Initial commit

.env.sample

@@ -0,0 +1,5 @@
+# !!! DO NOT UPDATE THIS FILE DIRECTLY. MAKE A COPY AND RENAME IT `.env` TO PROCEED !!! #
+HF_LLM_ENDPOINT="YOUR_LLM_ENDPOINT_URL_HERE"
+HF_EMBED_ENDPOINT="YOUR_EMBED_MODEL_ENDPOINT_URL_HERE"
+HF_TOKEN="YOUR_HF_TOKEN_HERE"
+# !!! DO NOT UPDATE THIS FILE DIRECTLY. MAKE A COPY AND RENAME IT `.env` TO PROCEED !!! #

.gitignore

@@ -0,0 +1,5 @@
+.env
+__pycache__/
+.chainlit
+*.pkl
+.files

Dockerfile

@@ -0,0 +1,27 @@
+FROM python:3.9
+
+RUN pip install --upgrade pip
+
+# Create a user and set up the environment
+RUN useradd -m -u 1000 user
+USER user
+
+ENV HOME=/home/user \
+    PATH=/home/user/.local/bin:$PATH
+
+WORKDIR $HOME/app
+
+# Add this line to copy the data directory
+COPY ./data /home/user/app/data  
+
+# Copy only requirements.txt first to leverage Docker cache
+COPY --chown=user requirements.txt $HOME/app/requirements.txt
+
+# Install dependencies
+RUN pip install --no-cache-dir -r requirements.txt
+
+# Copy the rest of the application code
+COPY --chown=user . $HOME/app
+
+# Run the application
+CMD ["chainlit", "run", "midterm_app.py", "--port", "7860"]

README.md

@@ -0,0 +1,152 @@
+#Note to self: Revise this later
+
+# Week 4: Tuesday
+
+In today's assignment, we'll be creating an Open Source LLM-powered LangChain RAG Application in Chainlit.
+
+There are 2 main sections to this assignment:
+
+## Build 🏗️
+
+### Build Task 1: Deploy LLM and Embedding Model to SageMaker Endpoint Through Hugging Face Inference Endpoints
+
+#### LLM Endpoint
+
+Select "Inference Endpoint" from the "Solutions" button in Hugging Face:
+
+![image](https://i.imgur.com/6KC9TCD.png)
+
+Create a "+ New Endpoint" from the Inference Endpoints dashboard.
+
+![image](https://i.imgur.com/G6Bq9KC.png)
+
+Select the `NousResearch/Meta-Llama-3-8B-Instruct` model repository and name your endpoint. Select N. Virginia as your region (`us-east-1`). Give your endpoint an appropriate name. Make sure to select *at least* a L4 GPU. 
+
+![image](https://i.imgur.com/X3YlUbh.png)
+
+Select the following settings for your `Advanced Configuration`.
+
+![image](https://i.imgur.com/c0HQ7g1.png)
+
+Create a `Protected` endpoint.
+
+![image](https://i.imgur.com/Ak8kchZ.png)
+
+If you were successful, you should see the following screen:
+
+![image](https://i.imgur.com/IBYG3wm.png)
+
+#### Embedding Model Endpoint
+We'll be using `Snowflake/snowflake-arctic-embed-m` for our embedding model today.
+
+The process is the same as the LLM - but we'll make a few specific tweaks:
+
+Let's make sure our set-up reflects the following screenshots:
+
+![image](https://i.imgur.com/IHh8FnC.png)
+
+After which, make sure the advanced configuration is set like so:
+
+![image](https://i.imgur.com/bbcrhUj.png)
+
+> #### NOTE: PLEASE SHUTDOWN YOUR INSTANCES WHEN YOU HAVE COMPLETED THE ASSIGNMENT TO PREVENT UNESSECARY CHARGES.
+
+### Build Task 2: Create RAG Pipeline with LangChain
+
+Follow the [notebook](https://colab.research.google.com/drive/1v1FYmvKH4gsqcdZwIT9wvbQe0GUjrc9d?usp=sharing) to create a LangChain pipeline powered by Hugging Face endpoints!
+
+Once you're done - please move on to Build Task 3!
+
+### Build Task 3: Create a Chainlit Application
+
+1. Create a new empty Docker space through Hugging Face - with the following settings:
+
+![image](https://i.imgur.com/0YzyQX7.png)
+
+> NOTE: You may notice the application builds slowly (~15min.) with the default free-tier hardware. The process will be faster using the `CPU upgrade` Space Hardware - though it is not required. 
+
+2. Clone the newly created space into a directory that is *NOT IN YOUR AI MAKERSPACE REPOSITORY* using the SSH option.
+
+> NOTE: You may need to ensure you've added your SSH key to Hugging Face, as well as GitHub. This should already be done.
+
+![image](https://i.imgur.com/5RyBdP5.png)
+
+3. Copy and Paste (`cp ...` or through UI) the contents of `Week 4/Day 1` into the newly cloned repository. 
+
+> NOTE: Please keep the `README.md` that was cloned from your space and delete the class `README.md`.
+
+4. Using the `ls` command or the `tree` command verify that you have copied over: 
+ - `app.py`
+ - `Dockerfile`
+ - `data/paul_graham_essays.txt`
+ - `chainlit.md`
+ - `.gitignore`
+ - `.env.sample`
+ - `solution_app.py`
+ - `requirements.txt`
+
+ Here is an example as the `ls -al` CLI command: 
+
+ ![image](https://i.imgur.com/vazGYeb.png)
+
+ 5. Work through the `app.py` file to migrate your LCEL LangChain RAG Chain from the Notebook to Chainlit!
+
+ 6. Be sure to modify your `README.md` and `chainlit.md` as you see fit!
+
+ > NOTE: If you get stuck, there is a working reference version in `solution_app.py`.
+
+ 7. When you are done with local testing - push your changes to your space. 
+
+ 8. Make sure you add your `HF_LLM_ENDPOINT`, `HF_EMBED_ENDPOINT`, `HF_TOKEN` as "Secrets" in your Hugging Face Space.
+
+### Terminating Your Resources
+
+Please head to the settings of each endpoint and select `Delete Endpoint`. You will need to type the name of the endpoint to delete the resources.
+
+### Deliverables
+
+- Completed Notebook
+- Chainlit Application in a Hugging Face Space Powered by Hugging Face Endpoints
+- Screenshot of endpoint usage
+
+Example Screen Shot:
+
+![image](https://i.imgur.com/qfbcVpS.png)
+
+## Ship 🚢
+
+Create a Hugging Face Space powered by Hugging Face Endpoints!
+
+### Deliverables
+
+- A short Loom of the space, and a 1min. walkthrough of the application in full
+
+## Share 🚀
+
+Make a social media post about your final application!
+
+### Deliverables
+
+- Make a post on any social media platform about what you built!
+
+Here's a template to get you started:
+
+```
+🚀 Exciting News! 🚀
+
+I am thrilled to announce that I have just built and shipped a open-source LLM-powered Retrieval Augmented Generation Application with LangChain! 🎉🤖
+
+🔍 Three Key Takeaways:
+1️⃣ 
+2️⃣ 
+3️⃣ 
+
+Let's continue pushing the boundaries of what's possible in the world of AI and question-answering. Here's to many more innovations! 🚀
+Shout out to @AIMakerspace !
+
+#LangChain #QuestionAnswering #RetrievalAugmented #Innovation #AI #TechMilestone
+
+Feel free to reach out if you're curious or would like to collaborate on similar projects! 🤝🔥
+```
+
+> #### NOTE: PLEASE SHUTDOWN YOUR INSTANCES WHEN YOU HAVE COMPLETED THE ASSIGNMENT TO PREVENT UNESSECARY CHARGES.

chainlit.md

@@ -0,0 +1,23 @@
+Welcome to the AirBnB 10k filing QnA Bot!
+
+This bot answers questions from Airbnb's Q1 2024 10-K filings to demonstrate the power of AI to process complex financial documents and provide precise insights. It's the midterm assignment for the AI Makerspace AI Engineering Bootcamp. 
+
+My mind is buzzing with the potential to harness this kind of application to drive social impact, and I can't wait to use what I'm learning co-create solutions with nonprofits, social enterprises, and government agencies across the US. 
+
+Here's a bit more on the assignment for those who are interested:
+
+Build 🏗️
+
+Data: Airbnb 10-k Filings from Q1, 2024
+LLM: You decide! (I picked OpenAI.)
+Embedding Model: You decide! (I picked OpenAI.)
+Infrastructure: LangChain
+Vector Store: QDrant
+Deployment: Chainlit, Hugging Face
+
+Ship 🚢
+
+Evaluate your answers to the following questions
+Q1 "What is Airbnb's 'Description of Business'?"
+Q2 "What was the total value of 'Cash and cash equivalents' as of December 31, 2023?"
+Q3 "What is the 'maximum number of shares to be sold under the 10b5-1 Trading plan' by Brian Chesky?"

midterm-app

@@ -0,0 +1 @@
+Subproject commit a6492b2481dbd143f30a0d5ebf707b3b070e7f54

midterm_app.py

@@ -0,0 +1,124 @@
+# Import Required Libraries
+import os
+from dotenv import load_dotenv
+
+import openai
+import fitz  # PyMuPDF
+import pandas as pd
+from transformers import pipeline
+from qdrant_client import QdrantClient
+from qdrant_client.http import models as qdrant_models
+import chainlit as cl
+import tiktoken
+
+# Specific imports from the libraries
+from langchain.document_loaders import PyMuPDFLoader
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.embeddings import OpenAIEmbeddings
+#old import from langchain_openai import OpenAIEmbeddings
+from langchain_community.vectorstores import Qdrant
+from langchain.prompts import ChatPromptTemplate
+from langchain.chat_models import ChatOpenAI
+#old import from langchain_openai import ChatOpenAI
+from operator import itemgetter
+from langchain.schema.output_parser import StrOutputParser
+from langchain.schema.runnable import RunnablePassthrough
+
+# Set Environment Variables
+load_dotenv()
+
+# Load environment variables
+OPENAI_API_KEY = os.getenv("OPENAI_API_KEY")
+
+# Initialize OpenAI client after loading the environment variables
+openai.api_key = OPENAI_API_KEY
+
+# Load and split documents
+loader = PyMuPDFLoader("/home/user/app/data/airbnb_q1_2024.pdf")
+#old file path is loader = PyMuPDFLoader("/Users/sampazar/AIE3-Midterm/data/airbnb_q1_2024.pdf")
+documents = loader.load()
+
+def tiktoken_len(text):
+    tokens = tiktoken.encoding_for_model("gpt-4o").encode(text)
+    return len(tokens)
+
+text_splitter = RecursiveCharacterTextSplitter(
+    chunk_size=150, 
+    chunk_overlap=100,
+    length_function = tiktoken_len
+)
+
+split_chunks = text_splitter.split_documents(documents)
+
+
+# Load OpenAI Embeddings Model
+embeddings = OpenAIEmbeddings(model="text-embedding-3-small")
+
+# Creating a Qdrant Vector Store
+qdrant_vector_store = Qdrant.from_documents(
+    split_chunks,
+    embeddings,
+    location=":memory:",
+    collection_name="Airbnb_Q1_2024",
+)
+
+# Create a Retriever
+retriever = qdrant_vector_store.as_retriever()
+
+# Create a prompt template
+template = """Answer the question based only on the following context. If you cannot answer the question with the context, please respond with 'I don't know':
+
+Context:
+{context}
+
+Question:
+{question}
+"""
+
+prompt = ChatPromptTemplate.from_template(template)
+
+# Define the primary LLM
+primary_llm = ChatOpenAI(model_name="gpt-4o", temperature=0)
+
+# Creating a Retrieval Augmented Generation (RAG) Chain
+retrieval_augmented_qa_chain = (
+    # INVOKE CHAIN WITH: {"question" : "<>"}
+    # "question" : populated by getting the value of the "question" key
+    # "context"  : populated by getting the value of the "question" key and chaining it into the base_retriever
+    {"context": itemgetter("question") | retriever, "question": itemgetter("question")}
+    # "context"  : is assigned to a RunnablePassthrough object (will not be called or considered in the next step)
+    #              by getting the value of the "context" key from the previous step
+    | RunnablePassthrough.assign(context=itemgetter("context"))
+    # "response" : the "context" and "question" values are used to format our prompt object and then piped
+    #              into the LLM and stored in a key called "response"
+    # "context"  : populated by getting the value of the "context" key from the previous step
+    | {"response": prompt | primary_llm, "context": itemgetter("context")}
+)
+
+# Chainlit integration for deployment
+@cl.on_chat_start  # marks a function that will be executed at the start of a user session
+async def start_chat():
+    settings = {
+        "model": "gpt-4o",
+        "temperature": 0,
+        "max_tokens": 500,
+        "top_p": 1,
+        "frequency_penalty": 0,
+        "presence_penalty": 0,
+    }
+    cl.user_session.set("settings", settings)
+
+@cl.on_message  # marks a function that should be run each time the chatbot receives a message from a user
+async def handle_message(message: cl.Message):
+    settings = cl.user_session.get("settings")
+
+    response = retrieval_augmented_qa_chain.invoke({"question": message.content})
+    
+    #msg = cl.Message(content=response["response"])
+    #await msg.send()
+
+    # Extracting and sending just the content
+    content = response["response"].content
+    pretty_content = content.strip()  # Remove any leading/trailing whitespace
+
+    await cl.Message(content=pretty_content).send()

requirements.txt

@@ -0,0 +1,21 @@
+chainlit==0.7.700
+langchain==0.2.5
+langchain_community==0.2.5
+langchain_core==0.2.9
+langchain_text_splitters==0.2.1
+python-dotenv==1.0.1
+
+#Adding OpenAI API client and Qdrant client
+openai==1.35.3 #Be sure to use the latest version 'pip show openai'
+qdrant-client==1.9.2 #Be sure to use the latest version 'pip show qdrant-client'
+
+# Adding PyMuPDF for PDF processing
+PyMuPDF==1.24.5 #Be sure to use the latest version 'pip show pymupdf'
+
+tiktoken==0.7.0
+#cohere==4.37
+transformers==4.37.0
+pandas==2.0.3
+#Removed Hugging Face and FAISS dependencies
+#langchain_huggingface==0.0.3
+#faiss-cpu