feat(agent-runner): support scoped token counting

src/langbot/pkg/agent/runner/context_builder.py

@@ -184,7 +184,7 @@ class AgentRunContextBuilder:
     def _is_llm_model_resource(model_resource: ModelResource) -> bool:
         operations = model_resource.get('operations')
         if isinstance(operations, list) and operations:
-            return bool({'invoke', 'stream'} & {str(operation) for operation in operations})
+            return bool({'invoke', 'stream', 'count_tokens'} & {str(operation) for operation in operations})
         return model_resource.get('model_type') != 'rerank'
 
     async def _build_model_context_window_tokens(self, resources: AgentResources) -> int | None:

src/langbot/pkg/agent/runner/resource_builder.py

@@ -101,9 +101,9 @@ class AgentResourceBuilder:
         seen_model_ids: set[str] = set()
 
         model_perms = set(manifest_perms.models)
-        include_llm = bool({'invoke', 'stream'} & model_perms)
+        include_llm = bool({'invoke', 'stream', 'count_tokens'} & model_perms)
         include_rerank = 'rerank' in model_perms
-        llm_operations = [operation for operation in ('invoke', 'stream') if operation in model_perms]
+        llm_operations = [operation for operation in ('invoke', 'stream', 'count_tokens') if operation in model_perms]
         if not include_llm and not include_rerank:
             return models
 

src/langbot/pkg/agent/runner/session_registry.py

@@ -13,7 +13,7 @@ from .context_builder import AgentResources
 MAX_STEERING_QUEUE_ITEMS = 100
 
 DEFAULT_RESOURCE_OPERATIONS: dict[str, set[str]] = {
-    'model': {'invoke', 'stream', 'rerank'},
+    'model': {'invoke', 'stream', 'rerank', 'count_tokens'},
     'tool': {'detail', 'call'},
     'knowledge_base': {'list', 'retrieve'},
     'skill': {'activate'},