How singhsidhukuldeep/supportdesk-rag-workshop works
SupportDesk-RAG is a hands-on workshop, not one integrated application: learners explore ticket retrieval in separate modules, from embeddings and document preparation through indexing and grounded RAG, then evaluation and agentic approaches. The README documents the capabilities; sampled demos show OpenAI embedding/chat use, Chroma-backed preparation, LlamaIndex indexing, and retrieval evaluation. Source excerpts are partial, so the graph does not assume that module outputs are wired together as a single runtime.
Main entry points: 4_rag_pipeline/demo.py Grounded RAG demos · 2_chunking/demo.py Chunking demos · 5_evaluation/demo.py RAG evaluation · 6_agentic_rag/demo.py Agentic RAG demos
Components 16
Retrieval Workshop
- Embedding demos
modules/1_embeddings/demo.py - Chunking demos
modules/2_chunking/demo.py - Index strategy demos
modules/3_indexing/demo.py - Grounded RAG demos
modules/4_rag_pipeline/demo.py
Evaluation and Agents
- RAG evaluation
modules/5_evaluation/demo.py - Evaluation queries
modules/5_evaluation/evaluation_queries.json - Agentic RAG demos
modules/6_agentic_rag/demo.py - Ticket tools
modules/6_agentic_rag/tools.py
External Systems
- OpenAI embeddings
- OpenAI chat models
- Chroma vector store
- LlamaIndex
- FAISS
- LangChain
Other components
- Workshop learner
- Synthetic tickets
data/synthetic_tickets.json
Connections 26
- Workshop learner → Embedding demos: runs demo
- Workshop learner → Chunking demos: runs demo
- Workshop learner → Index strategy demos: runs demo
- Workshop learner → Grounded RAG demos: runs demo
- Workshop learner → RAG evaluation: runs demo
- Workshop learner → Agentic RAG demos: runs demo
- Embedding demos → Synthetic tickets: embeds ticket text
- Embedding demos → OpenAI embeddings: requests vectors
- Chunking demos → Synthetic tickets: prepares documents
- Chunking demos → OpenAI embeddings: requests vectors
- Chunking demos → Chroma vector store: builds store
- Index strategy demos → LlamaIndex: builds indexes
- Index strategy demos → OpenAI chat models: requests summaries
- Grounded RAG demos → Synthetic tickets: loads tickets
- Grounded RAG demos → OpenAI embeddings: requests vectors
- Grounded RAG demos → Chroma vector store: creates vector store
- Grounded RAG demos → LangChain: builds RAG chain
- Grounded RAG demos → OpenAI chat models: generates answers
- RAG evaluation → Evaluation queries: reads queries
- RAG evaluation → FAISS: searches vectors
- RAG evaluation → OpenAI chat models: judges answers
- Agentic RAG demos → LangChain: binds tools
- Agentic RAG demos → OpenAI chat models: selects tools
- Agentic RAG demos → Ticket tools: dispatches calls
- Ticket tools → Synthetic tickets: reads tickets
- Ticket tools → OpenAI embeddings: requests vectors