Build a RAG App — Chat with Your PDFs

Build a command-line app that loads a PDF and answers questions about it using Google Gemini and vector search.

Intermediate

3 hours

5 lessons

Course Content

30 chapters across 5 lessons

Set Up Your Environment

Install the required tools and load your first PDF

5 chapters

Progress0%

What Is RAG and Why It Works

Continue

Understand the retrieve-then-generate pattern before writing any code

Create a Project Folder and Virtual Environment

Set up an isolated Python environment for your RAG app

Install Dependencies

Install the four libraries your RAG app needs

Store Your API Key Safely

Create a .env file and add it to .gitignore

Lesson Complete

You set up your environment and loaded your first PDF

Extract and Chunk Text

Pull raw text from a PDF and split it into manageable pieces

5 chapters

Progress0%

Why Chunking Matters

Understand token limits and the role of overlap before writing code

Extract Raw Text from a PDF

Iterate over every page and join the text into one string

Split Text into Fixed-Size Chunks with Overlap

Write the chunking function that divides text into overlapping pieces

Preview Your Chunks

Print the chunk count and inspect the first chunk

Lesson Complete

You extracted text and split it into chunks

Generate Embeddings

Convert text chunks into vectors using the Gemini API

5 chapters

Progress0%

What Embeddings Are

Understand vectors as meaning in number form — no linear algebra required

Create the Gemini Client

Read the API key from the environment and instantiate the Gemini client

Embed a Single Chunk

Call the Gemini embedding API and return the vector for one piece of text

Embed All Chunks

Loop over every chunk in batches to stay within the free tier rate limit

Lesson Complete

You converted text chunks into vectors using the Gemini API

Store and Search Vectors

Implement cosine similarity and retrieve the most relevant chunks

5 chapters

Progress0%

How Similarity Search Works

Understand cosine similarity with a plain-language explanation

Implement Cosine Similarity

Write the function that scores how similar two vectors are

Write the Search Function

Score every chunk against a query and return the top matches

Test the Search

Run the full extract-chunk-embed-search pipeline and print results

Lesson Complete

You implemented vector search and retrieved relevant chunks

Generate Answers and Finish the App

Build a prompt from retrieved chunks, ask Gemini to answer, and wire everything into a CLI app

10 chapters

Progress0%

How RAG Assembles the Final Prompt

Understand the prompt template before writing the generation code

Build the Prompt

Assemble the system instruction, context chunks, and question into one string

Call Gemini to Generate an Answer

Send the prompt to Gemini Flash and return the response text

Print the Answer and Sources

Display the generated answer, with optional source passages

Cache Embeddings to Disk

Save embeddings to a JSON file so the app skips re-embedding on subsequent runs

Parse Command-Line Arguments

Read the PDF path and question from the terminal using sys.argv

Load or Embed

Check for a cached embedding file and skip the API if one exists

Search and Answer

Wire up the final four calls — search, prompt, generate, and print

Run the Finished App

Execute app.py against a PDF and see the answer in your terminal

Lesson Complete

You built a working RAG app that answers questions about any PDF