app.py

import gradio as gr
import openai
from openai import OpenAI
import google.generativeai as genai
import os
import io
import base64

# Set API key and organization ID from environment variables
api_key = os.environ.get("API_KEY")
#base_url = os.environ.get("OPENAI_API_BASE")

# Define the model to be used
MODEL = os.environ.get("MODEL")
MODEL_NAME = MODEL.split("/")[-1] if "/" in MODEL else MODEL

def read(filename):
    with open(filename) as f:
        data = f.read()
    return data
    
SYS_PROMPT = read('system_prompt.txt')


DESCRIPTION = '''
<div>
<h1 style="text-align: center;">知觉demo</h1>
<p>🩺一个基于提示词和前沿多模态模型的AI，帮助您解读专业领域内容。</p>
<p>🔎 您可以选择领域，参考示例上传图像，或发送需要解读的文字内容。</p>
<p>🦕 生成解读内容仅供参考。</p>
</div>
'''


css = """
h1 {
    text-align: center;
    display: block;
}
footer {
    display:none !important
}
"""


LICENSE = '采用 ' + MODEL_NAME + ' 模型'

def endpoints(api_key):
    if api_key is not None:
        if api_key.startswith('sk-'):
            return 'OPENAI'
        else:
            return 'GOOGLE'

def process_text(text_input, unit):
    print(text_input)
    endpoint = endpoints(api_key)
    if text_input and endpoint == 'OPENAI':
        client = OpenAI(api_key=api_key)
        completion = client.chat.completions.create(
            model=MODEL,
            messages=[
                {"role": "system", "content": f" You are a experienced Analyst in {unit}." + SYS_PROMPT},
                {"role": "user", "content": f"Hello! Could you analysis {text_input}?"}
            ]
        )
        return completion.choices[0].message.content
    elif text_input and endpoint == 'GOOGLE':
        genai.configure(api_key=api_key)
        model = genai.GenerativeModel(model_name=MODEL)
        prompt = f" You are a experienced Analyst in {unit}." + SYS_PROMPT + f"Could you analysis {text_input}?"
        response = model.generate_content(prompt)
        return response.text        
    return ""

def encode_image_to_base64(image_input):
    buffered = io.BytesIO()
    image_input.save(buffered, format="JPEG")
    img_str = base64.b64encode(buffered.getvalue()).decode("utf-8")
    return img_str

def process_image(image_input, unit):
    endpoint = endpoints(api_key)
    if image_input is not None and endpoint == 'OPENAI':
        #with open(image_input.name, "rb") as f:
        #    base64_image = base64.b64encode(f.read()).decode("utf-8")
        client = OpenAI(api_key=api_key)
        base64_image = encode_image_to_base64(image_input)
        response = client.chat.completions.create(
            model=MODEL,
            messages=[
                {"role": "system", "content": f" You are a experienced Analyst in {unit}." + SYS_PROMPT},
                {"role": "user", "content": [
                    {"type": "text", "text": "Help me understand what is in this picture and analysis."},
                    {"type": "image_url", 
                     "image_url": {
                        "url": f"data:image/jpeg;base64,{base64_image}",
                        "detail":"low"}
                    }
                ]}
            ],
            temperature=0.0,
            max_tokens=1024,
        )
        return response.choices[0].message.content
    elif image_input is not None and endpoint == 'GOOGLE':
        print(image_input)
        genai.configure(api_key=api_key)
        model = genai.GenerativeModel(model_name=MODEL)
        prompt = f" You are a experienced Analyst in {unit}." + SYS_PROMPT + "Help me understand what is in this picture and analysis it."
        response = model.generate_content([prompt, image_input],request_options={"timeout": 60})
        return response.text        


def main(text_input="", image_input=None, unit=""):
    if text_input and image_input is None:
        return process_text(text_input,unit)
    elif image_input is not None:
        return process_image(image_input,unit)
    else:
        gr.Error("请输入内容或者上传图片")

EXAMPLES = [
        ["./docs/estate.jpeg","",],
        ["./docs/pop.jpeg","",],
        ["./docs/debt.jpeg","",],
        [None,"中国央行表示高度关注当前债券市场变化及潜在风险，必要时会进行卖出低风险债券包括国债操作",],
]

with gr.Blocks(theme='shivi/calm_seafoam', css=css, title="知觉demo") as iface:
    with gr.Accordion(""):
        gr.Markdown(DESCRIPTION)
        unit = gr.Dropdown(label="领域", value='财经', elem_id="units",
                            choices=["财经", "法律", "政治", "体育", "医疗", \
                                     "SEO", "评估", "科技", "交通", "行情"])
    with gr.Row():
        output_box = gr.Markdown(label="分析")  # Create an output textbox
    with gr.Row():
        image_input = gr.Image(type="pil", label="上传图片")  # Create an image upload button
        text_input = gr.Textbox(label="输入")  # Create a text input box
    with gr.Row():
        submit_btn = gr.Button("🚀 确认")  # Create a submit button
        clear_btn = gr.ClearButton([output_box,image_input,text_input], value="🗑️ 清空") # Create a clear button

    # Set up the event listeners
    submit_btn.click(main, inputs=[text_input, image_input, unit], outputs=output_box)
    gr.Examples(examples=EXAMPLES, inputs=[image_input, text_input])
    gr.Markdown(LICENSE)
    
#gr.close_all()

iface.queue().launch(show_api=False)  # Launch the Gradio interface