init

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.faiss filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,173 @@
+# local
+data/
+data_all/
+data_simple/
+compose.ipynb
+deal_data_org.py
+search.ipynb
+
+.env.local
+.env
+
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# poetry
+#   Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
+#   This is especially recommended for binary packages to ensure reproducibility, and is more
+#   commonly ignored for libraries.
+#   https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
+#poetry.lock
+
+# pdm
+#   Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
+#pdm.lock
+#   pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
+#   in version control.
+#   https://pdm.fming.dev/latest/usage/project/#working-with-version-control
+.pdm.toml
+.pdm-python
+.pdm-build/
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# PyCharm
+#  JetBrains specific template is maintained in a separate JetBrains.gitignore that can
+#  be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
+#  and can be added to the global gitignore or merged into this file.  For a more nuclear
+#  option (not recommended) you can uncomment the following to ignore the entire idea folder.
+#.idea/

app.py

@@ -0,0 +1,346 @@
+import os
+import time
+import json
+import asyncio
+import gradio as gr
+
+# set the env
+from dotenv import load_dotenv
+load_dotenv()
+
+# get the root path of the project
+current_file_path = os.path.dirname(os.path.abspath(__file__))
+root_path = os.path.abspath(current_file_path)
+
+from textwrap import dedent
+from langchain_openai import ChatOpenAI
+from langchain_openai import OpenAIEmbeddings
+from langchain_community.vectorstores import FAISS
+from langchain_core.prompts import ChatPromptTemplate
+
+class OurLLM:
+    def __init__(self, model="gpt-4o"):
+        '''
+        params: 
+            model: str, 
+                模型名称 ["GLM-4-Flash", "GLM-4V-Flash", 
+                         "gpt-4o-mini", "gpt-4o", "o1-mini", 
+                         "gemini-1.5-flash-002", "gemini-1.5-pro-002",
+                         "Qwen/Qwen2.5-7B-Instruct", "Qwen/Qwen2.5-Coder-7B-Instruct"]
+        '''
+
+        self.model_name = model
+
+        OPENAI_API_KEY = os.getenv('OPENAI_API_KEY')
+        OPENAI_API_KEY_DF = os.getenv('OPENAI_API_KEY_DF', OPENAI_API_KEY)
+        OPENAI_API_KEY_AZ = os.getenv('OPENAI_API_KEY_AZ', OPENAI_API_KEY)
+        OPENAI_API_KEY_CD = os.getenv('OPENAI_API_KEY_CD')
+        OPENAI_API_KEY_O1 = os.getenv('OPENAI_API_KEY_O1')
+        OPENAI_API_KEY_GLM = os.getenv('OPENAI_API_KEY_GLM')
+        OPENAI_API_KEY_SC = os.getenv('OPENAI_API_KEY_SC')
+
+        OPENAI_BASE_URL = os.getenv('OPENAI_BASE_URL')
+        OPENAI_BASE_URL_GLM = os.getenv('OPENAI_BASE_URL_GLM')
+        OPENAI_BASE_URL_SC = os.getenv('OPENAI_BASE_URL_SC')
+
+        # 创建 API Key 映射
+        apiKeyMap = {
+            'gemini': {"base_url": OPENAI_BASE_URL, "api_key": OPENAI_API_KEY_DF},
+            'gpt': {"base_url": OPENAI_BASE_URL, "api_key": OPENAI_API_KEY_AZ},
+            'o1': {"base_url": OPENAI_BASE_URL, "api_key": OPENAI_API_KEY_O1},
+            'claude': {"base_url": OPENAI_BASE_URL, "api_key": OPENAI_API_KEY_CD},
+            'glm': {"base_url": OPENAI_BASE_URL_GLM, "api_key": OPENAI_API_KEY_GLM},
+            'qwen': {"base_url": OPENAI_BASE_URL_SC, "api_key": OPENAI_API_KEY_SC},
+        }
+
+        for name, info in apiKeyMap.items():
+            if name in model.lower():
+                self.base_url = info["base_url"]
+                self.api_key = info["api_key"]
+                break
+        assert self.base_url is not None, f"Base URL not found for model: {model}"
+        assert self.api_key is not None, f"API key not found for model: {model}"
+
+        chat_prompt = ChatPromptTemplate.from_messages(
+            [
+                ("system", "{system_prompt}"),
+                ("human", "{input}"),
+                # ("ai", "{chat_history}"),
+            ]
+        )
+        self.chat_prompt = chat_prompt
+        self.llm = self.get_llm(model)
+
+    def clean_json(self, s):
+        return s.replace("```json", "").replace("```", "").strip()
+
+    def get_system_prompt(self, mode="assistant"):
+        prompt_map = {
+            "assistant": dedent("""
+                你是一个智能助手，擅长用简洁的中文回答用户的问题。
+                请确保你的回答准确、清晰、有条理，并且符合中文的语言习惯。
+                重要提示：
+                1. 回答要简洁明了，避免冗长
+                2. 使用适当的专业术语
+                3. 保持客观中立的语气
+                4. 如果不确定，要明确指出
+            """),
+            # search
+            "keyword_expand": dedent("""
+                你是一个搜索关键词扩展专家，擅长将用户的搜索意图转化为多个相关的搜索词或短语。
+                用户会输入一段描述他们搜索需求的文本，请你生成与之相关的关键词列表。
+                你需要返回一个可以直接被 json 库解析的响应，包含以下内容：
+                {
+                    "keywords": [关键词列表],
+                }
+                重要提示：
+                1. 关键词应该包含同义词、近义词、上位词、下位词
+                2. 短语要体现不同的表达方式和组合
+                3. 描述句子要涵盖不同的应用场景和用途
+                4. 所有内容必须与原始搜索意图高度相关
+                5. 扩展搜索意图到相关的应用场景和工具，例如:
+                    - 如果搜索"PDF转MD"，应包含PDF内容提取、PDF解析工具、PDF数据处理等
+                    - 如果搜索"图片压缩"，应包含批量压缩工具、图片格式转换等
+                    - 如果搜索"代码格式化"，应包含代码美化工具、语法检查器、代码风格统一等
+                    - 如果搜索"文本翻译"，应包含机器翻译API、多语言翻译工具、离线翻译软件等
+                    - 如果搜索"数据可视化"，应包含图表生成工具、数据分析库、交互式图表等
+                    - 如果搜索"网络爬虫"，应包含数据采集框架、反爬虫绕过、数据解析工具等
+                    - 如果搜索"API测试"，应包含接口测试工具、性能监控、自动化测试框架等
+                6. 所有内容主要使用英文表达，并对部分关键词添加额外的中文表示
+                7. 返回内容不要使用任何 markdown 格式 以及任何特殊字符
+            """),
+            "zh2en": dedent("""
+                你是一个专业的中译英翻译专家，尤其擅长学术论文的翻译工作。
+                请将用户提供的中文内容翻译成地道、专业的英文。
+
+                重要提示：
+                1. 使用学术论文常用的表达方式和术语
+                2. 保持专业、正式的语气
+                3. 确保译文的准确性和流畅性
+                4. 对专业术语进行准确翻译
+                5. 遵循英文学术写作的语法规范
+                6. 保持原文的逻辑结构
+                7. 适当使用学术论文常见的过渡词和连接词
+                8. 如遇到模糊的表达，选择最符合学术上下文的翻译
+                9. 避免使用口语化或非正式的表达
+                10. 注意时态和语态的准确使用
+            """),
+            "github_score": dedent("""
+                你是一个语义匹配评分专家，擅长根据用户需求和仓库描述进行语义匹配度评分。
+                用户会输入两部分内容:
+                1. 用户的具体需求描述
+                2. 多个仓库的描述列表(以1,2,3等数字开头)
+                
+                请你仔细分析用户需求，并对每个仓库进行评分。
+                确保返回一个可以直接被 json 库解析的响应，包含以下内容：
+                {
+                    "indices": [仓库编号列表，按分数从高到低],
+                    "scores": [编号对应的匹配度评分列表，0-100的整数，表示匹配程度]
+                }
+                
+                重要提示：
+                1. 评分范围为0-100的整数，高于60分表示具有明显相关性
+                2. 评分要客观反映仓库与需求的契合度
+                3. 只返回评分大于 60 的仓库
+                4. 返回内容不要使用任何 markdown 格式 以及任何特殊字符
+            """)
+        }
+        return prompt_map[mode]
+
+    def get_llm(self, model="gpt-4o-mini"):
+        '''
+        params:
+            model: str, 模型名称 ["gpt-4o-mini", "gpt-4o", "o1-mini", "gemini-1.5-flash-002"]
+        '''
+        llm = ChatOpenAI(
+            model=model,
+            base_url=self.base_url,
+            api_key=self.api_key,
+        )
+        print(f"Init model {model} successfully!")
+        return llm
+    
+    def ask_question(self, question, system_prompt=None):
+        # 1. 获取系统提示
+        if system_prompt is None:
+            system_prompt = self.get_system_prompt()
+        
+        # 2. 生成聊天提示
+        prompt = self.chat_prompt.format(input=question, system_prompt=system_prompt)
+        config = {
+            "configurable": {"response_format": {"type": "json_object"}}
+        }
+        
+        # 3. 调用 LLM 进行回答
+        for _ in range(10):
+            try:
+                response = self.llm.invoke(prompt, config=config)
+                response.content = self.clean_json(response.content)
+                return response
+            except Exception as e:
+                print(e)
+                time.sleep(10)
+                continue
+        print(f"Failed to call llm for prompt: {prompt[0:10]}")
+        return None
+    
+    async def ask_questions_parallel(self, questions, system_prompt=None):
+        # 1. 获取系统提示
+        if system_prompt is None:
+            system_prompt = self.get_system_prompt()
+
+        # 2. 定义异步函数
+        async def call_llm(prompt):
+            for _ in range(10):
+                try:
+                    response = await self.llm.ainvoke(prompt)
+                    response.content = self.clean_json(response.content)
+                    return response
+                except Exception as e:
+                    print(e)
+                    await asyncio.sleep(10)
+                    continue
+            print(f"Failed to call llm for prompt: {prompt[0:10]}")
+            return None
+
+        # 3. 构建 prompt
+        prompts = [self.chat_prompt.format(input=question, system_prompt=system_prompt) for question in questions]
+
+        # 4. 异步调用
+        tasks = [call_llm(prompt) for prompt in prompts]
+        results = await asyncio.gather(*tasks)
+
+        return results
+
+class RepoSearch:
+    def __init__(self):
+        db_path = os.path.join(root_path, "database", "init")
+        embeddings = OpenAIEmbeddings(api_key=os.getenv("OPENAI_API_KEY"), 
+                                      base_url=os.getenv("OPENAI_BASE_URL"),
+                                      model="text-embedding-3-small")
+        
+        assert os.path.exists(db_path), f"Database not found: {db_path}"
+        self.vector_db = FAISS.load_local(db_path, embeddings, 
+                                          index_name="init",
+                                          allow_dangerous_deserialization=True)
+        
+    def search(self, query, k=10):
+        '''
+            name + description + html_url + topics
+        '''
+        results = self.vector_db.similarity_search(query + " technology", k=k)
+
+        simple_str = ""
+        simple_list = []
+        for i, doc in enumerate(results):
+            content = json.loads(doc.page_content)
+            metadata = doc.metadata
+            if content["description"] is None:
+                content["description"] = ""
+            # desc = content["description"] if len(content["description"]) < 300 else content["description"][:300] + "..."
+            simple_str += f"\t**{i+1}. {content['name']}** || {content['description']}\n" # 用于大模型匹配
+            simple_list.append({
+                "name": content["name"],
+                "description": content["description"],
+                **metadata,  # 解包所有 metadata 字段
+            })
+
+        return simple_str, simple_list
+
+def main():
+    search = RepoSearch()
+    llm = OurLLM(model="gpt-4o")
+
+    def respond(
+        prompt: str,
+        history,
+        is_llm_filter: bool = False,
+        is_keyword_expand: bool = False,
+        match_num: int = 40
+    ):
+        # 1. 初始化历史记录
+        if not history:
+            history = [{"role": "system", "content": "You are a friendly chatbot"}]
+        history.append({"role": "user", "content": prompt})
+        response = {"role": "assistant", "content": ""}
+        yield history
+
+        # 2. 扩展用户问题关键词
+        if is_keyword_expand:
+            response["content"] = "开始扩展关键词..."
+            yield history + [response]
+
+            query = llm.ask_question(prompt, system_prompt=llm.get_system_prompt("keyword_expand")).content
+            prompt = ", ".join(json.loads(query)["keywords"])
+
+        # 3. 语义向量匹配
+        response["content"] = "开始语义向量匹配..."
+        yield history + [response]
+        match_str, simple_list = search.search(prompt, match_num)
+
+        # 4. 通过 LLM 评分得到最匹配的仓库索引
+        if not is_llm_filter:
+            simple_strs = [f"\t**{i+1}. {repo['name']}** [✨ {repo['star_count'] // 1000}k] || **Description:** {repo['description']} || **Url:** {repo['html_url']} \n" for i, repo in enumerate(simple_list)]
+            response["content"] = "".join(simple_strs)
+            yield history + [response]
+        else:
+            response["content"] = "开始通过 LLM 评分得到最匹配的仓库..."
+            yield history + [response]
+
+            query = ' ## 用户需要的仓库内容：' + prompt + '\n ## 搜索结果列表：' + match_str
+            out = llm.ask_question(query, system_prompt=llm.get_system_prompt("github_score")).content
+            matched_index = json.loads(out)["indices"]
+
+            # 5. 通过索引得到最匹配的仓库
+            result = [simple_list[idx-1] for idx in matched_index]
+            simple_strs = [f"\t**{i+1}. {repo['name']}** [✨ {repo['star_count'] // 1000}k] || **Description:** {repo['description']} || **Url:** {repo['html_url']} \n" for i, repo in enumerate(result)]
+            response["content"] = "".join(simple_strs)
+            yield history + [response]
+
+    with gr.Blocks() as demo:
+        gr.Markdown("## Github semantic search (基于语义的 github 仓库搜索) 🌐")
+        
+        with gr.Row():
+            with gr.Column(scale=1):
+                # 添加控制参数
+                llm_filter = gr.Checkbox(
+                    label="使用LLM过滤结果",
+                    value=False,
+                    info="是否使用 LLM 对搜索结果进行二次过滤"
+                )
+                keyword_expand = gr.Checkbox(
+                    label="扩展关键词搜索",
+                    value=False,
+                    info="是否使用 LLM 扩展搜索关键词"
+                )
+                match_number = gr.Slider(
+                    minimum=10,
+                    maximum=100,
+                    value=40,
+                    step=10,
+                    label="语义匹配数量",
+                    info="进行语义匹配后返回的仓库数量，若使用 LLM 过滤，建议适当增加数量"
+                )
+            
+            with gr.Column(scale=3):
+                chatbot = gr.Chatbot(
+                    label="Agent",
+                    type="messages",
+                    avatar_images=(None, "https://img1.baidu.com/it/u=2193901176,1740242983&fm=253&fmt=auto&app=138&f=JPEG?w=500&h=500"),
+                    height="65vh"
+                )
+                prompt = gr.Textbox(max_lines=2, label="Chat Message")
+                
+        # 更新submit调用，包含新的参数
+        prompt.submit(
+            respond, 
+            [prompt, chatbot, llm_filter, keyword_expand, match_number], 
+            [chatbot]
+        )
+        prompt.submit(lambda: "", None, [prompt])
+
+    demo.launch(share=False)
+
+
+if __name__ == "__main__":
+    main()

database/init/init.faiss

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:007ee89904bde801789ec8fd0c3b56dc1d6b2d684d3c4c80f808cf1614c38ad7
+size 279736365

database/init/init.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3e3d8fd7c484feab60214829cf3e9f1edbb99e48d6453aed86a6899223bf379
+size 19696870

deal_data.py

@@ -0,0 +1,136 @@
+import os
+import json
+import asyncio
+import requests
+
+from tqdm import tqdm
+from dotenv import load_dotenv
+load_dotenv()
+
+from langchain_openai import ChatOpenAI
+from langchain_core.prompts import ChatPromptTemplate
+from langchain_core.documents import Document
+from langchain_openai import OpenAIEmbeddings
+from langchain_community.vectorstores import FAISS
+
+# 获取当前目录根路径
+current_file_path = os.path.dirname(os.path.abspath(__file__))
+root_path = os.path.abspath(current_file_path)
+data_path = os.path.join(root_path, "data_simple") 
+db_path = os.path.join(root_path, "database", "init")
+
+# 1. 根据 star 数量区间获取 GitHub 仓库，同时根据 star 数量从多到少排序（闭区间）并保存 GitHub 仓库
+def get_top_repo_by_star(per_page=1000, page=1, min_star_num=0, max_star_num=500000):
+    query = f'stars:{min_star_num}..{max_star_num} pushed:>2021-01-01'
+    sort = 'stars'
+    order = 'desc'
+    search_url = f'{os.getenv('GITHUB_API_URL')}/search/repositories?q={query}&sort={sort}&order={order}&per_page={per_page}&page={page}'
+    headers = {"Authorization": f"token {os.getenv('GITHUB_TOKEN')}"}
+
+    response = requests.get(search_url, headers=headers)
+    if response.status_code == 200:
+        total_count = response.json()['total_count']
+        total_page = total_count // per_page + 1
+        print(f"Total page: {total_page}, current page: {page}")
+        if response.json()['incomplete_results']: print("Incomplete results")
+        return response.json()['items'], response.json()['items'][-1]['stargazers_count'], total_count
+    else:
+        print(f"Failed to retrieve repositories: {response.status_code}")
+        print("")
+        # 直接退出
+        exit(1)
+
+def save_repo_by_star(max_star=500000):
+    # github 限制每次请求最多得到 100 个仓库，因此 page 固定为 1
+    top_repositories, max_star, count = get_top_repo_by_star(per_page=1000, page=1, min_star_num=1000, max_star_num=max_star)
+
+    for i, repo in enumerate(top_repositories):
+        owner = repo['owner']['login']
+        name = repo['name']
+        unique_id = f"{name} -- {owner}"
+        stars = repo['stargazers_count']
+        print(f"Repository {i}: {name}, Stars: {stars}")
+
+        # 存储为 json 格式
+        with open(os.path.join(data_path, f'{unique_id}.json'), 'w') as f:
+            json.dump(repo, f, indent=4)
+
+    if count < 100: exit(1)
+
+    return max_star
+
+def main_repo():
+    max_star = 500000 # 最多 star 的仓库有 500k
+    num = 1
+    while True:
+        print("=" * 50)
+        print(f"Round {num}, Max star: {max_star}")
+        max_star = save_repo_by_star(max_star)
+        num += 1
+
+# 2. 将数据转换为向量
+async def create_vector_db(docs, embeddings, batch_size=800):
+    # 初始化第一批数据
+    vector_db = await FAISS.afrom_documents(docs[0:batch_size], embeddings)
+    if len(docs) < batch_size: return vector_db
+    
+    # 创建任务x``
+    tasks = []
+    for start_idx in range(batch_size, len(docs), batch_size):
+        end_idx = min(start_idx + batch_size, len(docs))
+        tasks.append(FAISS.afrom_documents(docs[start_idx:end_idx], embeddings))
+
+    # 执行任务
+    results = await asyncio.gather(*tasks)
+
+    # 合并结果
+    for temp_db in results:
+        vector_db.merge_from(temp_db)
+    return vector_db
+
+async def main_convert_to_vector():
+    # 读取文件
+    files = os.listdir(data_path)
+
+    # 构建 document
+    docs = []
+    for file in tqdm(files):
+        if not file.endswith(".json"): continue
+        with open(os.path.join(data_path, file), "r", encoding="utf-8") as f:
+            data = json.load(f)
+        
+        content_map = {
+            "name": data["name"],
+            "description": data["description"],
+        }
+        content = json.dumps(content_map)
+        doc = Document(page_content=content, metadata={"html_url": data["html_url"], 
+                                                    "topics": data["topics"],
+                                                    "created_at": data["created_at"],
+                                                    "updated_at": data["updated_at"],
+                                                    "star_count": data["stargazers_count"]})
+        docs.append(doc)
+    print(f"Total {len(docs)} documents.")
+
+    # 初始化 Embedding 实例
+    embeddings = OpenAIEmbeddings(api_key=os.getenv("OPENAI_API_KEY"),
+                                  base_url=os.getenv("OPENAI_BASE_URL"),
+                                  model="text-embedding-3-small")
+    print("Embedding model success: text-embedding-3-small")
+
+    # 文档嵌入
+    if os.path.exists(os.path.join(db_path, "init.faiss")):
+        vector_db = FAISS.load_local(db_path, embeddings=embeddings,
+                                        index_name="init",
+                                        allow_dangerous_deserialization=True)
+    else:
+        vector_db = await create_vector_db(docs, embeddings=embeddings)
+        vector_db.save_local(db_path, index_name="init")
+    return vector_db
+
+if __name__ == "__main__":
+    # 1. 获取仓库信息
+    # main_repo()
+
+    # 2. 构建向量数据库
+    asyncio.run(main_convert_to_vector())

requirements.txt

@@ -0,0 +1,9 @@
+langchain_community
+langchain_core
+langchain_openai
+
+faiss-cpu
+tqdm
+python-dotenv
+
+gradio