diff --git a/.env.example b/.env.example
index 5a5f6429..849b94ad 100644
--- a/.env.example
+++ b/.env.example
@@ -66,6 +66,10 @@ EMBEDDING_MODEL=
 # OPTIONAL url for openrag link to langflow in the UI
 LANGFLOW_PUBLIC_URL=
 
+# OPTIONAL: Override the full docling-serve URL (e.g., for remote instances)
+# If not set, auto-detects host and uses port 5001
+# DOCLING_SERVE_URL=http://my-docling-server:5001
+
 # OPTIONAL: Override host for docling service (for special networking setups)
 # HOST_DOCKER_INTERNAL=host.containers.internal
 
diff --git a/.github/workflows/build-multiarch.yml b/.github/workflows/build-multiarch.yml
index 077b66ec..1f06553c 100644
--- a/.github/workflows/build-multiarch.yml
+++ b/.github/workflows/build-multiarch.yml
@@ -69,8 +69,7 @@ jobs:
             tag: langflowai/openrag-backend
             platform: linux/arm64
             arch: arm64
-            #runs-on: [self-hosted, linux, ARM64, langflow-ai-arm64-2]
-            runs-on: RagRunner
+            runs-on: [self-hosted, Linux, ARM64, langflow-ai-arm64-40gb-ephemeral]
 
           # frontend
           - image: frontend
@@ -84,8 +83,7 @@ jobs:
             tag: langflowai/openrag-frontend
             platform: linux/arm64
             arch: arm64
-            #runs-on: [self-hosted, linux, ARM64, langflow-ai-arm64-2]
-            runs-on: RagRunner
+            runs-on: [self-hosted, Linux, ARM64, langflow-ai-arm64-40gb-ephemeral]
 
           # langflow
           - image: langflow
@@ -99,8 +97,7 @@ jobs:
             tag: langflowai/openrag-langflow
             platform: linux/arm64
             arch: arm64
-            #runs-on: self-hosted
-            runs-on: RagRunner
+            runs-on: [self-hosted, Linux, ARM64, langflow-ai-arm64-40gb-ephemeral]
 
           # opensearch
           - image: opensearch
@@ -114,9 +111,7 @@ jobs:
             tag: langflowai/openrag-opensearch
             platform: linux/arm64
             arch: arm64
-            #runs-on: [self-hosted, linux, ARM64, langflow-ai-arm64-2]
-            #runs-on: self-hosted
-            runs-on: RagRunner
+            runs-on: [self-hosted, Linux, ARM64, langflow-ai-arm64-40gb-ephemeral]
 
     runs-on: ${{ matrix.runs-on }}
 
diff --git a/docker-compose.gpu.yml b/docker-compose.gpu.yml
index 4496c4ac..b849cd3b 100644
--- a/docker-compose.gpu.yml
+++ b/docker-compose.gpu.yml
@@ -3,5 +3,10 @@ services:
     environment:
       - NVIDIA_DRIVER_CAPABILITIES=compute,utility
       - NVIDIA_VISIBLE_DEVICES=all
-    gpus: all
-
+    deploy:
+      resources:
+        reservations:
+          devices:
+            - driver: nvidia
+              count: all
+              capabilities: [gpu]
diff --git a/docker-compose.yml b/docker-compose.yml
index c99205a6..30574506 100644
--- a/docker-compose.yml
+++ b/docker-compose.yml
@@ -29,7 +29,7 @@ services:
       - "9200:9200"
       - "9600:9600"
     volumes:
-      - ${OPENSEARCH_DATA_PATH:-./opensearch-data}:/usr/share/opensearch/data:Z
+      - ${OPENSEARCH_DATA_PATH:-./opensearch-data}:/usr/share/opensearch/data:U,z
 
   dashboards:
     image: opensearchproject/opensearch-dashboards:3.0.0
diff --git a/docs/docs/reference/api-sdk-overview.mdx b/docs/docs/reference/api-sdk-overview.mdx
new file mode 100644
index 00000000..d3a7418f
--- /dev/null
+++ b/docs/docs/reference/api-sdk-overview.mdx
@@ -0,0 +1,12 @@
+---
+title: OpenRAG APIs and SDKs
+slug: /reference/api-sdk-overview
+---
+
+You can use OpenRAG's APIs and SDKs to integrate and extend OpenRAG's capabilities:
+
+* [Python SDK](https://github.com/langflow-ai/openrag/tree/main/sdks/python)
+* [TypeScript/JavaScript SDK](https://github.com/langflow-ai/openrag/tree/main/sdks/typescript)
+
+<!-- TBD: MCP: See https://github.com/langflow-ai/openrag/pull/729 -->
+<!-- TBD: API Reference: See https://github.com/langflow-ai/openrag/issues/734 -->
\ No newline at end of file
diff --git a/docs/docusaurus.config.js b/docs/docusaurus.config.js
index 8525447c..5bc29f58 100644
--- a/docs/docusaurus.config.js
+++ b/docs/docusaurus.config.js
@@ -236,6 +236,15 @@ const config = {
           },
         ],
       },
+      algolia: {
+        appId: "SMEA51Q5OL",
+        // public key, safe to commit
+        apiKey: "b2ec302e9880e8979ad6a68f0c36271e",
+        indexName: "openrag-algolia",
+        contextualSearch: true,
+        searchParameters: {},
+        searchPagePath: "search",
+      },
       prism: {
         theme: prismThemes.github,
         darkTheme: prismThemes.dracula,
diff --git a/docs/sidebars.js b/docs/sidebars.js
index bc11ffa6..0cc0be91 100644
--- a/docs/sidebars.js
+++ b/docs/sidebars.js
@@ -75,6 +75,11 @@ const sidebars = {
       label: "Chat",
     },
     "reference/configuration",
+    {
+      type: "doc",
+      id: "reference/api-sdk-overview",
+      label: "APIs and SDKs",
+    },
     "support/contribute",
     "support/troubleshoot",
   ],
diff --git a/src/api/docling.py b/src/api/docling.py
index 22b709ef..2f9adcbd 100644
--- a/src/api/docling.py
+++ b/src/api/docling.py
@@ -1,5 +1,6 @@
 """Docling service proxy endpoints."""
 
+import os
 import socket
 import struct
 from pathlib import Path
@@ -73,9 +74,15 @@ def determine_docling_host() -> str:
     return "localhost"
 
 
-# Detect the host IP once at startup
-HOST_IP = determine_docling_host()
-DOCLING_SERVICE_URL = f"http://{HOST_IP}:5001"
+# Use explicit URL if provided, otherwise auto-detect host
+_docling_url_override = os.getenv("DOCLING_SERVE_URL")
+if _docling_url_override:
+    DOCLING_SERVICE_URL = _docling_url_override.rstrip("/")
+    HOST_IP = _docling_url_override  # For display in health responses
+    logger.info("Using DOCLING_SERVE_URL override: %s", DOCLING_SERVICE_URL)
+else:
+    HOST_IP = determine_docling_host()
+    DOCLING_SERVICE_URL = f"http://{HOST_IP}:5001"
 
 
 async def health(request: Request) -> JSONResponse: