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/ci-tests/build.yaml

@@ -48,6 +48,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
     batches:
     - provider_type: inline::reference

llama_stack/distributions/ci-tests/run.yaml

@@ -216,6 +216,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
   batches:
@@ -261,6 +263,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/dell/build.yaml

@@ -26,6 +26,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
 image_type: venv
 additional_pip_packages:
 - aiosqlite

llama_stack/distributions/dell/dell.py

@@ -45,6 +45,7 @@ def get_distribution_template() -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
+            BuildProvider(provider_type="inline::rag-runtime"),
         ],
     }
     name = "dell"
@@ -97,6 +98,10 @@ def get_distribution_template() -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="brave-search",
         ),
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
     ]
 
     return DistributionTemplate(

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

@@ -87,6 +87,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
 storage:
   backends:
     kv_default:
@@ -131,6 +133,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: brave-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/dell/run.yaml

@@ -83,6 +83,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
 storage:
   backends:
     kv_default:
@@ -122,6 +124,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: brave-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/meta-reference-gpu/build.yaml

@@ -24,6 +24,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
 image_type: venv
 additional_pip_packages:

llama_stack/distributions/meta-reference-gpu/meta_reference.py

@@ -47,6 +47,7 @@ def get_distribution_template() -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
+            BuildProvider(provider_type="inline::rag-runtime"),
             BuildProvider(provider_type="remote::model-context-protocol"),
         ],
     }
@@ -91,6 +92,10 @@ def get_distribution_template() -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="tavily-search",
         ),
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
     ]
 
     return DistributionTemplate(

llama_stack/distributions/meta-reference-gpu/run-with-safety.yaml

@@ -98,6 +98,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
 storage:
@@ -144,6 +146,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/meta-reference-gpu/run.yaml

@@ -88,6 +88,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
 storage:
@@ -129,6 +131,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

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:

llama_stack/distributions/open-benchmark/build.yaml

@@ -28,6 +28,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
 image_type: venv
 additional_pip_packages:

llama_stack/distributions/open-benchmark/open_benchmark.py

@@ -118,6 +118,7 @@ def get_distribution_template() -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
+            BuildProvider(provider_type="inline::rag-runtime"),
             BuildProvider(provider_type="remote::model-context-protocol"),
         ],
     }
@@ -153,6 +154,10 @@ def get_distribution_template() -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="tavily-search",
         ),
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
     ]
 
     models, _ = get_model_registry(available_models)

llama_stack/distributions/open-benchmark/run.yaml

@@ -118,6 +118,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
 storage:
@@ -242,6 +244,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/postgres-demo/build.yaml

@@ -14,6 +14,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
 image_type: venv
 additional_pip_packages:

llama_stack/distributions/postgres-demo/postgres_demo.py

@@ -45,6 +45,7 @@ def get_distribution_template() -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
+            BuildProvider(provider_type="inline::rag-runtime"),
             BuildProvider(provider_type="remote::model-context-protocol"),
         ],
     }
@@ -65,6 +66,10 @@ def get_distribution_template() -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="tavily-search",
         ),
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
     ]
 
     default_models = [

llama_stack/distributions/postgres-demo/run.yaml

@@ -54,6 +54,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
 storage:
@@ -105,6 +107,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/starter-gpu/build.yaml

@@ -49,6 +49,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
     batches:
     - provider_type: inline::reference

llama_stack/distributions/starter-gpu/run.yaml

@@ -219,6 +219,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
   batches:
@@ -264,6 +266,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/starter/build.yaml

@@ -49,6 +49,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
     batches:
     - provider_type: inline::reference

llama_stack/distributions/starter/run.yaml

@@ -216,6 +216,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
   batches:
@@ -261,6 +263,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/starter/starter.py

@@ -140,6 +140,7 @@ def get_distribution_template(name: str = "starter") -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
+            BuildProvider(provider_type="inline::rag-runtime"),
             BuildProvider(provider_type="remote::model-context-protocol"),
         ],
         "batches": [
@@ -161,6 +162,10 @@ def get_distribution_template(name: str = "starter") -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="tavily-search",
         ),
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
     ]
     default_shields = [
         # if the

llama_stack/distributions/watsonx/build.yaml

@@ -23,6 +23,7 @@ distribution_spec:
     tool_runtime:
     - provider_type: remote::brave-search
     - provider_type: remote::tavily-search
+    - provider_type: inline::rag-runtime
     - provider_type: remote::model-context-protocol
     files:
     - provider_type: inline::localfs

llama_stack/distributions/watsonx/run.yaml

@@ -83,6 +83,8 @@ providers:
     config:
       api_key: ${env.TAVILY_SEARCH_API_KEY:=}
       max_results: 3
+  - provider_id: rag-runtime
+    provider_type: inline::rag-runtime
   - provider_id: model-context-protocol
     provider_type: remote::model-context-protocol
   files:
@@ -123,6 +125,8 @@ registered_resources:
   tool_groups:
   - toolgroup_id: builtin::websearch
     provider_id: tavily-search
+  - toolgroup_id: builtin::rag
+    provider_id: rag-runtime
 server:
   port: 8321
 telemetry:

llama_stack/distributions/watsonx/watsonx.py

@@ -33,6 +33,7 @@ def get_distribution_template(name: str = "watsonx") -> DistributionTemplate:
         "tool_runtime": [
             BuildProvider(provider_type="remote::brave-search"),
             BuildProvider(provider_type="remote::tavily-search"),
+            BuildProvider(provider_type="inline::rag-runtime"),
             BuildProvider(provider_type="remote::model-context-protocol"),
         ],
         "files": [BuildProvider(provider_type="inline::localfs")],
@@ -49,6 +50,10 @@ def get_distribution_template(name: str = "watsonx") -> DistributionTemplate:
             toolgroup_id="builtin::websearch",
             provider_id="tavily-search",
         ),
+        ToolGroupInput(
+            toolgroup_id="builtin::rag",
+            provider_id="rag-runtime",
+        ),
     ]
 
     files_provider = Provider(