叫板GPT-4的Gemini，我做了一个聊天网页，可图片输入，附教程

# 先安装google-generativeai
pip install -q -U google-generativeai

import google.generativeai as genai
GOOGLE_API_KEY='这里填写上一步获取的api'
genai.configure(api_key=GOOGLE_API_KEY)
model = genai.GenerativeModel('gemini-pro')
response = model.generate_content("你好")
print(response.text)

import PIL.Image
import google.generativeai as genai
img = PIL.Image.open('img.png')
GOOGLE_API_KEY='这里填写上一步获取的api'
genai.configure(api_key=GOOGLE_API_KEY)
model = genai.GenerativeModel('gemini-pro-vision')
response = model.generate_content(["请评价一下这张照片", img])
response.resolve()
print(response.text)

import time
from typing import List, Tuple, Optional

import google.generativeai as genai
import gradio as gr
from PIL import Image

print("google-generativeai:", genai.__version__)

TITLE = """<h1 align="center">Gemini App</h1>"""
SUBTITLE = """<h2 align="center">仅做试玩，不定期下线</h2>"""
GOOGLE_API_KEY='这里填写上一步获取的api'

AVATAR_IMAGES = (
    None,
    "image.png"
)


def preprocess_stop_sequences(stop_sequences: str) -> Optional[List[str]]:
    if not stop_sequences:
        return None
    return [sequence.strip() for sequence in stop_sequences.split(",")]


def user(text_prompt: str, chatbot: List[Tuple[str, str]]):
    return "", chatbot + [[text_prompt, None]]


def bot(
    #google_key: str,
    image_prompt: Optional[Image.Image],
    temperature: float,
    max_output_tokens: int,
    stop_sequences: str,
    top_k: int,
    top_p: float,
    chatbot: List[Tuple[str, str]]
):

    text_prompt = chatbot[-1][0]
    genai.configure(api_key=GOOGLE_API_KEY)
    generation_config = genai.types.GenerationConfig(
        temperature=temperature,
        max_output_tokens=max_output_tokens,
        stop_sequences=preprocess_stop_sequences(stop_sequences=stop_sequences),
        top_k=top_k,
        top_p=top_p)

    if image_prompt is None:
        model = genai.GenerativeModel('gemini-pro')
        response = model.generate_content(
            text_prompt,
            stream=True,
            generation_config=generation_config)
        response.resolve()
    else:
        model = genai.GenerativeModel('gemini-pro-vision')
        response = model.generate_content(
            [text_prompt, image_prompt],
            stream=True,
            generation_config=generation_config)
        response.resolve()

    # streaming effect
    chatbot[-1][1] = ""
    for chunk in response:
        for i in range(0, len(chunk.text), 10):
            section = chunk.text[i:i + 10]
            chatbot[-1][1] += section
            time.sleep(0.01)
            yield chatbot


image_prompt_component = gr.Image(type="pil", label="Image", scale=1)
chatbot_component = gr.Chatbot(
    label='Gemini',
    bubble_full_width=False,
    avatar_images=AVATAR_IMAGES,
    scale=2
)
text_prompt_component = gr.Textbox(
    placeholder="你好！",
    label="请在这里提问···"
)
run_button_component = gr.Button()
temperature_component = gr.Slider(
    minimum=0,
    maximum=1.0,
    value=0.4,
    step=0.05,
    label="Temperature",
    info=(
        "Temperature 控制令牌选择的随机程度 "
        "较低的Temperature适用于期望获得真实或正确回答的提示, "
        "而较高的Temperature可以导致更多样化或意外的结果 "
    ))
max_output_tokens_component = gr.Slider(
    minimum=1,
    maximum=2048,
    value=1024,
    step=1,
    label="Token limit",
    info=(
        "Token 限制确定每个提示可以获得的最大文本输出量 "
        "每个 Token 大约为四个字符，默认值为 2048 "
    ))

stop_sequences_component = gr.Textbox(
    label="Add stop sequence",
    value="",
    type="text",
    placeholder="STOP, END",
    info=(
        "停止序列是一系列字符(包括空格),如果模型遇到它,会停止生成响应"
        "该序列不作为响应的一部分，"
        "可以添加多达5个停止序列"
    ))
top_k_component = gr.Slider(
    minimum=1,
    maximum=40,
    value=32,
    step=1,
    label="Top-K",
    info=(
        "Top-k 改变了模型为输出选择 token 的方式 "
        "Top-k 为 1 意味着所选 token 在模型词汇表中所有 token 中是最可能的(也称为贪心解码)"
        "而 top-k 为 3 意味着下一个 token 从最可能的 3 个 token 中选取(使用temperature)"
    ))
top_p_component = gr.Slider(
    minimum=0,
    maximum=1,
    value=1,
    step=0.01,
    label="Top-P",
    info=(
        "Top-p 改变了模型为输出选择 token 的方式 "
        "token 从最可能到最不可能选择,直到它们的概率之和等于 top-p 值 "
        "如果 token A、B 和 C 的概率分别为 0.3、0.2 和 0.1,top-p 值为 0.5 "
        "那么模型将选择 A 或 B 作为下一个 token(使用temperature) "
    ))

user_inputs = [
    text_prompt_component,
    chatbot_component
]

bot_inputs = [
    image_prompt_component,
    temperature_component,
    max_output_tokens_component,
    stop_sequences_component,
    top_k_component,
    top_p_component,
    chatbot_component
]

with gr.Blocks() as demo:
    gr.HTML(TITLE)
    gr.HTML(SUBTITLE)
    with gr.Column():
        with gr.Row():
            image_prompt_component.render()
            chatbot_component.render()
        text_prompt_component.render()
        run_button_component.render()
        with gr.Accordion("Parameters", open=False):
            temperature_component.render()
            max_output_tokens_component.render()
            stop_sequences_component.render()
            with gr.Accordion("Advanced", open=False):
                top_k_component.render()
                top_p_component.render()

    run_button_component.click(
        fn=user,
        inputs=user_inputs,
        outputs=[text_prompt_component, chatbot_component],
        queue=False
    ).then(
        fn=bot, inputs=bot_inputs, outputs=[chatbot_component],
    )

    text_prompt_component.submit(
        fn=user,
        inputs=user_inputs,
        outputs=[text_prompt_component, chatbot_component],
        queue=False
    ).then(
        fn=bot, inputs=bot_inputs, outputs=[chatbot_component],
    )

demo.queue(max_size=99).launch(auth=("用户名", "密码"),debug=True)