RAG System Demo - Interactive Retrieval Augmented Generation

1. Initialize RAG Pipeline

Load the embedding model, retriever, generator, and Wikipedia articles. This may take a moment.

Wikipedia Articles to Load:

Initializing...

Generator Model: - -

2. Document Collection

Dataset Loaded: 100 real Wikipedia articles across diverse topics

Loaded Wikipedia articles from various topics including science, history, arts, technology, and more. These articles provide a rich knowledge base for testing RAG capabilities.

View Sample Articles (Click to expand)

Click to upload text or JSON files

Or use the 100 Wikipedia articles

3. Document Chunking & Indexing

Chunking Strategy

Chunk Size (words)

Overlap (words)

Chunking Statistics

4. Embedding Visualization

Visualize document chunks in vector space (dimensionality reduced to 2D/3D)

5. Query & Retrieval

Enter your question

Number of documents to retrieve (Top K)

Mode

Processing query...

6. Results

Retrieved Documents

Generated Answer

Retrieval Time (ms)

Generation Time (ms)

Total Time (ms)

WITH RAG

Retrieved Documents:

Answer

Retrieval (ms)

Generation (ms)

Total (ms)

WITHOUT RAG

No context retrieval - answering from model knowledge only

Answer

Retrieval (ms)

Generation (ms)

Total (ms)

Note: Compare the quality and factual accuracy of both answers. RAG typically provides more accurate, grounded responses with specific information from the retrieved documents.

Sample Queries to Try

Try these queries against the Wikipedia articles to see RAG in action:

Retrieval Augmented Generation (RAG)

1. Initialize RAG Pipeline

2. Document Collection

3. Document Chunking & Indexing

Chunking Statistics

4. Embedding Visualization

5. Query & Retrieval

6. Results

Retrieved Documents

Generated Answer

WITH RAG

Answer

WITHOUT RAG

Answer

Sample Queries to Try