init qa

.gitignore

@@ -0,0 +1,5 @@
+__pycache__/
+.env
+.gradio
+data/
+faiss_index/

README.md

@@ -4,7 +4,7 @@ emoji: 💬
 colorFrom: yellow
 colorTo: purple
 sdk: gradio
-sdk_version: 5.0.1
+sdk_version: 5.0.2
 app_file: app.py
 pinned: false
 ---

app.py

@@ -1,64 +1,206 @@
+from langchain.memory import ConversationBufferMemory
+
+import pandas as pd
 import gradio as gr
-from huggingface_hub import InferenceClient
 
-"""
-For more information on `huggingface_hub` Inference API support, please check the docs: https://huggingface.co/docs/huggingface_hub/v0.22.2/en/guides/inference
-"""
-client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
+from langchain.embeddings import OpenAIEmbeddings
+from langchain.vectorstores import FAISS
+
+import os
+
+
+import pandas as pd
+from langchain.embeddings import OpenAIEmbeddings
+from langchain.vectorstores import FAISS
+from langchain.schema import Document
+import os
+
+from utils import make_html_source, make_pairs, get_llm, reset_textbox
+
+from prompt import PROMPT_INTERPRATE_INTENTION, ANSWER_PROMPT
+
+
+try:
+    from dotenv import load_dotenv
+    load_dotenv()
+except Exception:
+    pass
+
+
+# Load your OpenAI API key
+import os
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+assert OPENAI_API_KEY, "Please set your OpenAI API key"
+
+embeddings = OpenAIEmbeddings(openai_api_key=OPENAI_API_KEY)
+
+
+
+
+new_vector_store = FAISS.load_local(
+    "faiss_index", embeddings, allow_dangerous_deserialization=True
+)
+
+retriever = new_vector_store.as_retriever()
+
+
 
+llm = get_llm()
 
-def respond(
-    message,
-    history: list[tuple[str, str]],
-    system_message,
-    max_tokens,
-    temperature,
-    top_p,
+memory = ConversationBufferMemory(
+    return_messages=True, output_key="answer", input_key="question"
+)
+
+def make_qa_chain(
+
+) :
+    final_inputs = {
+        "context": lambda x: x["context"],
+        "question": lambda x: x["question"],
+    }
+        
+        
+        
+    return final_inputs | ANSWER_PROMPT | llm
+
+
+def load_documents_meeting(meeting_number):
+    # Step 1: Load the CSV data
+    csv_file_path = "../data/mfls.xlsx"  
+    df = pd.read_excel(csv_file_path)
+    
+    df["meeting_number"]= df["Meeting"].apply(lambda x: x.split(" ")[0][:-2])
+    df_meeting = df[df["meeting_number"] == meeting_number]
+    def combine_title_and_content(row):
+        return f"{row['Meeting']} {row['Issues']} {row['Content']}"
+
+    df_meeting['combined'] = df_meeting.apply(combine_title_and_content, axis=1)
+
+    # Step 3: Generate embeddings using OpenAI
+    embeddings = OpenAIEmbeddings(openai_api_key=OPENAI_API_KEY)
+
+    # Generate embeddings for each document
+    documents = [
+        Document(
+            page_content=row['combined'], 
+            metadata={
+                "Issues": row['Issues'], 
+                "Title": row['Title'],
+                "meeting_number": row["Meeting"].split(" ")[0][:-2],
+                "Agencies": row["Agencies"],
+                "project": row["Projects"],
+            }
+        ) for i,row in df_meeting.iterrows()]
+    return documents
+
+
+async def chat(
+    query: str,
+    history: list = [],
 ):
-    messages = [{"role": "system", "content": system_message}]
-
-    for val in history:
-        if val[0]:
-            messages.append({"role": "user", "content": val[0]})
-        if val[1]:
-            messages.append({"role": "assistant", "content": val[1]})
-
-    messages.append({"role": "user", "content": message})
-
-    response = ""
-
-    for message in client.chat_completion(
-        messages,
-        max_tokens=max_tokens,
-        stream=True,
-        temperature=temperature,
-        top_p=top_p,
-    ):
-        token = message.choices[0].delta.content
-
-        response += token
-        yield response
-
-
-"""
-For information on how to customize the ChatInterface, peruse the gradio docs: https://www.gradio.app/docs/chatinterface
-"""
-demo = gr.ChatInterface(
-    respond,
-    additional_inputs=[
-        gr.Textbox(value="You are a friendly Chatbot.", label="System message"),
-        gr.Slider(minimum=1, maximum=2048, value=512, step=1, label="Max new tokens"),
-        gr.Slider(minimum=0.1, maximum=4.0, value=0.7, step=0.1, label="Temperature"),
-        gr.Slider(
-            minimum=0.1,
-            maximum=1.0,
-            value=0.95,
-            step=0.05,
-            label="Top-p (nucleus sampling)",
-        ),
-    ],
+    """taking a query and a message history, use a pipeline (reformulation, retriever, answering) to yield a tuple of:
+    (messages in gradio format, messages in langchain format, source documents)"""
+    source_string = ""
+    gradio_format = make_pairs([a.content for a in history]) + [(query, "")]
+    qa_chain = make_qa_chain()
+    
+    # reset memory
+    memory.clear()
+    for message in history:
+        memory.chat_memory.add_message(message)
+    
+    inputs = {"question": query}
+    
+    ## INTENT
+    intent = await llm.abatch([PROMPT_INTERPRATE_INTENTION.format_prompt(query = query)])
+    intent = intent[0].content
+    print("intent", intent)
+    
+    ## RETRIEVER
+    if intent.split(" ")[0] == "meeting":
+        meeting_number = intent.split(" ")[-1]
+        sources = load_documents_meeting(meeting_number)
+    else :
+        sources = new_vector_store.search(query, search_type="similarity", k=5)
+    
+    source_string = "\n\n".join([make_html_source(doc, i) for i, doc in enumerate(sources, 1)])
+
+    ## RAG
+    inputs_rag = {"question": query, "context": sources}
+    
+    result = qa_chain.astream_log(inputs_rag)
+
+    reformulated_question_path_id = "/logs/ChatOpenAI/streamed_output_str/-"
+    retriever_path_id =             "/logs/VectorStoreRetriever/final_output"
+    final_answer_path_id =          "/streamed_output/-"
+    
+    async for op in result:
+        op = op.ops[0]
+        # print(op["path"])
+        if op['path'] == reformulated_question_path_id: # reforulated question
+            new_token = op['value'] # str
+            
+        elif op['path'] == retriever_path_id: # documents
+            sources = op['value']['documents'] # List[Document]
+            source_string = "\n\n".join([make_html_source(i, doc) for i, doc in enumerate(sources, 1)])
+
+        elif op['path'] == final_answer_path_id: # final answer
+            new_token = op['value'].content # str
+            answer_yet = gradio_format[-1][1]
+            gradio_format[-1] = (query, answer_yet + new_token  )          
+
+        yield gradio_format, history, source_string
+
+    memory.save_context(inputs, {"answer": gradio_format[-1][1]})
+    yield gradio_format, memory.load_memory_variables({})["history"], source_string
+    
+    
+### GRADIO UI
+    
+theme = gr.themes.Soft(
+    primary_hue="sky",
+    font=[gr.themes.GoogleFont("Poppins"), "ui-sans-serif", "system-ui", "sans-serif"],
 )
 
+demo_name = "UNEP Q&A"
+
+with gr.Blocks(title=f"{demo_name}", theme=theme, css_paths=os.getcwd()+ "/style.css") as demo:
+ 
+    gr.Markdown(f"<h1><center>{demo_name}</center></h1>")
+
+    with gr.Row():
+        with gr.Column(scale=2):
+            chatbot = gr.Chatbot(
+                value = [("","Hello ! How can I help you today ?")],
+                elem_id="chatbot", 
+                label=f"{demo_name} chatbot", 
+                show_label=False
+            )
+            state = gr.State([])
+ 
+            with gr.Row():
+                ask = gr.Textbox(
+                    show_label=False,
+                    placeholder="Input your question then press enter",
+                )
+            
+        with gr.Column(scale=1, variant="panel"):
+            gr.Markdown("### Sources")
+            sources_textbox = gr.HTML(show_label=False)
+ 
+        ask.submit(
+            fn=chat,
+            inputs=[
+                ask,
+                state,
+            ],
+            outputs=[chatbot, state, sources_textbox],
+        )
+
+    ask.submit(reset_textbox, [], [ask])
 
-if __name__ == "__main__":
-    demo.launch()
+demo.queue()
+demo.launch(
+    share=True,
+    debug=True
+)

prompt.py

@@ -0,0 +1,53 @@
+"""
+Prompt configuration.
+"""
+from datetime import datetime
+from langchain.prompts.prompt import PromptTemplate
+from langchain.prompts import ChatPromptTemplate
+
+from langchain_core.prompts import ChatPromptTemplate
+
+interprate_question_sharepoint_template = """
+    whatever is asked, just answer only {{}}"""
+
+PROMPT_INTERPRATE_INTENTION_SHAREPOINT = ChatPromptTemplate.from_template(
+    interprate_question_sharepoint_template
+)
+
+interprate_question_template = (
+    """You are an assistant that have to identify the object of a question.
+    A user asks a question about meeting decisions. 
+    If the question is about a particular meeting, identified by a meeting number, answer only 'meeting <meeting number>'.
+    Otherwise answer only 'other'.
+    
+    Example:
+    Q: What decision was taken at meeting 123th?
+    R: meeting 123
+    Q: Give me an example of a decision that applied a penalty to a country?
+    R: autre
+    
+    """
+    "La question est la suivante: {query}."
+)
+
+PROMPT_INTERPRATE_INTENTION = ChatPromptTemplate.from_template(
+    interprate_question_template
+)
+
+current_date = datetime.now().strftime('%d/%m/%Y')
+company_name = "UNEP" # to change
+
+answering_template = (
+    f"You are an AI Assistant by Ekimetrics for {company_name}. "
+    f"Your task is to help {company_name} employees. "
+    "You will be given a question and extracted parts of documents."
+    "Provide a clear and structured answer based on the context provided. "
+    "When relevant, use bullet points and lists to structure your answers. "
+    "Whenever you use information from a document, reference it at the end of the sentence (ex: [doc 2]). "
+    "You don't have to use all documents, only if it makes sense in the conversation. "
+    "If no relevant information to answer the question is present in the documents, "
+    "just say you don't have enough information to answer.\n\n"
+    "{context}\n\n"
+    "Question: {question}"
+    )
+ANSWER_PROMPT = ChatPromptTemplate.from_template(answering_template)

requirements.txt

@@ -1 +1,13 @@
-huggingface_hub==0.25.2
+huggingface_hub==0.25.2
+# gradio==5.0.2
+# python-dotenv==1.0.0
+# langchain==0.2.1
+# langchain-community==0.2
+# langchain_openai==0.1.7
+# faiss-cpu==1.9.0
+python-dotenv==1.0.1
+gradio==5.0.2
+langchain==0.3.3
+langchain-community==0.3.2
+langchain-openai==0.2.2
+faiss-cpu==1.9.0

style.css

@@ -0,0 +1,598 @@
+
+/* :root {
+    --user-image: url('https://ih1.redbubble.net/image.4776899543.6215/st,small,507x507-pad,600x600,f8f8f8.jpg');
+  } */
+  .avatar-container.svelte-1x5p6hu:not(.thumbnail-item) img {
+    width: 100%;
+    height: 100%;
+    object-fit: cover;
+    border-radius: 50%;
+    padding: 0px;
+    margin: 0px;
+}
+
+.gradio-container {
+    width: 100%!important;
+    max-width: 100% !important;
+}
+
+/* fix for huggingface infinite growth*/
+main.flex.flex-1.flex-col { 
+    max-height: 95vh !important;
+}
+
+button#show-figures{
+    /* Base styles */
+    background-color: #f5f5f5;
+    border: 1px solid #e0e0e0;
+    border-radius: 4px;
+    color: #333333;
+    cursor: pointer;
+    width: 100%;
+    text-align: center;
+}
+
+.avatar-container.svelte-1x5p6hu:not(.thumbnail-item) img {
+    width: 100%;
+    height: 100%;
+    object-fit: cover;
+    border-radius: 50%;
+    padding: 0px;
+    margin: 0px;
+}
+
+.warning-box {
+    background-color: #fff3cd;
+    border: 1px solid #ffeeba;
+    border-radius: 4px;
+    padding: 15px 20px;
+    font-size: 14px;
+    color: #856404;
+    display: inline-block;
+    margin-bottom: 15px;
+  }
+
+
+.tip-box {
+    background-color: #f0f9ff;
+    border: 1px solid #80d4fa;
+    border-radius: 4px;
+    margin-top:20px;
+    padding: 15px 20px;
+    font-size: 14px;
+    display: inline-block;
+    margin-bottom: 15px;
+    width: auto;
+    color:black !important;
+}
+
+body.dark .warning-box * {
+    color:black !important;
+}
+
+
+body.dark .tip-box * {
+    color:black !important;
+}
+
+
+.tip-box-title {
+    font-weight: bold;
+    font-size: 14px;
+    margin-bottom: 5px;
+}
+
+.light-bulb {
+    display: inline;
+    margin-right: 5px;
+}
+
+.gr-box {border-color: #d6c37c}
+
+#hidden-message{
+    display:none;
+}
+
+.message{
+    font-size:14px !important;
+    
+}
+.card-content img {
+    display: block;
+    margin: auto;
+    max-width: 100%; /* Ensures the image is responsive */
+    height: auto;
+}
+
+a {
+    text-decoration: none;
+    color: inherit;
+}
+
+.doc-ref sup{
+    color:#dc2626!important;
+    /* margin-right:1px; */
+}
+
+
+.card {
+    background-color: white;
+    border-radius: 10px;
+    box-shadow: 0 4px 6px rgba(0, 0, 0, 0.1);
+    overflow: hidden;
+    display: flex;
+    flex-direction: column;
+    margin:20px;
+}
+
+.card-content {
+    padding: 20px;
+}
+
+.card-content h2 {
+    font-size: 14px !important;
+    font-weight: bold;
+    margin-bottom: 10px;
+    margin-top:0px !important;
+    color:#dc2626!important;;
+}
+
+.card-content p {
+    font-size: 12px;
+    margin-bottom: 0;
+}
+
+.card-footer {
+    background-color: #f4f4f4;
+    font-size: 10px;
+    padding: 10px;
+    display: flex;
+    justify-content: space-between;
+    align-items: center;
+}
+
+.card-footer span {
+    flex-grow: 1;
+    text-align: left;
+    color: #999 !important;
+}
+
+.pdf-link {
+    display: inline-flex;
+    align-items: center;
+    margin-left: auto;
+    text-decoration: none!important;
+    font-size: 14px;
+}
+
+
+
+.message.user{
+    /* background-color:#7494b0 !important; */
+    border:none;
+    /* color:white!important; */
+}
+
+.message.bot{
+    /* background-color:#f2f2f7 !important; */
+    border:none;
+}
+
+/* .gallery-item > div:hover{
+    background-color:#7494b0 !important;
+    color:white!important;
+}
+
+.gallery-item:hover{
+    border:#7494b0 !important;
+}
+
+.gallery-item > div{
+    background-color:white !important;
+    color:#577b9b!important;
+}
+
+.label{
+    color:#577b9b!important;
+} */
+
+/* .paginate{
+    color:#577b9b!important;
+} */
+
+
+
+/* span[data-testid="block-info"]{
+    background:none !important;
+    color:#577b9b;
+  } */
+
+/* Pseudo-element for the circularly cropped picture */
+/* .message.bot::before {
+    content: '';
+    position: absolute;
+    top: -10px;
+    left: -10px;
+    width: 30px;
+    height: 30px;
+    background-image: var(--user-image);
+    background-size: cover;
+    background-position: center;
+    border-radius: 50%;
+    z-index: 10;
+  }
+   */
+
+label.selected{
+  background:none !important;
+}
+
+#submit-button{
+    padding:0px !important;
+}
+
+@media screen and (min-width: 1024px) {
+    .gradio-container {
+        max-height: calc(100vh - 190px) !important;
+        overflow: hidden;
+    }
+    /* div#chatbot{
+        height:calc(100vh - 170px) !important;
+        max-height:calc(100vh - 170px) !important;
+
+    } */
+
+    div#tab-examples{
+        height:calc(100vh - 190px) !important;
+        overflow-y: scroll !important;
+        /* overflow-y: auto; */
+    }
+
+    div#sources-textbox{
+        height:calc(100vh - 190px) !important;
+        overflow-y: scroll !important;
+        /* overflow-y: auto !important; */
+    }
+
+    div#sources-figures{
+        height:calc(100vh - 300px) !important;
+        max-height: 90vh !important;
+        overflow-y: scroll !important;
+    }
+
+    div#tab-config{
+        height:calc(100vh - 190px) !important;
+        overflow-y: scroll !important;
+        /* overflow-y: auto !important; */
+    }
+
+    /* Force container to respect height limits */
+    .main-component{
+        contain: size layout;
+        overflow: hidden;
+    }
+
+
+    div#chatbot-row{
+        max-height:calc(100vh - 90px) !important;
+    }
+/*
+
+
+    .max-height{
+        height:calc(100vh - 90px) !important;
+        max-height:calc(100vh - 90px) !important;
+        overflow-y: auto;
+    }
+*/
+
+}
+
+footer {
+    visibility: hidden;
+    display:none !important;
+}
+
+
+@media screen and (max-width: 767px) {
+    /* Your mobile-specific styles go here */
+
+    div#chatbot{
+        height:500px !important;
+    }
+
+    #submit-button{
+        padding:0px !important;
+        min-width: 80px;
+    }
+
+    /* This will hide all list items */
+    div.tab-nav button {
+        display: none !important;
+    }
+
+    /* This will show only the first list item */
+    div.tab-nav button:first-child {
+        display: block !important;
+    }
+    
+    /* This will show only the first list item */
+    div.tab-nav button:nth-child(2) {
+        display: block !important;
+    }
+    
+    #right-panel button{
+        display: block !important;
+    }
+
+    /* ... add other mobile-specific styles ... */
+}
+
+@media (prefers-color-scheme: dark) {
+    .card{
+        background-color: #374151;
+    }
+    .card-image > .card-content{
+        background-color: rgb(55, 65, 81) !important;
+    }
+
+    .card-footer {
+        background-color: #404652;
+    }
+
+    .container > .wrap{
+        background-color: #374151 !important;
+        color:white !important;
+    }
+    .card-content h2{
+        color:#e7754f !important;
+    }
+    .doc-ref sup{
+        color:rgb(235 109 35)!important;
+        /* margin-right:1px; */
+    }
+    .card-footer span {
+        color:white !important;
+    }
+    
+}
+
+
+.doc-ref{
+    color:#dc2626!important;
+    margin-right:1px;
+}
+
+.tabitem{
+    border:none !important;
+}    
+
+.other-tabs > div{
+    padding-left:40px;
+    padding-right:40px;
+    padding-top:10px;
+}
+
+.gallery-item > div{
+    white-space: normal !important; /* Allow the text to wrap */
+    word-break: break-word !important; /* Break words to prevent overflow */
+    overflow-wrap: break-word !important; /* Break long words if necessary */
+  }
+
+span.chatbot > p > img{
+    margin-top:40px !important;
+    max-height: none !important;
+    max-width: 80% !important;
+    border-radius:0px !important;
+}
+
+
+.chatbot-caption{
+    font-size:11px;
+    font-style:italic;
+    color:#508094;
+}
+
+.ai-generated{
+    font-size:11px!important;
+    font-style:italic;
+    color:#73b8d4 !important;
+}
+
+.card-image > .card-content{
+    background-color:#f1f7fa;
+}
+
+
+
+.tab-nav > button.selected{
+    color:#4b8ec3;
+    font-weight:bold;
+    border:none;
+}
+
+.tab-nav{
+    border:none !important;
+}
+
+#input-textbox > label > textarea{
+    border-radius:40px;
+    padding-left:30px;
+    resize:none;
+}
+
+#input-message > div{
+    border:none;
+}
+
+#dropdown-samples{
+  /*! border:none !important; */
+  /*! border-width:0px !important; */
+  background:none !important;
+  
+}
+
+#dropdown-samples > .container > .wrap{
+  background-color:white;
+}
+
+
+#tab-examples > div > .form{
+  border:none;
+  background:none !important;
+}
+
+.a-doc-ref{
+	text-decoration: none !important;
+}
+
+
+.dropdown {
+    position: relative;
+    display:inline-block;
+    margin-bottom: 10px;
+  }
+  
+  .dropdown-toggle {
+    background-color: #f2f2f2;
+    color: black;
+    padding: 10px;
+    font-size: 16px;
+    cursor: pointer;
+    display: block;
+    width: 400px; /* Adjust width as needed */
+    position: relative;
+    display: flex;
+    align-items: center; /* Vertically center the contents */
+    justify-content: left;
+  }
+  
+  .dropdown-toggle .caret {
+    content: "";
+    position: absolute;
+    right: 10px;
+    top: 50%;
+    border-left: 5px solid transparent;
+    border-right: 5px solid transparent;
+    border-top: 5px solid black;
+    transform: translateY(-50%);
+  }
+  
+  input[type="checkbox"] {
+    display: none !important;
+  }
+  
+  input[type="checkbox"]:checked + .dropdown-content {
+    display: block;
+  }
+  
+  .dropdown-content {
+    display: none;
+    position: absolute;
+    background-color: #f9f9f9;
+    min-width: 300px;
+    box-shadow: 0 8px 16px 0 rgba(0,0,0,0.2);
+    z-index: 1;
+    padding: 12px;
+    border: 1px solid #ccc;
+  }
+  
+  input[type="checkbox"]:checked + .dropdown-toggle + .dropdown-content {
+    display: block;
+  }
+  
+  input[type="checkbox"]:checked + .dropdown-toggle .caret {
+    border-top: 0;
+    border-bottom: 5px solid black;
+  }
+  
+  .loader {
+    border: 1px solid #d0d0d0 !important; /* Light grey background */
+    border-top: 1px solid #db3434 !important; /* Blue color */
+    border-right: 1px solid #3498db !important; /* Blue color */
+    border-radius: 50%;
+    width: 20px;    
+    height: 20px;
+    animation: spin 2s linear infinite;
+    display:inline-block;
+    margin-right:10px !important;
+  }
+  
+  .checkmark{
+    color:green !important;
+    font-size:18px;
+    margin-right:10px !important;
+  }
+  
+  @keyframes spin {
+    0% { transform: rotate(0deg); }
+    100% { transform: rotate(360deg); }
+  }
+
+
+  .relevancy-score{
+    margin-top:10px !important;
+    font-size:10px !important;
+    font-style:italic;
+  }
+  
+  .score-green{
+    color:green !important;
+  }
+  
+  .score-orange{
+    color:orange !important;
+  }
+  
+  .score-orange{
+    color:red !important;
+  }
+
+/* Additional style for scrollable tab content */
+div#tab-recommended_content {
+    overflow-y: auto; /* Enable vertical scrolling */
+    max-height: 80vh; /* Adjust height as needed */
+}
+
+/* Mobile specific adjustments */
+@media screen and (max-width: 767px) {
+    div#tab-recommended_content {
+        max-height: 50vh; /* Reduce height for smaller screens */
+        overflow-y: auto;
+    }
+}
+
+/* Additional style for scrollable tab content */
+div#tab-saved-graphs {
+    overflow-y: auto; /* Enable vertical scrolling */
+    max-height: 80vh; /* Adjust height as needed */
+}
+
+/* Mobile specific adjustments */
+@media screen and (max-width: 767px) {
+    div#tab-saved-graphs {
+        max-height: 50vh; /* Reduce height for smaller screens */
+        overflow-y: auto;
+    }
+}
+.message-buttons-left.panel.message-buttons.with-avatar {
+    display: none;
+}
+  .score-red{
+    color:red !important;
+  }
+.message-buttons-left.panel.message-buttons.with-avatar {
+    display: none;
+}
+
+/* Specific fixes for Hugging Face Space iframe */
+.h-full {
+    height: auto !important;
+    min-height: 0 !important;
+}
+
+.space-content {
+    height: auto !important;
+    max-height: 100vh !important;
+    overflow: hidden;
+}

utils.py

@@ -0,0 +1,81 @@
+from langchain_openai import ChatOpenAI
+from langchain.prompts.prompt import PromptTemplate
+from typing import Tuple, List
+from langchain.schema import format_document
+
+
+import gradio as gr
+
+from langchain.chat_models import ChatOpenAI
+import os
+from langchain_openai import ChatOpenAI
+import os
+
+DEFAULT_DOCUMENT_PROMPT = PromptTemplate.from_template(template="{page_content}")
+
+
+def make_pairs(lst):
+    """from a list of even lenght, make tupple pairs"""
+    return [(lst[i], lst[i + 1]) for i in range(0, len(lst), 2)]
+
+def reset_textbox():
+    return gr.update(value="")
+
+def _combine_documents(
+    docs, document_prompt=DEFAULT_DOCUMENT_PROMPT, document_separator="\n\n"
+):
+    doc_strings = [f"Document {i}: \n'''\n{format_document(doc, document_prompt)}\n'''" for i, doc in enumerate(docs, 1)]
+    return document_separator.join(doc_strings)
+
+
+def _format_chat_history(chat_history: List[Tuple]) -> str:
+    buffer = ""
+    for dialogue_turn in chat_history:
+        human = "Human: " + dialogue_turn[0]
+        ai = "Assistant: " + dialogue_turn[1]
+        buffer += "\n" + "\n".join([human, ai])
+    return buffer
+
+def _format_chat_history(chat_history: List[Tuple]) -> str:
+    turn = 1
+    buffer = []
+    for dialogue in chat_history:
+        buffer.append(("Human: " if turn else "Assistant: ") + dialogue.content)
+        turn ^= 1
+    return "\n".join(buffer) + "\n"
+
+def get_llm(model="gpt-4o-mini",max_tokens=1024, temperature=0.0, streaming=True,timeout=30, **kwargs):
+    
+    llm = ChatOpenAI(
+        model=model,
+        api_key=os.environ.get("OPENAI_API_KEY", None),
+        max_tokens = max_tokens,
+        streaming = streaming,
+        temperature=temperature,
+        timeout = timeout,
+        **kwargs,
+    )
+
+    return llm
+
+
+
+def make_html_source(source,i):
+    meta = source.metadata
+    # content = source.page_content.split(":",1)[1].strip()
+    content = source.page_content.strip()
+
+
+    card = f"""
+    <div class="card" id="doc{i}">
+        <div class="card-content">
+            <h2>Document {i} - Meeting {meta["meeting_number"]} - title {meta['Title']} - Issues {meta['Issues']}</h2>
+            <p>{content}</p>
+        </div>
+
+    </div>
+    """
+    return card
+    
+
+