Step3 / app.py
Zenith Wang
Support multimodal input with paste image, add CoT streaming display
5f37fb9
raw
history blame
9.62 kB
import gradio as gr
import time
import base64
from openai import OpenAI
import os
from io import BytesIO
from PIL import Image
import re
# 配置
BASE_URL = "https://api.stepfun.com/v1"
# 从环境变量获取API密钥
STEP_API_KEY = os.environ.get("STEP_API_KEY", "")
def image_to_base64(image):
"""将PIL图像转换为base64字符串"""
if image is None:
return None
if isinstance(image, Image.Image):
buffered = BytesIO()
image.save(buffered, format="PNG")
img_str = base64.b64encode(buffered.getvalue()).decode('utf-8')
return img_str
return None
def extract_cot_and_answer(text):
"""从响应中提取CoT推理过程和最终答案"""
# 匹配<reasoning>标签内的内容
reasoning_pattern = re.compile(r'<reasoning>(.*?)</reasoning>', re.DOTALL)
match = reasoning_pattern.search(text)
if match:
cot = match.group(1).strip()
# 移除reasoning标签及其内容,得到最终答案
answer = reasoning_pattern.sub('', text).strip()
return cot, answer
else:
# 如果没有reasoning标签,整个响应就是答案
return "", text
def format_message_with_image(message_text, image_path=None):
"""格式化包含图片的消息"""
if image_path:
# 创建包含图片和文本的消息
return f'<img src="{image_path}" style="max-width: 200px; max-height: 200px; border-radius: 8px; margin-bottom: 10px;"><br>{message_text}'
return message_text
def call_step_api_stream(message, history):
"""调用Step API进行流式对话,支持多模态输入"""
print(f"[DEBUG] Starting API call - Message type: {type(message)}")
if not message:
print("[DEBUG] No message provided")
yield history, "", ""
return
if not STEP_API_KEY:
print("[DEBUG] API key not configured")
error_msg = "❌ API key not configured. Please add STEP_API_KEY in Settings."
history.append([message if isinstance(message, str) else "Message", error_msg])
yield history, "", ""
return
print(f"[DEBUG] API Key exists: {bool(STEP_API_KEY)}")
# 处理多模态输入
text_content = ""
image_content = None
display_message = ""
# Gradio MultimodalTextbox 返回一个字典
if isinstance(message, dict):
text_content = message.get("text", "")
files = message.get("files", [])
# 处理图片文件
if files and len(files) > 0:
image_path = files[0] # 取第一张图片
try:
img = Image.open(image_path)
image_content = image_to_base64(img)
# 创建显示消息,包含图片缩略图
display_message = format_message_with_image(text_content, image_path)
print(f"[DEBUG] Image processed successfully")
except Exception as e:
print(f"[DEBUG] Failed to process image: {e}")
display_message = text_content
else:
display_message = text_content
else:
# 纯文本消息
text_content = str(message)
display_message = text_content
# 添加用户消息到历史
history.append([display_message, ""])
yield history, "", ""
# 构造API消息
messages = []
# 添加历史对话(只提取文本部分,不包含HTML)
for h in history[:-1]: # 不包含当前消息
if h[0]: # 用户消息
# 从HTML中提取纯文本
user_text = re.sub(r'<[^>]+>', '', h[0]) if '<' in h[0] else h[0]
messages.append({"role": "user", "content": user_text})
if h[1]: # 助手回复
messages.append({"role": "assistant", "content": h[1]})
# 构造当前消息
if image_content:
# 有图片的情况
current_content = [
{"type": "image_url", "image_url": {"url": f"data:image/jpg;base64,{image_content}", "detail": "high"}}
]
if text_content:
current_content.append({"type": "text", "text": text_content})
messages.append({"role": "user", "content": current_content})
else:
# 纯文本
messages.append({"role": "user", "content": text_content})
print(f"[DEBUG] Messages count: {len(messages)}")
# 创建客户端
try:
client = OpenAI(api_key=STEP_API_KEY, base_url=BASE_URL)
print("[DEBUG] Client created successfully")
except Exception as e:
print(f"[DEBUG] Client initialization failed: {e}")
history[-1][1] = f"❌ Client initialization failed: {str(e)}"
yield history, "", ""
return
# 调用API
try:
print("[DEBUG] Calling API...")
response = client.chat.completions.create(
model="step-3",
messages=messages,
temperature=0.7,
max_tokens=2000,
stream=True
)
print("[DEBUG] API call successful, processing stream...")
# 处理流式响应
full_response = ""
current_cot = ""
current_answer = ""
chunk_count = 0
for chunk in response:
chunk_count += 1
if chunk.choices and len(chunk.choices) > 0:
delta = chunk.choices[0].delta
if hasattr(delta, 'content') and delta.content:
full_response += delta.content
# 实时提取CoT和答案
current_cot, current_answer = extract_cot_and_answer(full_response)
# 更新历史中的回复
if current_cot and current_answer:
# 如果有CoT,显示完整格式
history[-1][1] = f"💭 **Reasoning Process:**\n\n{current_cot}\n\n---\n\n📝 **Answer:**\n\n{current_answer}"
elif current_cot:
# 只有CoT,还没有答案
history[-1][1] = f"💭 **Reasoning Process:**\n\n{current_cot}\n\n---\n\n📝 **Answer:**\n\n*Generating...*"
else:
# 没有CoT,直接显示答案
history[-1][1] = current_answer
print(f"[DEBUG] Chunk {chunk_count}: processed")
yield history, current_cot, current_answer
if not full_response:
print("[DEBUG] No response content received")
history[-1][1] = "⚠️ No response received from API"
yield history, "", ""
else:
print(f"[DEBUG] Final response length: {len(full_response)} chars")
except Exception as e:
print(f"[DEBUG] API request failed: {e}")
import traceback
traceback.print_exc()
history[-1][1] = f"❌ API request failed: {str(e)}"
yield history, "", ""
def clear_history():
"""Clear conversation history"""
return [], None
# 创建Gradio界面
with gr.Blocks(title="Step-3", theme=gr.themes.Soft()) as demo:
gr.Markdown("""
# 🤖 Step-3
Hello, I am Step-3!
""")
with gr.Row():
with gr.Column(scale=2):
# 对话界面
chatbot = gr.Chatbot(
height=600,
show_label=False,
elem_id="chatbot",
bubble_full_width=False,
render_markdown=True
)
with gr.Row():
# 多模态输入框 - 支持文本和图片
msg = gr.MultimodalTextbox(
placeholder="Type your message here... (You can paste images directly)",
show_label=False,
file_types=["image"],
container=False,
submit_btn="Send"
)
clear_btn = gr.Button("Clear", scale=0)
with gr.Column(scale=1):
# CoT推理过程展示
gr.Markdown("### 💭 Chain of Thought")
cot_display = gr.Textbox(
label="Reasoning Process",
lines=10,
max_lines=15,
show_label=False,
interactive=False,
show_copy_button=True
)
gr.Markdown("### 📝 Final Answer")
answer_display = gr.Textbox(
label="Answer",
lines=10,
max_lines=15,
show_label=False,
interactive=False,
show_copy_button=True
)
# 事件处理
msg.submit(
call_step_api_stream,
[msg, chatbot],
[chatbot, cot_display, answer_display]
)
clear_btn.click(
clear_history,
None,
[chatbot, msg]
)
# 页脚
gr.Markdown("""
---
<div style="text-align: center;">
<img src="https://huggingface.co/stepfun-ai/step3/resolve/main/figures/stepfun-logo.png" alt="StepFun Logo" style="height: 40px; margin: 10px;">
<br>
Powered by <a href="https://www.stepfun.com/" target="_blank">StepFun</a>
</div>
""")
# 启动应用
if __name__ == "__main__":
print(f"[DEBUG] Starting app with API key: {'Set' if STEP_API_KEY else 'Not set'}")
print(f"[DEBUG] Base URL: {BASE_URL}")
demo.queue(max_size=10)
demo.launch(
share=False,
debug=True
)