revert: "chore(cleanup)!: remove tool_runtime.rag_tool" (#3877)

Reverts llamastack/llama-stack#3871

This PR broke RAG (even from Responses -- there _is_ a dependency)

llama_stack/distributions/nvidia/build.yaml

@@ -19,7 +19,8 @@ distribution_spec:
     - provider_type: remote::nvidia
     scoring:
     - provider_type: inline::basic
-    tool_runtime: []
+    tool_runtime:
+    - provider_type: inline::rag-runtime
     files:
     - provider_type: inline::localfs
 image_type: venv

llama_stack/distributions/nvidia/nvidia.py

@@ -28,7 +28,7 @@ def get_distribution_template(name: str = "nvidia") -> DistributionTemplate:
             BuildProvider(provider_type="remote::nvidia"),
         ],
         "scoring": [BuildProvider(provider_type="inline::basic")],
-        "tool_runtime": [],
+        "tool_runtime": [BuildProvider(provider_type="inline::rag-runtime")],
         "files": [BuildProvider(provider_type="inline::localfs")],
     }
 
@@ -66,7 +66,12 @@ def get_distribution_template(name: str = "nvidia") -> DistributionTemplate:
         provider_id="nvidia",
     )
 
-    default_tool_groups: list[ToolGroupInput] = []
+    default_tool_groups = [
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
+    ]
 
     return DistributionTemplate(
         name=name,

llama_stack/distributions/nvidia/run-with-safety.yaml

@@ -80,7 +80,9 @@ providers:
   scoring:
   - provider_id: basic
     provider_type: inline::basic
-  tool_runtime: []
+  tool_runtime:
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   files:
   - provider_id: meta-reference-files
     provider_type: inline::localfs
@@ -126,7 +128,9 @@ registered_resources:
   datasets: []
   scoring_fns: []
   benchmarks: []
-  tool_groups: []
+  tool_groups:
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/nvidia/run.yaml

@@ -69,7 +69,9 @@ providers:
   scoring:
   - provider_id: basic
     provider_type: inline::basic
-  tool_runtime: []
+  tool_runtime:
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   files:
   - provider_id: meta-reference-files
     provider_type: inline::localfs
@@ -105,7 +107,9 @@ registered_resources:
   datasets: []
   scoring_fns: []
   benchmarks: []
-  tool_groups: []
+  tool_groups:
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry: