How ragook6/rag_production works
This is a documentation-focused RAG assistant. Its user workflow is query → retrieval → grounded prompt → cited answer; its separate ingestion workflow chunks Markdown, embeds chunks, and persists them for retrieval. The graph also preserves character-versus-semantic chunking comparison and standalone vector search. Query-stage calls are supported by excerpts; ingestion-to-storage links reflect README-described behavior and are marked unverified because the ingestion orchestrators were not excerpted.
Main entry points: nodes.py Pipeline nodes · answer_generator.py Answer generation · embedder.py Text embeddings · hybrid_search.py Hybrid retrieval
Components 17
Document ingestion
- Markdown documents
docs - Character splitting
chunking/character_splitter.py - Semantic splitting
chunking/semantic_splitter.py - Chunk comparison
chunking/comparison.py - Text embeddings
vectorstore/embedder.py - ChromaDB collection
vectorstore/store.py
Retrieval
- Hybrid retrieval
retrieval/hybrid_search.py - Vector search
retrieval/semantic_search.py
Answer generation
- Grounded prompt
generation/prompt_builder.py - Answer generation
generation/answer_generator.py - OpenAI API
- Cited answer
generation/answer_generator.py
Pipeline runtime
- Runtime settings
config/settings.py - RAG state
pipeline/state.py - Pipeline nodes
pipeline/nodes.py - Query entry point
pipeline/graph.py
Other components
- Questioner
Connections 21
- Questioner → Query entry point: asks query
README.md - Query entry point → Pipeline nodes: runs graph
pipeline/graph.py - Pipeline nodes → RAG state: uses state
pipeline/nodes.py - Pipeline nodes → Hybrid retrieval: retrieves chunks
pipeline/nodes.py - Hybrid retrieval → ChromaDB collection: reads collection
retrieval/hybrid_search.py - Hybrid retrieval → Text embeddings: encodes query
retrieval/hybrid_search.py - Pipeline nodes → Grounded prompt: builds prompt
pipeline/nodes.py - Grounded prompt → Vector search: uses chunk type
generation/prompt_builder.py - Pipeline nodes → Answer generation: generates answer
pipeline/nodes.py - Answer generation → Grounded prompt: consumes prompt
generation/answer_generator.py - Answer generation → OpenAI API: calls API
generation/answer_generator.py - Answer generation → Cited answer: returns answer
generation/answer_generator.py - Markdown documents → Character splitting: splits characters
README.md - Markdown documents → Semantic splitting: splits sections
README.md - Character splitting → Text embeddings: embeds chunks
README.md - Semantic splitting → Text embeddings: embeds chunks
README.md - Text embeddings → ChromaDB collection: stores vectors
README.md - Chunk comparison → Character splitting: compares strategy
chunking/comparison.py - Chunk comparison → Semantic splitting: compares strategy
chunking/comparison.py - Vector search → ChromaDB collection: queries vectors
retrieval/semantic_search.py - Runtime settings → Pipeline nodes: configures runtime
pipeline/nodes.py