How xbl2602/rag-redo works
RAG NAVIGATION builds a local, content-based map of note libraries so a person or AI assistant can locate a file and section without reading the whole collection. The principal workflow is library selection, extraction, chunking and indexing, followed by semantic and lexical retrieval, fusion, reranking and result delivery. The README also documents independent PDF page navigation, OCR, library summaries, note relations, duplicate detection and import/export. Unsampled plugin internals are represented at their exact implementation paths; relationships use README evidence where capabilities are documented but source calls were not excerpted.
Main entry points: pipeline.py RAG pipeline · tools.py MCP tools · gui_main.py Desktop app · cli.py Command line
Components 22
User entry points
- Desktop app
gui_main.py - MCP tools
plugins/official-mcp-server/official_mcp_server/tools.py - Command line
core/cli.py
Library indexing
- RAG pipeline
core/pipeline.py - Library manager
plugins/official-library-manager/official_library_manager/plugin.py - Document extraction
plugins/official-extractor-pdf-text/official_extractor_pdf_text/extract.py - OCR extraction
plugins/official-ocr-mineru-cloud/official_ocr_mineru_cloud/extract.py - Chunking
plugins/official-chunker/official_chunker/chunk.py - Embedding
plugins/official-embedder-bge-m3/official_embedder_bge_m3/embed.py
Search and navigation
- BM25 indexing
plugins/official-lexical-bm25/official_lexical_bm25/bm25.py - Vector store
plugins/official-vector-store-chroma/official_vector_store_chroma/store.py - Rank fusion
plugins/official-fusion-rrf/official_fusion_rrf/rrf.py - Reranking
plugins/official-reranker/official_reranker/rerank.py - PDF page navigation
plugins/official-visual-wemm/official_visual_wemm/plugin.py
Optional capabilities
- Library summaries
plugins/official-library-summary/official_library_summary/plugin.py - Note relations
core/note_relations.py - Duplicate detection
plugins/official-dedup/official_dedup/dedup.py - Import / export
plugins/official-import-export/official_import_export/archive.py
Runtime and state
- Plugin runtime
core/runtime.py - Index workers
core/index_progress.py - Conversion cache
core/conversion_cache.py
Other components
- Person / AI
Connections 28
- Person / AI → Desktop app: opens
- Person / AI → MCP tools: asks
- Person / AI → Command line: runs
- Desktop app → RAG pipeline: uses
gui_main.py - MCP tools → RAG pipeline: calls
plugins/official-mcp-server/official_mcp_server/tools.py - Command line → RAG pipeline: uses
core/cli.py - RAG pipeline → Library manager: selects files
- RAG pipeline → Document extraction: extracts text
- RAG pipeline → OCR extraction: routes scans
- RAG pipeline → Chunking: chunks text
- RAG pipeline → Embedding: embeds chunks
- RAG pipeline → BM25 indexing: indexes terms
- RAG pipeline → Vector store: writes vectors
- RAG pipeline → Index workers: reports progress
core/pipeline.py - RAG pipeline → Conversion cache: uses cache
core/pipeline.py - RAG pipeline → PDF page navigation: builds page index
- RAG pipeline → BM25 indexing: retrieves terms
- RAG pipeline → Vector store: retrieves vectors
- RAG pipeline → Rank fusion: fuses ranks
- RAG pipeline → Reranking: reranks results
- MCP tools → Library summaries: offers summaries
- MCP tools → Note relations: offers links
plugins/official-mcp-server/official_mcp_server/tools.py - MCP tools → Duplicate detection: offers duplicate check
- MCP tools → Import / export: offers archive transfer
plugins/official-mcp-server/official_mcp_server/tools.py - Desktop app → PDF page navigation: shows page map
- Desktop app → Library summaries: shows summaries
- Desktop app → Note relations: shows note links
- Plugin runtime → RAG pipeline: hosts plugins
core/runtime.py