20) | |
with gr.Column(): | |
code_tabs = gr.Tabs() | |
with code_tabs: | |
with gr.TabItem("app.py"): | |
app_py_content = gr.Code( | |
language="python", | |
label="app.py", | |
lines=50 | |
) | |
with gr.TabItem("requirements.txt"): | |
requirements_content = gr.Textbox( | |
label="requirements.txt", | |
lines=50 | |
) | |
with gr.TabItem("AI ์ฝ๋์ฑ"): | |
gr.Markdown("## ์์ ๋ฅผ ์ ๋ ฅ ๋๋ ์์ค ์ฝ๋๋ฅผ ๋ถ์ฌ๋ฃ๊ณ ์ง๋ฌธํ์ธ์") | |
# Chatbot์ ๋จ์ง ์ถ๋ ฅ๋ง ๋ด๋น(ํํ์ ๋ฐ์ ํ์) | |
chatbot = gr.Chatbot( | |
label="๋ํ", | |
height=400 | |
) | |
msg = gr.Textbox( | |
label="๋ฉ์์ง", | |
placeholder="๋ฉ์์ง๋ฅผ ์ ๋ ฅํ์ธ์..." | |
) | |
# ์จ๊ฒจ์ง ํ๋ผ๋ฏธํฐ | |
max_tokens = gr.Slider( | |
minimum=1, maximum=8000, | |
value=4000, label="Max Tokens", | |
visible=False | |
) | |
temperature = gr.Slider( | |
minimum=0, maximum=1, | |
value=0.7, label="Temperature", | |
visible=False | |
) | |
top_p = gr.Slider( | |
minimum=0, maximum=1, | |
value=0.9, label="Top P", | |
visible=False | |
) | |
examples = [ | |
["์์ธํ ์ฌ์ฉ ๋ฐฉ๋ฒ์ 4000 ํ ํฐ ์ด์ ์์ธํ ์ค๋ช "], | |
["FAQ 20๊ฑด์ 4000 ํ ํฐ ์ด์ ์์ฑ"], | |
["๊ธฐ์ ์ฐจ๋ณ์ , ๊ฐ์ ์ ์ค์ฌ์ผ๋ก 4000 ํ ํฐ ์ด์ ์ค๋ช "], | |
["ํนํ ์ถ์์ ํ์ฉ ๊ฐ๋ฅํ ํ์ ์์ด๋์ด๋ฅผ 4000 ํ ํฐ ์ด์ ์์ฑ"], | |
["๋ ผ๋ฌธ ํ์์ผ๋ก 4000 ํ ํฐ ์ด์ ์์ฑ"], | |
["๊ณ์ ์ด์ด์ ๋ต๋ณํ๋ผ"] | |
] | |
gr.Examples(examples, inputs=msg) | |
# ๋ํ ์ํ(์ฑํ ๊ธฐ๋ก)๋ ChatMessage ๊ฐ์ฒด๋ก๋ง ์ ์ง | |
conversation_state = gr.State([]) | |
# ์ด๋ฒคํธ ์ฒด์ธ | |
# 1) ์ ์ ๋ฉ์์ง -> user_submit_message -> (์ ๋ ฅ์ฐฝ ๋น์ + state์ถ๊ฐ) | |
# 2) respond_wrapper -> Gemini ์คํธ๋ฆฌ๋ฐ -> ๋ํ state ๊ฐฑ์ -> (user,assistant) ํํ ๋ณํ | |
msg.submit( | |
user_submit_message, | |
inputs=[msg, conversation_state], | |
outputs=[msg, conversation_state], | |
queue=False | |
).then( | |
respond_wrapper, | |
inputs=[msg, conversation_state, max_tokens, temperature, top_p], | |
outputs=[msg, chatbot], | |
) | |
with gr.TabItem("Recommended Best"): | |
gr.Markdown( | |
"Discover recommended HuggingFace Spaces [here](https://huggingface.co/spaces/openfree/Korean-Leaderboard)." | |
) | |
# ๋ถ์ ๋ฒํผ ๋ก์ง | |
space_id_state = gr.State() | |
tree_structure_state = gr.State() | |
app_py_content_lines = gr.State() | |
analyze_button.click( | |
analyze_space, | |
inputs=[url_input], | |
outputs=[ | |
app_py_content, | |
tree_view_output, | |
tree_structure_state, | |
space_id_state, | |
summary_output, | |
analysis_output, | |
usage_output, | |
app_py_content_lines | |
] | |
).then( | |
lambda space_id: get_file_content(space_id, "requirements.txt"), | |
inputs=[space_id_state], | |
outputs=[requirements_content] | |
).then( | |
lambda lines: gr.update(lines=lines), | |
inputs=[app_py_content_lines], | |
outputs=[app_py_content] | |
) | |
return demo | |
except Exception as e: | |
print(f"Error in create_ui: {str(e)}") | |
print(traceback.format_exc()) | |
raise | |
if __name__ == "__main__": | |
try: | |
print("Starting HuggingFace Space Analyzer...") | |
demo = create_ui() | |
print("UI created successfully.") | |
print("Configuring Gradio queue...") | |
demo.queue() | |
print("Gradio queue configured.") | |
print("Launching Gradio app...") | |
demo.launch( | |
server_name="0.0.0.0", | |
server_port=7860, | |
share=False, | |
debug=True, | |
show_api=False | |
) | |
print("Gradio app launched successfully.") | |
except Exception as e: | |
print(f"Error in main: {str(e)}") | |
print("Detailed error information:") | |
print(traceback.format_exc()) | |
raise | |