feat: integrate RAG rag_jrxml submodule and fix Anthropic API key

Add rag submodule for semantic JRXML chunk retrieval, refactor
retrieve node to use RAGSearcher, and fix missing api_key in
Anthropic SDK client initialization.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

.env.example

@@ -22,6 +22,26 @@ VALIDATION_SERVICE_URL=http://localhost:8001/validate
 # Chroma 持久化目录
 CHROMA_PERSIST_DIR=./db/chroma
 
+# ---- RAG / 向量知识库 (rag_jrxml 子模块) ----
+# 嵌入模型
+RAG_EMBED_MODEL=sentence-transformers/paraphrase-multilingual-MiniLM-L12-v2
+# JRXML 模板源目录 (rag 子模块内已含 107 个模板)
+RAG_JRXML_SOURCE=./rag/jrxml_source
+# 分块输出目录
+RAG_CHUNKER_OUTPUT=./rag/jrxml_chunker_output
+# 向量输出目录
+RAG_EMBEDDINGS_DIR=./rag/embeddings
+# ChromaDB 知识库路径
+RAG_CHROMA_PATH=./db/chroma
+# ChromaDB 集合名称
+RAG_COLLECTION_NAME=jrxml_chunks
+# GPU 加速
+RAG_USE_GPU=true
+# FP16 半精度
+RAG_USE_FP16=true
+# 向量化批处理大小
+RAG_BATCH_SIZE=64
+
 # 最大自动修正尝试次数
 MAX_RETRY=3