> RAG_MANUALS_v1.0
8 Canonical Engineering Manuals
Comprehensive technical manuals spanning layout parsing, vector spaces, hybrid reranking, and zero-downtime production operations.
Executive Architecture Overview
Architectural foundations, the RAG vs Fine-Tuning vs Long-Context decision frontier, and lifecycle stages.
Document Ingestion & Layout Parsing
Layout-aware visual document parsing, OCR error tolerance, tabular structure recovery, and chunking invariants.
Embeddings & Vector Space Geometry
Dense bi-encoders, sparse SPLADE, late-interaction ColBERT, Matryoshka Representation Learning (MRL), and quantization.
Hybrid Search & Cross-Encoder Reranking
BM25 lexical search, Reciprocal Rank Fusion (RRF), and deep cross-encoder token-level reranking.
Context Engineering & Prompt Memory
Mitigating lost-in-the-middle attention degradation, prompt packing, prefix KV caching in vLLM, and episodic memory.
Evaluation, Benchmarking & Observability
The RAG Triad of metrics, automated LLM judges, Prediction-Powered Inference (PPI), and OpenTelemetry tracing.
Guardrails, Safety & Regulatory Governance
Indirect prompt injection defense, multi-tenant isolation, logit masking, and tamper-evident audit logs.
Production Engineering & Operational Hardening
Bridging the notebook-to-production gap, Change Data Capture (CDC), and blue-green zero-downtime vector reindexing.