Source Tokens
0
Vocabulary Size
0
System Latency
0ms
Vector Count
0

Visualization of RAG Process

Interactive demo using real TF-IDF vectorization and cosine similarity for retrieval.

1 Knowledge Base Ingestion

2 Data Chunking

3 TF-IDF Vector Embeddings

Each chunk is represented as a normalized sparse vector over the vocabulary. Top terms by TF-IDF weight are shown below each chunk.

4 User Query

5 Cosine Similarity Retrieval

Top K Results: 2

6 Context Injection

7 AI Response