Skip to content

A curated collection of practical AI projects implementing OCR systems, RAG, AI agents, and other AI use cases.

audit: fix image_question_answering - add env comment, rewrite em dashes, add module docstring, add docstrings

Tiioluwani5bdbc4bf66cb
diff --git a/multimodal/image_question_answering/.env.example b/multimodal/image_question_answering/.env.example
index fa40a94..1fd7173 100644
--- a/multimodal/image_question_answering/.env.example
+++ b/multimodal/image_question_answering/.env.example
@@ -1 +1,3 @@
-GOOGLE_API_KEY=your_api_key_here
+# Google AI Studio API key (required)
+# Get your free key at https://aistudio.google.com
+GOOGLE_API_KEY=your_api_key_here
diff --git a/multimodal/image_question_answering/README.md b/multimodal/image_question_answering/README.md
index 8d50626..a563026 100644
--- a/multimodal/image_question_answering/README.md
+++ b/multimodal/image_question_answering/README.md
@@ -4,7 +4,7 @@
## Overview
-Upload a PDF, select a page, and ask a natural-language question about its visual content. PyMuPDF renders each page into a high-quality image, which is sent alongside your question to Gemma 4 (`gemma-4-31b-it`) via the Gemini API. The model reasons over the rendered image — including charts, tables, diagrams, and figures — and returns a grounded answer in the Gradio interface.
+Upload a PDF, select a page, and ask a natural-language question about its visual content. PyMuPDF renders each page into a high-quality image, which is sent alongside your question to Gemma 4 (`gemma-4-31b-it`) via the Gemini API. The model reasons over the rendered image, including charts, tables, diagrams, and figures, and returns a grounded answer in the Gradio interface.
## Demo
@@ -16,7 +16,7 @@ Upload a PDF, select a page, and ask a natural-language question about its visua
- Page slider lets you navigate and preview any page before asking
- Answers visual questions grounded in the actual rendered content (charts, tables, figures)
- Gemma 4 thinking mode (`high`) for multi-step reasoning over complex layouts
-- Clean two-column Gradio UI — page viewer on the left, Q&A on the right
+- Clean two-column Gradio UI with the page viewer on the left and Q&A on the right
## Tech Stack
@@ -88,7 +88,7 @@ Create a `.env` file in the project root with the following key:
| Variable | Description |
|---|---|
-| `GOOGLE_API_KEY` | Gemini API key — obtain one at [aistudio.google.com](https://aistudio.google.com) |
+| `GOOGLE_API_KEY` | Gemini API key; obtain one at [aistudio.google.com](https://aistudio.google.com) |
```env
GOOGLE_API_KEY=your_api_key_here
@@ -103,3 +103,7 @@ image-question-answering/
├── .env.example # Environment variable template
└── requirements.txt
```
+
+---
+
+[Back to Top](#image-question-answering)
diff --git a/multimodal/image_question_answering/app.py b/multimodal/image_question_answering/app.py
index 1919fb8..528fc24 100644
--- a/multimodal/image_question_answering/app.py
+++ b/multimodal/image_question_answering/app.py
@@ -1,3 +1,5 @@
+"""Image Question Answering: upload a PDF, select a page, and ask visual questions answered by Gemma 4 via the Gemini API."""
+
import io
import os
@@ -23,6 +25,7 @@ _page_images: list[Image.Image] = []
def pdf_to_images(path: str, dpi: int = 150) -> list[Image.Image]:
+ """Render every page of the PDF at the given DPI and return them as a list of PIL images."""
doc = fitz.open(path)
zoom = dpi / 72.0
mat = fitz.Matrix(zoom, zoom)
@@ -38,6 +41,7 @@ def pdf_to_images(path: str, dpi: int = 150) -> list[Image.Image]:
def on_pdf_upload(pdf_file):
+ """Handle a PDF upload by rendering all pages and initialising the page slider."""
global _page_images
if pdf_file is None:
@@ -62,6 +66,7 @@ def on_pdf_upload(pdf_file):
def on_page_change(page_num: int):
+ """Return the PIL image for the selected 1-based page number."""
idx = int(page_num) - 1
if _page_images and 0 <= idx < len(_page_images):
return _page_images[idx]
@@ -69,6 +74,7 @@ def on_page_change(page_num: int):
def on_submit(page_num: int, question: str):
+ """Send the selected page image and question to Gemma 4 and return the answer text."""
if not _page_images:
return "Please upload a PDF first."
if not question.strip():