i18n: complete EN/KO parity for Beyond RAG section

Add the 17 rag.* keys (rag.label, rag.title, rag.intro, rag.viz.*,
rag.chunk.*, rag.doc.*, rag.complementary.*, rag.card1-3.*, rag.pipeline.*)
to both EN and KO dictionaries. The HTML markup referenced these keys but
they were missing from the i18n tables, so the Beyond RAG section fell
back to the literal key strings on language toggle.

EN dict: 162 keys, KO dict: 162 keys, HTML data-i18n keys: 162 — verified
matching by diff.

Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>

Author: unamedkr

site/index.html

@@ -896,7 +896,24 @@ <h2 style="margin-bottom:1rem" data-i18n="cta.title">Try It Yourself</h2>
     "glossary.gguf.term": "GGUF",
     "glossary.gguf.def": "The standard file format for quantized LLM model weights, created by the llama.cpp project. quant.cpp loads GGUF models directly.",
     "cta.title": "Try It Yourself",
-    "cta.desc": "Three lines of Python. No GPU, no API key, no setup."
+    "cta.desc": "Three lines of Python. No GPU, no API key, no setup.",
+    "rag.label": "Movement",
+    "rag.title": "Beyond RAG",
+    "rag.intro": "Traditional RAG splits documents into 512-token chunks, embeds them in a vector database, and retrieves fragments. This was a reasonable engineering compromise when LLMs had 2K context windows. <strong>Now they have 128K. The compromise should have started disappearing.</strong>",
+    "rag.viz.title": "Chunk-Level RAG vs Document-Level RAG",
+    "rag.chunk.title": "Chunk-Level RAG",
+    "rag.chunk.result": "✗ Cross-page info lost",
+    "rag.doc.title": "Document-Level RAG",
+    "rag.doc.result": "✓ Full document understanding",
+    "rag.complementary.title": "Complementary, Not Competitive",
+    "rag.complementary.desc": "RAG decides <strong>which documents</strong> to look at. Long-context decides <strong>how deeply</strong> to understand them. Each does what it's best at.",
+    "rag.card1.t": "RAG's weakness → Long-Context solves",
+    "rag.card1.d": "Chunk boundaries lose cross-page relationships. Multi-hop reasoning fails. Long-context keeps the full document — no information loss.",
+    "rag.card2.t": "Long-Context's weakness → RAG solves",
+    "rag.card2.d": "Can't fit 100K documents in context. Prefill is slow. RAG narrows the search to 2-3 relevant documents that DO fit.",
+    "rag.card3.t": "Read Once, Query Forever",
+    "rag.card3.d": "Pre-process documents into .kv files (GPU, once). Load instantly on any laptop (0.5s). Query offline, unlimited, private.",
+    "rag.pipeline.title": "Pre-computed KV Library Pattern"
   },
   ko: {
     "nav.problem": "\uBB38\uC81C\uC810",
@@ -1043,7 +1060,24 @@ <h2 style="margin-bottom:1rem" data-i18n="cta.title">Try It Yourself</h2>
     "glossary.gguf.term": "GGUF",
     "glossary.gguf.def": "\uC591\uC790\uD654\uB41C LLM \uBAA8\uB378 \uAC00\uC911\uCE58\uC758 \uD45C\uC900 \uD30C\uC77C \uD615\uC2DD. llama.cpp \uD504\uB85C\uC81D\uD2B8\uC5D0\uC11C \uB9CC\uB4E4\uC5C8\uC2B5\uB2C8\uB2E4. quant.cpp\uB294 GGUF \uBAA8\uB378\uC744 \uC9C1\uC811 \uB85C\uB4DC\uD569\uB2C8\uB2E4.",
     "cta.title": "\uC9C1\uC811 \uD574\uBCF4\uAE30",
-    "cta.desc": "Python 3\uC904. GPU\uB3C4, API \uD0A4\uB3C4, \uC124\uCE58\uB3C4 \uD544\uC694 \uC5C6\uC2B5\uB2C8\uB2E4."
+    "cta.desc": "Python 3\uC904. GPU\uB3C4, API \uD0A4\uB3C4, \uC124\uCE58\uB3C4 \uD544\uC694 \uC5C6\uC2B5\uB2C8\uB2E4.",
+    "rag.label": "운동",
+    "rag.title": "Beyond RAG",
+    "rag.intro": "전통적인 RAG는 문서를 512토큰 청크로 나누고, 벡터 DB에 임베딩하고, 조각을 검색합니다. 이것은 LLM이 2K 컨텍스트만 가졌을 때 합리적인 엔지니어링 타협이었습니다. <strong>지금은 128K입니다. 그 타협은 사라지기 시작했어야 합니다.</strong>",
+    "rag.viz.title": "Chunk-Level RAG vs Document-Level RAG",
+    "rag.chunk.title": "Chunk-Level RAG",
+    "rag.chunk.result": "✗ 페이지 간 정보 손실",
+    "rag.doc.title": "Document-Level RAG",
+    "rag.doc.result": "✓ 전체 문서 이해",
+    "rag.complementary.title": "경쟁이 아닌 상호 보완",
+    "rag.complementary.desc": "RAG는 <strong>어떤 문서를</strong> 볼지 결정합니다. Long-context는 <strong>얼마나 깊이</strong> 이해할지 결정합니다. 각자 잘하는 것을 합니다.",
+    "rag.card1.t": "RAG의 약점 → Long-Context가 해결",
+    "rag.card1.d": "청크 경계에서 페이지 간 관계가 사라집니다. Multi-hop 추론은 실패합니다. Long-context는 전체 문서를 유지합니다 — 정보 손실 없음.",
+    "rag.card2.t": "Long-Context의 약점 → RAG가 해결",
+    "rag.card2.d": "100K 문서를 한 번에 컨텍스트에 넣을 수 없습니다. Prefill이 느립니다. RAG는 검색을 2-3개 관련 문서로 좁혀줍니다.",
+    "rag.card3.t": "한 번 읽고, 영원히 질문",
+    "rag.card3.d": "문서를 .kv 파일로 사전 처리 (GPU, 1회). 어떤 노트북에서든 즉시 로드 (0.5초). 오프라인, 무제한, 프라이빗 질문.",
+    "rag.pipeline.title": "사전 계산된 KV 라이브러리 패턴"
   }
 };