The Dr. Who Oracle

A retrieval-augmented chatbot that answers questions about the Doctor Who universe using Wikipedia data.

The Oracle combines vector search, cross-encoder re-ranking, and a large language model to produce concise answers with citations. A desktop interface built with Tkinter allows users to ask questions interactively and inspect the sources used to generate each response.

---

Overview

The Dr. Who Oracle is a Retrieval-Augmented Generation (RAG) system designed to answer questions about Doctor Who lore.

Instead of relying solely on a language model's internal knowledge, the system retrieves relevant documents from a locally stored vector database constructed from Wikipedia pages. These documents are then re-ranked and supplied to the language model as context when generating an answer.

Each response includes links to the sources used by the model so users can verify the information.

---

Features

• Interactive desktop interface built with Tkinter
• Retrieval-Augmented Generation pipeline
• FAISS vector database for fast similarity search
• Cross-encoder re-ranking for improved document relevance
• Concise answers with explicit citations
• Separate Sources tab showing the documents used to generate each response

---

Launch Screen

---

Example Interaction

The Oracle answering questions about Doctor Who lore:

---

Source Attribution

Every answer is accompanied by citations to the underlying documents used by the model.

---

Architecture

The system follows a standard Retrieval-Augmented Generation pipeline:

User Question
↓
Vector Search (FAISS)
↓
Retrieve Top Documents
↓
Cross-Encoder Re-Ranking
↓
Select Most Relevant Context
↓
LLM Generation
↓
Answer + Citations

---

Project Structure

DrWhoOracle/

Dr_Who_Oracle.py
Tkinter application and chatbot interface

Dr_Who_FAISS_VectorStore_Create.py
Script used to build the FAISS vector database

DATA/Wiki_Data/
Wikipedia text files used to create the vectorstore

Images/
Intro_Screen.png
Oracle_Tab.png
Sources_Tab.png

README.md

---

Setting the API Key

The project requires an OpenAI API key.

Set the environment variable before running the program:

export OPENAI_API_KEY="your-api-key"

Verify that the variable is set:

echo $OPENAI_API_KEY

---

Building the Vector Database

The FAISS vector database can be created from the Wikipedia dataset using:

python Dr_Who_FAISS_VectorStore_Create.py

This script:

1. Loads the documents from DATA/Wiki_Data/
2. Splits them into overlapping text chunks
3. Generates embeddings
4. Stores the results in a FAISS vector database

---

Running the Oracle

Launch the chatbot with:

python Dr_Who_Oracle.py

Click Enter on the launch screen to open the chat interface.

---

Example Questions

Try asking the Oracle questions such as:

• Why did Donna Noble stop traveling with the Doctor?
• What is the first appearance of K-9?
• Who founded Time Lord society?
• Didn't the Twelfth Doctor meet Davros as a child?

---

Technologies Used

• Python
• Tkinter
• LangChain
• FAISS
• SentenceTransformers Cross-Encoder
• OpenAI embeddings and chat models

---

Image Credits

Some images used in the user interface were obtained from publicly available sources.

• tardis_thumb.png
  Source: https://www.flaticon.com/free-icon/tardis_1600954

• gallif_resize.png
  Source: https://www.clipartmax.com/max/m2H7d3Z5H7d3d3d3/

• gallifreyan2_resize.png
  Source: https://favpng.com/png_view/doctor-tenth-doctor-rassilon-gallifrey-time-lord-png/raFa2RUm

• drwhobg_resize.png
  Source: https://www.pxfuel.com/en/desktop-wallpaper-qgwrs

Images remain the property of their respective copyright holders and are used here for demonstration purposes.

---

Disclaimer

This project is intended as a demonstration of retrieval-augmented generation techniques using publicly available information about Doctor Who.
All source material originates from Wikipedia.