How anwar-genai/rag-architecture works
RAG Architecture Canvas is a single-file interactive explainer, not a RAG processing service. Its user workflow is to explore the zoomable canvas, click a pipeline block for its details, and inspect the latency waterfall; users can also switch themes. The diagram documents ingestion, retrieval, generation, and evaluation feedback as a conceptual production pipeline. The README is the only available evidence, so the arrows describe documented interactions and conceptual flow, not inspected implementation calls.
Main entry points: index.html Eval feedback
Components 14
Interactive canvas
- Zoomable canvas
index.html - Pipeline blocks
index.html - Block details
index.html - Latency waterfall
index.html - Theme toggle
index.html
RAG pipeline
- Ingestion
index.html - Chunking
index.html - Embedding config
index.html - Hybrid retrieval
index.html - Retrieval config
index.html - Reranking
index.html - Generation
index.html
Evaluation loop
- Eval feedback
index.html
Other components
- User
Connections 16
- User → Zoomable canvas: explores
README.md - User → Pipeline blocks: clicks blocks
README.md - Pipeline blocks → Block details: shows details
README.md - User → Latency waterfall: inspects timings
README.md - User → Theme toggle: switches theme
README.md - Zoomable canvas → Pipeline blocks: displays
README.md - Ingestion → Chunking: feeds
README.md - Chunking → Embedding config: uses config
README.md - Embedding config → Hybrid retrieval: supports
README.md - Hybrid retrieval → Reranking: feeds
README.md - Reranking → Generation: feeds
README.md - Generation → Eval feedback: evaluated by
README.md - Eval feedback → Chunking: retunes
README.md - Eval feedback → Embedding config: retunes
README.md - Eval feedback → Retrieval config: retunes
README.md - Retrieval config → Hybrid retrieval: configures
README.md