Upload 3 files

Dockerfile

@@ -0,0 +1,23 @@
+
+FROM python:3.9-slim
+
+WORKDIR /app
+
+RUN apt-get update && apt-get install -y \
+    build-essential \
+    curl \
+    software-properties-common \
+    git \
+    && rm -rf /var/lib/apt/lists/*
+
+COPY requirements.txt ./
+COPY app.py ./
+COPY src/ ./src/
+
+RUN pip3 install -r requirements.txt
+
+EXPOSE 8501
+
+HEALTHCHECK CMD curl --fail http://localhost:8501/_stcore/health
+
+ENTRYPOINT ["streamlit", "run", "app.py", "--server.port=8501", "--server.address=0.0.0.0","--server.enableXsrfProtection=false"]

app.py

@@ -0,0 +1,139 @@
+
+import streamlit as st
+import os
+import json
+import requests
+from langchain_community.document_loaders import PyMuPDFLoader
+from openai import OpenAI
+import tiktoken
+import pandas as pd
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain_community.embeddings.openai import OpenAIEmbeddings
+from langchain_community.vectorstores import Chroma
+import tempfile
+
+
+OPENAI_API_KEY = os.environ.get("API_KEY")
+OPENAI_API_BASE = os.environ.get("API_BASE")
+
+# Initialize OpenAI client
+client = OpenAI(
+    api_key=OPENAI_API_KEY,
+    base_url=OPENAI_API_BASE
+)
+
+# Define the system prompt for the model
+qna_system_message = """
+# You are an AI assistant designed to support the HR team at Flykite Airlines. Your task is to provide evidence-based, concise, and relevant answers to employee queries based on the context provided.
+
+User input will include the necessary context for you to answer their questions. This context will begin with the token: ###Context. The context contains references to specific portions of the official Flykite Airlines HR Policy Handbook and related documentation.
+
+When crafting your response:
+1. Use only the provided context to answer the question.
+2. If the answer is found in the context, respond with concise and actionable HR policy information.
+3. Include the source reference with the section name, heading, or clause number, as provided in the context.
+4. If the question is unrelated to the context or the context is empty, clearly respond with: "Sorry, this is out of my knowledge base."
+
+Please adhere to the following response guidelines:
+- Provide clear, direct answers using only the given context.
+- Do not include any additional information outside of the context.
+- Avoid rephrasing or summarizing the context unless explicitly relevant to the question.
+- If no relevant answer exists in the context, respond with: "Sorry, this is out of my knowledge base."
+- If the context is not provided, your response should also be: "Sorry, this is out of my knowledge base."
+
+Here is an example of how to structure your response:
+
+Answer:
+[Answer based on context]
+
+Source:
+[Source details with section, clause, or heading]
+"""
+
+# Define the user message template
+qna_user_message_template = """
+# ###Context
+Here are some excerpts from the Flykite Airlines HR Policy Handbook and their sources that are relevant to the employee's question mentioned below:
+{context}
+
+###Question
+{question}
+"""
+
+# Processing PDF files
+@st.cache_resource
+def load_and_process_pdfs(uploaded_files):
+    all_documents = []
+    for uploaded_file in uploaded_files:
+        with tempfile.NamedTemporaryFile(delete=False, suffix=".pdf") as tmp_file:
+            tmp_file.write(uploaded_file.getvalue())
+            tmp_file_path = tmp_file.name
+        loader = PyMuPDFLoader(tmp_file_path)
+        documents = loader.load()
+        all_documents.extend(documents)
+        os.remove(tmp_file_path) # Clean up the temporary file
+    text_splitter = RecursiveCharacterTextSplitter.from_tiktoken_encoder(
+        encoding_name='cl100k_base',
+        chunk_size=1000,
+    )
+    document_chunks = text_splitter.split_documents(all_documents)
+
+    embedding_model = OpenAIEmbeddings(
+        openai_api_key=OPENAI_API_KEY,
+        openai_api_base=OPENAI_API_BASE
+    )
+
+    # Create an in-memory vector store (or use a persistent one if needed)
+    vectorstore = Chroma.from_documents(
+        document_chunks,
+        embedding_model
+    )
+    return vectorstore.as_retriever(search_type='similarity', search_kwargs={'k': 3})
+
+def generate_rag_response(user_input, retriever, max_tokens=500, temperature=0, top_p=0.95):
+    # Retrieve relevant document chunks
+    relevant_document_chunks = retriever.get_relevant_documents(query=user_input)
+    context_list = [d.page_content for d in relevant_document_chunks]
+
+    # Combine document chunks into a single context
+    context_for_query = ". ".join(context_list)
+
+    user_message = qna_user_message_template.replace('{context}', context_for_query)
+    user_message = user_message.replace('{question}', user_input)
+
+    # Generate the response
+    try:
+        response = client.chat.completions.create(
+            model="gpt-4o-mini",
+            messages=[
+                {"role": "system", "content": qna_system_message},
+                {"role": "user", "content": user_message}
+            ],
+            max_tokens=max_tokens,
+            temperature=temperature,
+            top_p=top_p
+        )
+        response = response.choices[0].message.content.strip()
+    except Exception as e:
+        response = f'Sorry, I encountered the following error: \n {e}'
+
+    return response
+
+# Streamlit App
+st.title("LLM-Powered Support bot")
+
+uploaded_files = st.file_uploader("Upload PDF files", type=["pdf"], accept_multiple_files=True)
+
+retriever = None
+if uploaded_files:
+    st.info("Processing uploaded PDFs...")
+    retriever = load_and_process_pdfs(uploaded_files)
+    st.success("PDFs processed and ready for questioning!")
+
+
+if retriever:
+    user_question = st.text_input("Ask a question about the uploaded documents:")
+    if user_question:
+        with st.spinner("Generating response..."):
+            rag_response = generate_rag_response(user_question, retriever)
+            st.write(rag_response)

requirements.txt

@@ -0,0 +1,11 @@
+
+langchain_community==0.3.27
+langchain==0.3.27
+chromadb==1.0.15
+pymupdf==1.26.3
+tiktoken==0.9.0
+datasets==4.0.0
+evaluate==0.4.5
+streamlit==1.35.0
+openai==1.99.1
+langchain_openai==0.3.28