diff --git a/apps/api/src/api/v1/rag.py b/apps/api/src/api/v1/rag.py
index a385570a..ad806ff6 100644
--- a/apps/api/src/api/v1/rag.py
+++ b/apps/api/src/api/v1/rag.py
@@ -8,9 +8,10 @@ leWOOOgo 原則: Router 只做 HTTP 轉發，業務邏輯在 KnowledgeRAGService
 建立者: Claude Code (Phase 33 ADR-067)
 """
 
-from fastapi import APIRouter, BackgroundTasks, HTTPException
+from fastapi import APIRouter, BackgroundTasks
 from pydantic import BaseModel
 
+from src.core.config import get_settings
 from src.services.knowledge_rag_service import get_knowledge_rag_service
 
 router = APIRouter(prefix="/rag", tags=["RAG Knowledge Base"])
@@ -43,9 +44,10 @@ async def trigger_index(background_tasks: BackgroundTasks) -> RagIndexResponse:
     - .agents/skills/*.md
     """
     background_tasks.add_task(_run_index)
+    model = get_settings().OLLAMA_EMBEDDING_MODEL
     return RagIndexResponse(
         status="accepted",
-        message="索引已排程，背景執行中（nomic-embed-text @ Ollama 111）",
+        message=f"索引已排程，背景執行中（{model} @ Ollama GCP-A/GCP-B/111）",
     )
 
 
@@ -76,15 +78,16 @@ async def rag_debug() -> dict:
     try:
         async with httpx.AsyncClient(timeout=10.0) as c:
             from src.core.config import get_settings as _gs
+            settings = _gs()
             r = await c.post(
-                f"{_gs().OLLAMA_URL}/api/embeddings",
-                json={"model": "nomic-embed-text", "prompt": "test"},
+                f"{settings.OLLAMA_URL}/api/embeddings",
+                json={"model": settings.OLLAMA_EMBEDDING_MODEL, "prompt": "test"},
             )
             ollama_ok = r.status_code == 200 if r.status_code == 200 else f"http_{r.status_code}"
     except Exception as e:
         ollama_ok = f"error: {type(e).__name__}: {e}"
 
-    return {"cwd": os.getcwd(), "paths": paths_check, "ollama_111_embed": ollama_ok}
+    return {"cwd": os.getcwd(), "paths": paths_check, "ollama_embedding": ollama_ok}
 
 
 @router.get("/stats", summary="索引統計")
diff --git a/apps/api/src/core/config.py b/apps/api/src/core/config.py
index 351bddc6..e9e6d465 100644
--- a/apps/api/src/core/config.py
+++ b/apps/api/src/core/config.py
@@ -376,6 +376,10 @@ class Settings(BaseSettings):
         default="gemma3:4b",
         description="OllamaHealthMonitor 推理測試使用模型（P1.1）",
     )
+    OLLAMA_EMBEDDING_MODEL: str = Field(
+        default="bge-m3:latest",
+        description="Ollama embedding model. ADR-110 migrated embeddings from nomic-embed-text to bge-m3.",
+    )
     # 2026-04-12 ogt: 心跳必須確認載入的 Ollama 模型清單
     # 2026-05-04 ogt + Claude Sonnet 4.6: ADR-110 GCP 升級，更新必要模型清單（nomic→bge-m3 + 新增 qwen3:14b + hermes3）
     OLLAMA_REQUIRED_MODELS: list[str] = Field(
diff --git a/apps/api/src/services/knowledge_rag_service.py b/apps/api/src/services/knowledge_rag_service.py
index 93191722..57c046df 100644
--- a/apps/api/src/services/knowledge_rag_service.py
+++ b/apps/api/src/services/knowledge_rag_service.py
@@ -1,13 +1,13 @@
 """
 AWOOOI — Knowledge RAG Service (Phase 33, ADR-067)
 ==================================================
-本地 RAG 知識庫：nomic-embed-text 768維向量 + pgvector
+本地 RAG 知識庫：bge-m3 1024維向量 + pgvector
 
 索引策略:
   - 初期 < 100 筆: 線性搜尋
   - 超過 100 筆: 執行 CREATE INDEX ivfflat (手動觸發)
 
-向量模型: nomic-embed-text (Ollama 111, 768維) — 188:11434 被 NetworkPolicy v1.3 封閉
+向量模型: bge-m3 (GCP-A/GCP-B/111 Ollama lane, 1024維)
 生成模型: qwen2.5:7b-instruct (Ollama 111)
 
 leWOOOgo: Service 層只處理業務邏輯，DB 存取委派 rag_chunk_repository
@@ -21,11 +21,12 @@ import httpx
 import structlog
 
 import src.repositories.rag_chunk_repository as rag_repo
+from src.core.config import settings
 from src.services.ollama_endpoint_resolver import resolve_ollama_endpoint
 
 logger = structlog.get_logger(__name__)
 
-_EMBED_MODEL = "nomic-embed-text"
+_EMBED_MODEL = "bge-m3:latest"
 _GEN_MODEL = "qwen2.5:7b-instruct"
 _TOP_K = 5
 
@@ -131,7 +132,10 @@ class KnowledgeRAGService:
             http = await self._get_http()
             resp = await http.post(
                 f"{resolve_ollama_endpoint('embedding')}/api/embeddings",
-                json={"model": _EMBED_MODEL, "prompt": text},
+                json={
+                    "model": getattr(settings, "OLLAMA_EMBEDDING_MODEL", _EMBED_MODEL),
+                    "prompt": text,
+                },
             )
             if resp.status_code == 200:
                 return resp.json().get("embedding")
diff --git a/apps/api/src/services/playbook_rag.py b/apps/api/src/services/playbook_rag.py
index 62fe3b57..db818c73 100644
--- a/apps/api/src/services/playbook_rag.py
+++ b/apps/api/src/services/playbook_rag.py
@@ -4,7 +4,7 @@ Playbook RAG Service - Phase 3 向量化語意搜尋
 ADR-030: 智能自動修復系統
 
 使用 Embedding 進行 Playbook 語意搜尋:
-1. Ollama nomic-embed-text 生成向量
+1. Ollama bge-m3 生成向量
 2. Redis 儲存向量 (JSON 格式)
 3. 餘弦相似度搜尋
 
@@ -41,9 +41,9 @@ logger = structlog.get_logger(__name__)
 # Constants
 # =============================================================================
 
-# Embedding Model (Ollama 本地)
-EMBEDDING_MODEL = "nomic-embed-text"
-EMBEDDING_DIM = 768  # nomic-embed-text 向量維度
+# Embedding Model (Ollama)
+EMBEDDING_MODEL = "bge-m3:latest"
+EMBEDDING_DIM = 1024  # bge-m3 向量維度
 
 
 def _dedupe_urls(urls: list[str]) -> list[str]:
@@ -170,7 +170,7 @@ class PlaybookRAGService:
                 getattr(settings, "OLLAMA_FALLBACK_URL", ""),
             ]
         )
-        self.embedding_model = EMBEDDING_MODEL
+        self.embedding_model = str(getattr(settings, "OLLAMA_EMBEDDING_MODEL", EMBEDDING_MODEL) or EMBEDDING_MODEL)
 
     # =========================================================================
     # Embedding Operations
diff --git a/docs/LOGBOOK.md b/docs/LOGBOOK.md
index 78dc65e3..4a3c5b12 100644
--- a/docs/LOGBOOK.md
+++ b/docs/LOGBOOK.md
@@ -14,7 +14,7 @@
 - 新增 `INCIDENT_LLM_TIMEOUT_SECONDS`，production 設為 240s；Incident LLM 外層 guard 不再硬編 25s，且不得低於 `OPENCLAW_TIMEOUT`。
 - 新增 `AGENT_DEBATE_GLOBAL_TIMEOUT_SEC`，production 設為 260s；Phase 2 debate 不再被 90s 固定值卡死。
 - `ollama_endpoint_resolver` 改為非敏感工作（embedding/RAG/deep_rca/Hermes/code_review 等）GCP-A 優先、GCP-B 備援、111 兜底；只有 `local_required` / `privacy_sensitive` / `dr` 維持 local-first。
-- `PlaybookRAGService.embed_text()` 改為依序嘗試配置的 Ollama endpoints，單一 endpoint 失敗不再直接放棄 RAG。
+- `PlaybookRAGService.embed_text()` 改為依序嘗試配置的 Ollama endpoints，單一 endpoint 失敗不再直接放棄 RAG；Playbook/Knowledge RAG embedding model 改為 ADR-110 的 `bge-m3:latest`，避免 GCP-A/B 因舊 `nomic-embed-text` 回 404 後再掉到不可達的 111。
 
 **驗證**：
 - `py_compile` touched backend files OK；ruff `E9,F401,F821,F841` OK。
diff --git a/k8s/awoooi-prod/04-configmap.yaml b/k8s/awoooi-prod/04-configmap.yaml
index dd75197d..6ea0b5ec 100644
--- a/k8s/awoooi-prod/04-configmap.yaml
+++ b/k8s/awoooi-prod/04-configmap.yaml
@@ -64,6 +64,7 @@ data:
   ALERT_AI_ENFORCE_OLLAMA_FIRST: "true"
   ALERT_OLLAMA_MODEL: "qwen3:14b"
   OLLAMA_HEALTH_CHECK_MODEL: "gemma3:4b"
+  OLLAMA_EMBEDDING_MODEL: "bge-m3:latest"
   OPENCLAW_DEFAULT_MODEL: "qwen2.5:7b-instruct"
   OPENCLAW_TIMEOUT: "120"
   INCIDENT_LLM_TIMEOUT_SECONDS: "240"
diff --git a/k8s/awoooi-prod/06-deployment-api.yaml b/k8s/awoooi-prod/06-deployment-api.yaml
index 209f3695..91a73da1 100644
--- a/k8s/awoooi-prod/06-deployment-api.yaml
+++ b/k8s/awoooi-prod/06-deployment-api.yaml
@@ -81,6 +81,8 @@ spec:
               value: "qwen3:14b"  # 2026-05-05 Codex: 告警以解決問題為目標，可等待深度診斷
             - name: OLLAMA_HEALTH_CHECK_MODEL
               value: "gemma3:4b"  # 2026-05-05 Codex: 避免 health probe 載入 qwen2.5 7B 污染 GCP alert lane
+            - name: OLLAMA_EMBEDDING_MODEL
+              value: "bge-m3:latest"
             - name: OPENCLAW_DEFAULT_MODEL
               value: "qwen2.5:7b-instruct"
             - name: OPENCLAW_TIMEOUT