Google_L400_Labs_Code_Solutions

Build and Deploy a Generative AI solution using a RAG framework

⚠️ Warning: Make sure to replace all instances of PROJECT_ID = "qwiklabs-gcp-00-e4970b8c386a" with your own Google Cloud project ID. Failing to do so will result in errors or unintended behavior when accessing Google Cloud resources.

Cymbal Ingest to Vector Database

This guide provides step-by-step instructions on how to ingest documents into a vector database, create embeddings, and set up a Flask API for interacting with the data. This process involves integrating various Google Cloud services, such as Firestore, Vertex AI, and Cloud Run.

Required Libraries and Modules

import vertexai
from vertexai.language_models import TextEmbeddingModel
from vertexai.generative_models import GenerativeModel
import pickle
from IPython.display import display, Markdown
from langchain_google_vertexai import VertexAIEmbeddings
from langchain_community.document_loaders import PyMuPDFLoader
from langchain_experimental.text_splitter import SemanticChunker
from google.cloud import firestore
from google.cloud.firestore_v1.vector import Vector
from google.cloud.firestore_v1.base_vector_query import DistanceMeasure

Explanation

This code imports the necessary modules and libraries. It includes:

• vertexai: For utilizing Google Cloud's Vertex AI functionalities.
• pickle: For loading and saving serialized data.
• IPython.display: For displaying Markdown content in notebooks.
• Firestore: To create and manage a vector database using Firestore.
• LangChain: For integrating with the Vertex AI Embeddings model.

Setting Up the Project and Embedding Model

PROJECT_ID = "qwiklabs-gcp-00-e4970b8c386a"
LOCATION = "us-central1"

vertexai.init(project=PROJECT_ID, location=LOCATION)
embedding_model = VertexAIEmbeddings(model_name="text-embedding-004")

Explanation

• Initializes Vertex AI using your project ID and location.
• Sets up the embedding model text-embedding-004 for generating embeddings.

Loading and Cleaning the Document

!gcloud storage cp gs://partner-genai-bucket/genai069/nyc_food_safety_manual.pdf .

loader = PyMuPDFLoader("./nyc_food_safety_manual.pdf")
data = loader.load()

def clean_page(page):
  return page.page_content.replace("-\n","")\
                          .replace("\n"," ")\
                          .replace("\x02","")\
                          .replace("\x03","")\
                          .replace("fo d P R O T E C T I O N  T R A I N I N G  M A N U A L","")\
                          .replace("N E W  Y O R K  C I T Y  D E P A R T M E N T  O F  H E A L T H  &  M E N T A L  H Y G I E N E","")

Explanation

• Downloads the PDF document from a Google Cloud Storage bucket.
• Uses PyMuPDF to load the document and defines a clean_page function to remove unwanted characters and text from each page.

Chunking and Creating Embeddings

cleaned_pages = []
for pages in data:
  cleaned_pages.append(clean_page(pages))

text_splitter = SemanticChunker(embedding_model)
docs = text_splitter.create_documents(cleaned_pages[0:4])
chunked_content = [doc.page_content for doc in docs]
chunked_embeddings = embedding_model.embed_documents(chunked_content)

Explanation

• Cleans the document pages and uses the SemanticChunker to split the text into meaningful chunks.
• Generates embeddings for each chunk using the Vertex AI Embedding model.

Storing in Firestore

db = firestore.Client(project=PROJECT_ID)
collection = db.collection('food-safety')

for i, (content, embedding) in enumerate(zip(chunked_content, chunked_embeddings)):
    doc_ref = collection.document(f"doc_{i}")
    doc_ref.set({
        "content": content,
        "embedding": Vector(embedding)
    })

Explanation

• Creates a Firestore client and sets up a collection named food-safety.
• Iterates through the chunked content and embeddings, adding them as documents in Firestore.

Vector Search Function

def search_vector_database(query: str):
  query_embedding = embedding_model.embed_query(query)
  vector_query = collection.find_nearest(
    vector_field="embedding",
    query_vector=Vector(query_embedding),
    distance_measure=DistanceMeasure.EUCLIDEAN,
    limit=5,
  )
  docs = vector_query.stream()
  context = [result.to_dict()['content'] for result in docs]
  return context

Explanation

• This function searches the vector database for the closest matching documents based on the query.
• Retrieves the top 5 documents using the Euclidean distance measure and returns the matching content.

Flask API Setup (main.py)

import os
import json
import logging
import google.cloud.logging
from flask import Flask, render_template, request
from google.cloud import firestore
from google.cloud.firestore_v1.vector import Vector
from google.cloud.firestore_v1.base_vector_query import DistanceMeasure
import vertexai
from vertexai.generative_models import GenerativeModel, GenerationConfig
from langchain_google_vertexai import VertexAIEmbeddings

Explanation

• Imports the necessary libraries for setting up a Flask API.
• Sets up Google Cloud logging and Firestore connection.

Search and Response Function

def ask_gemini(question):
    prompt_template = "Using the context provided below, answer the following question:\nContext: {context}\nQuestion: {question}\nAnswer:"
    context = search_vector_database(question)
    formatted_prompt = prompt_template.format(context=context, question=question)
    generation_config = GenerationConfig(
        temperature=0.7,
        max_output_tokens=256,
        response_mime_type="application/json",
    )
    contents = [{"role": "user", "parts": [{"text": formatted_prompt}]}]
    response = gen_model.generate_content(
        contents=contents,
        generation_config=generation_config
    )
    response_text = response.text if response else "{}"
    try:
        response_json = json.loads(response_text)
        answer = response_json.get("answer", "No answer found.")
    except json.JSONDecodeError:
        answer = "Error: Unable to parse response."
    return answer

Explanation

• The ask_gemini function utilizes the Gemini model to answer user queries using the context retrieved from the vector database.
• It formats a prompt with the context and question, sends it to the generative model, and returns the model’s response.

Docker and Cloud Run Deployment

Dockerfile

docker build -t cymbal-docker-image -f Dockerfile .
docker tag cymbal-docker-image us-central1-docker.pkg.dev/qwiklabs-gcp-01-a365df012da7/cymbal-artifact-repo/cymbal-docker-image
gcloud auth configure-docker us-central1-docker.pkg.dev
docker push us-central1-docker.pkg.dev/qwiklabs-gcp-01-a365df012da7/cymbal-artifact-repo/cymbal-docker-image

Explanation

• Builds the Docker image for the application and tags it with the artifact repository URL.
• Configures Docker to use gcloud credentials and pushes the image to Google Cloud's Artifact Registry.

Cloud Run Deployment

gcloud run deploy cymbal-freshbot     --image=us-central1-docker.pkg.dev/qwiklabs-gcp-01-a365df012da7/cymbal-artifact-repo/cymbal-docker-image     --platform=managed     --region=us-central1     --allow-unauthenticated

Explanation

• Deploys the Docker image to Cloud Run, allowing unauthenticated access to the service.