Dev

project-UI/chainlitmain.py

@@ -0,0 +1,109 @@
+import chainlit as cl
+from langchain_core.prompts import PromptTemplate
+from langchain_core.runnables import RunnablePassthrough
+from langchain_core.output_parsers import StrOutputParser
+from langchain_openai import ChatOpenAI
+from langchain.vectorstores import FAISS
+from langchain.embeddings import HuggingFaceEmbeddings
+import os
+import json
+
+os.environ["OPENAI_API_KEY"] = "OPENAI_API_KEY"  
+
+# Initialize Embeddings and FAISS
+modelPath = "sentence-transformers/all-MiniLM-l6-v2"
+
+embeddings = HuggingFaceEmbeddings(
+    model_name=modelPath,
+    model_kwargs={'device': 'cpu'},
+    encode_kwargs={'normalize_embeddings': False}
+)
+
+# Load FAISS Index
+loaded_db = FAISS.load_local("code/faiss_index", embeddings, allow_dangerous_deserialization=True)
+retriever = loaded_db.as_retriever(search_type="similarity", search_kwargs={"k": 4})
+
+# LLM Configuration
+llm = ChatOpenAI(model="gpt-4o-mini")
+
+# Prompt Template
+template = """
+You are a highly knowledgeable and professional policy advisor for Boston Public Schools. Your role is to provide precise, context-specific answers to questions based solely on the information provided in the context. 
+
+### Guidelines:
+1. **Scope Limitation**: Only use the information provided in the "Context" to answer the question. Do not infer, assume, or incorporate external information.
+2. **Out-of-Scope Questions**: If a question is unrelated to any policy, politely respond that it is beyond the scope of your knowledge as a policy advisor and feel free to continue the answer based on the "question". Finally, append "[0]__[0]" at the end of the answer for the developer to use, only if the "question" is unrelated to the task. 
+3. **Citing Policy**: Always conclude your response by explicitly citing the policy name(s) used to formulate your answer. If no policy is applicable, don't mention anything.
+
+### Additional Considerations:
+- **Ambiguities**: If the context lacks sufficient information to answer the question definitively, mention this and provide a response based on the provided context.
+- **Clarity and Professionalism**: Ensure all responses are concise but comprehensive, clear, and professional.
+
+### Input Structure:
+Context: {context}
+Question: {question}
+"""
+
+custom_rag_prompt = PromptTemplate.from_template(template)
+
+def format_docs(docs):
+    return "\n\n".join(
+        f"{doc.page_content}\n\nSource: {doc.metadata.get('file_name', 'Unknown File')}, Folder: {doc.metadata.get('folder_name', 'Unknown Folder')}"
+        for doc in docs
+    )
+
+rag_chain = (
+    {
+        "context": retriever | format_docs, 
+        "question": RunnablePassthrough()
+    }
+    | custom_rag_prompt
+    | llm
+    | StrOutputParser()
+)
+
+@cl.on_chat_start
+async def on_chat_start():
+    """
+    Initializes the chatbot session.
+    """
+    await cl.Message(
+        content="Hi! I am the policy advisor for Boston Public School. How can I assist you today?"
+    ).send()
+
+@cl.on_message
+async def on_message(message: cl.Message):
+    """
+    Handles user queries and returns answers based on retrieved documents.
+    """
+    try:
+        # Invoke the RAG chain to get the main response
+        response = rag_chain.invoke(message.content)
+
+        marker = "[0]__[0]"
+        # Check if the response contains the marker
+        if marker in response:
+            # Remove the marker from the response
+            full_response = response.replace(marker, "").strip()
+        else:
+            # Retrieve relevant documents
+            retrieved_documents = retriever.invoke(message.content)
+
+            # Initialize a list to collect formatted links
+            links = []
+            for i, item in enumerate(retrieved_documents, 1):  # Iterate through the list of Document objects
+                link = item.metadata.get('link')  # Access the 'link' key within the 'metadata' attribute
+                file_name = item.metadata.get('file_name')  # Access the 'file_name' key within the 'metadata' attribute
+                if link:
+                    # Append each link in Markdown format for clickability
+                    links.append(f"**Source {i}:** [{file_name}]({link})")
+
+            # Join the response with links, separating links by a new line
+            full_response = f"{response}\n\n**References:**\n" + "\n".join(links) if links else response
+
+        # Send the response back to the user
+        await cl.Message(content=full_response).send()
+
+    except Exception as e:
+        # Send the error message back to the user
+        await cl.Message(content=f"An error occurred: {str(e)}").send()

project-UI/main.py

@@ -0,0 +1,91 @@
+from fastapi import FastAPI
+from fastapi.responses import HTMLResponse
+from fastapi.staticfiles import StaticFiles
+from chainlit.utils import mount_chainlit
+
+app = FastAPI()
+
+# Mount static files directory
+app.mount("/static", StaticFiles(directory="code/static"), name="static")
+
+# Mount Chainlit app at /chainlit
+mount_chainlit(app=app, target="code/chainlitmain.py", path="/chainlit")
+
+# Serve the main webpage
+@app.get("/", response_class=HTMLResponse)
+async def get_homepage():
+    html_content = """
+    <!DOCTYPE html>
+    <html>
+    <head>
+        <title>FastAPI with Chainlit</title>
+        <style>
+            body {
+                font-family: Arial, sans-serif;
+                text-align: center;
+                margin-top: 50px;
+            }
+            #chat-popup {
+                display: none;
+                position: fixed;
+                bottom: 80px;
+                right: 20px;
+                width: 400px;
+                height: 600px;
+                border: 2px solid #ccc;
+                background: #fff;
+                z-index: 1000;
+                box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2);
+                transition: all 0.3s ease;
+                border-radius: 10px;
+            }
+            iframe {
+                width: 100%;
+                height: 100%;
+                border: none;
+                border-radius: 10px;
+            }
+            #chat-button {
+                position: fixed;
+                bottom: 20px;
+                right: 20px;
+                background-color: #007bff;
+                color: white;
+                border: none;
+                padding: 10px 20px;
+                font-size: 16px;
+                border-radius: 5px;
+                cursor: pointer;
+                box-shadow: 0 4px 8px rgba(0, 0, 0, 0.2);
+                z-index: 1001;
+            }
+            #chat-button img {
+                width: 30px;  /* Increased from 10px */
+                height: auto;
+            }
+        </style>
+        <script>
+            window.onload = function() {
+                var popup = document.getElementById("chat-popup");
+                popup.style.display = "none";
+            }
+
+            function toggleChat() {
+                var popup = document.getElementById("chat-popup");
+                var currentDisplay = window.getComputedStyle(popup).display;
+                popup.style.display = currentDisplay === "none" ? "block" : "none";
+            }
+        </script>
+    </head>
+    <body>
+        <h1>Welcome to Boston Public School Policy Chatbot</h1>
+        <div id="chat-button" onclick="toggleChat()" style="cursor: pointer;">
+            <img src="/static/images/chainlit.png" alt="Chat with Chainlit">
+        </div>
+        <div id="chat-popup">
+            <iframe src="/chainlit"></iframe>
+        </div>
+    </body>
+    </html>
+    """
+    return HTMLResponse(content=html_content)