fix(export): make generated Strands runtimes deployable

src/assets/templates/strands-http-python/main.py

@@ -17,23 +17,21 @@
 {{/if}}
 {{/if}}
 import asyncio
+{{#if timeoutSeconds}}
+import threading
+{{/if}}
 {{#if hasShell}}
 import subprocess
 {{/if}}
 {{#if hasFileOperations}}
 import os
 {{/if}}
-{{#if hasExecutionLimits}}
-from strands.tools.executors import SequentialToolExecutor
-from strands.types.exceptions import EventLoopException
-from hooks.execution_limits import ExecutionLimitExceeded, ExecutionLimitsHook
-{{/if}}
 {{#if hasConfigBundle}}
 from strands.hooks import HookProvider, HookRegistry, BeforeInvocationEvent, BeforeToolCallEvent
 {{/if}}
 {{#if truncationStrategy}}
 {{#if (eq truncationStrategy "sliding_window")}}
-from strands.agent.conversation_manager.sliding_window_conversation_manager import SlidingWindowConversationManager
+from strands.agent.conversation_manager import SlidingWindowConversationManager
 {{/if}}
 {{#if (eq truncationStrategy "summarization")}}
 from strands.agent.conversation_manager.summarizing_conversation_manager import SummarizingConversationManager
@@ -413,18 +411,7 @@ def get_or_create_agent(session_id, user_id{{#if hasSkillsFetcher}}, skill_plugi
                 {{#if hasSkillsFetcher}}
                 plugins=skill_plugins or None,
                 {{/if}}
-                {{#if hasExecutionLimits}}
-                tool_executor=SequentialToolExecutor(),
-                callback_handler=None,
-                {{/if}}
                 hooks=[
-                    {{#if hasExecutionLimits}}
-                    ExecutionLimitsHook(
-                        {{#if maxIterations}}max_iterations={{maxIterations}},{{/if}}
-                        {{#if maxTokens}}max_tokens={{maxTokens}},{{/if}}
-                        {{#if timeoutSeconds}}timeout_seconds={{timeoutSeconds}},{{/if}}
-                    ),
-                    {{/if}}
                     {{#if hasConfigBundle}}
                     ConfigBundleHook(),
                     {{/if}}
@@ -457,18 +444,7 @@ def get_or_create_agent(session_id{{#if hasSkillsFetcher}}, skill_plugins=None{{
             {{#if hasSkillsFetcher}}
             plugins=skill_plugins or None,
             {{/if}}
-            {{#if hasExecutionLimits}}
-            tool_executor=SequentialToolExecutor(),
-            callback_handler=None,
-            {{/if}}
             hooks=[
-                {{#if hasExecutionLimits}}
-                ExecutionLimitsHook(
-                    {{#if maxIterations}}max_iterations={{maxIterations}},{{/if}}
-                    {{#if maxTokens}}max_tokens={{maxTokens}},{{/if}}
-                    {{#if timeoutSeconds}}timeout_seconds={{timeoutSeconds}},{{/if}}
-                ),
-                {{/if}}
                 {{#if hasConfigBundle}}
                 ConfigBundleHook(),
                 {{/if}}
@@ -639,24 +615,36 @@ async def invoke(payload, context):
     {{/if}}
 
     {{#if hasExecutionLimits}}
-    timeout_seconds = {{#if timeoutSeconds}}{{timeoutSeconds}}{{else}}None{{/if}}
+    limits = {
+        {{#if maxIterations}}"turns": {{maxIterations}},{{/if}}
+        {{#if maxTokens}}"output_tokens": {{maxTokens}},{{/if}}
+    } or None
+    cancel_signal = {{#if timeoutSeconds}}threading.Event(){{else}}None{{/if}}
     timeout_fired = False
     watchdog_task = None
-    if timeout_seconds is not None:
+    {{#if timeoutSeconds}}
+    if cancel_signal is not None:
         async def _timeout_watchdog():
             nonlocal timeout_fired
-            await asyncio.sleep(timeout_seconds)
+            await asyncio.sleep({{timeoutSeconds}})
             timeout_fired = True
-            agent.cancel()
+            cancel_signal.set()
         watchdog_task = asyncio.create_task(_timeout_watchdog())
+    {{/if}}
 
     try:
+        stop_reason = None
         {{#if inlineFunctionTools}}
         hit_inline_function = False
         {{/if}}
         async for event in agent.stream_async(
             prompt,
+            limits=limits,
+            cancel_signal=cancel_signal,
         ):
+            if isinstance(event, dict) and "result" in event:
+                stop_reason = getattr(event["result"], "stop_reason", None)
+                continue
             if not isinstance(event, dict) or "event" not in event:
                 continue
             cbs = event["event"].get("contentBlockStart")
@@ -674,11 +662,14 @@ async def _timeout_watchdog():
 
         if timeout_fired:
             yield {"event": {"messageStop": {"stopReason": "timeout_exceeded"}}}
-    except EventLoopException as e:
-        if isinstance(e.original_exception, ExecutionLimitExceeded):
-            yield {"event": {"messageStop": {"stopReason": str(e.original_exception)}}}
-            return
-        raise
+        {{#if maxIterations}}
+        elif stop_reason == "limit_turns":
+            yield {"event": {"messageStop": {"stopReason": "Max iterations exceeded: {{maxIterations}}"}}}
+        {{/if}}
+        {{#if maxTokens}}
+        elif stop_reason == "limit_output_tokens":
+            yield {"event": {"messageStop": {"stopReason": "Max output tokens exceeded: {{maxTokens}}"}}}
+        {{/if}}
     finally:
         if watchdog_task is not None:
             watchdog_task.cancel()

src/assets/templates/strands-http-python/mcp_client/client.py

@@ -69,29 +69,32 @@ def get_all_gateway_mcp_clients() -> list[MCPClient]:
 {{#if headerCredentials}}
 {{#each headerCredentials}}
 @requires_api_key(provider_name="{{credentialName}}")
-def _get_{{snakeCase ../name}}_{{snakeCase headerKey}}_key(api_key: str) -> str:
+def _get_{{pythonName}}_key(api_key: str) -> str:
     """Fetch {{headerKey}} credential for {{../name}} from AgentCore Identity."""
     return api_key
 
 {{/each}}
 {{/if}}
-def get_{{snakeCase name}}_mcp_client() -> MCPClient | None:
+def get_{{pythonName}}_mcp_client() -> MCPClient | None:
     """Returns an MCP Client for the {{name}} remote MCP server."""
     url = {{safeJson url}}
     {{#if headerCredentials}}
-    if os.getenv("LOCAL_DEV") == "1":
-        headers = { {{#each headerCredentials}}{{safeJson headerKey}}: os.environ.get("{{envVarName}}", ""){{#unless @last}}, {{/unless}}{{/each}} }
-    else:
-        headers = { {{#each headerCredentials}}{{safeJson headerKey}}: _get_{{snakeCase ../name}}_{{snakeCase headerKey}}_key(){{#unless @last}}, {{/unless}}{{/each}} }
-    return MCPClient(lambda: streamablehttp_client(url, headers=headers))
+    def transport():
+        if os.getenv("LOCAL_DEV") == "1":
+            headers = { {{#each headerCredentials}}{{safeJson headerKey}}: os.environ.get("{{envVarName}}", ""){{#unless @last}}, {{/unless}}{{/each}} }
+        else:
+            headers = { {{#each headerCredentials}}{{safeJson headerKey}}: _get_{{pythonName}}_key(){{#unless @last}}, {{/unless}}{{/each}} }
+        return streamablehttp_client(url, headers=headers)
+
+    return MCPClient(transport)
     {{else}}
     return MCPClient(lambda: streamablehttp_client(url))
     {{/if}}
 
 {{/each}}
 def get_all_remote_mcp_clients() -> list[MCPClient]:
     """Returns all configured remote MCP clients."""
-    clients = [{{#each remoteMcpTools}}get_{{snakeCase name}}_mcp_client(){{#unless @last}}, {{/unless}}{{/each}}]
+    clients = [{{#each remoteMcpTools}}get_{{pythonName}}_mcp_client(){{#unless @last}}, {{/unless}}{{/each}}]
     return [c for c in clients if c is not None]
 {{/if}}
 {{#unless (or hasGateway remoteMcpTools)}}

src/assets/templates/strands-http-python/memory/session.py

@@ -20,16 +20,16 @@ def get_memory_session_manager(
 {{#if memoryStrategies.length}}
     retrieval_config = {
 {{#if (includes memoryStrategies "SEMANTIC")}}
-        f"/users/{actor_id}/facts": RetrievalConfig(top_k=3, relevance_score=0.5),
+        f"/users/{actor_id}/facts": RetrievalConfig(top_k={{#if memoryRetrievalTopK}}{{memoryRetrievalTopK}}{{else}}3{{/if}}, relevance_score={{#if memoryRetrievalRelevanceScore}}{{memoryRetrievalRelevanceScore}}{{else}}0.5{{/if}}),
 {{/if}}
 {{#if (includes memoryStrategies "USER_PREFERENCE")}}
-        f"/users/{actor_id}/preferences": RetrievalConfig(top_k=3, relevance_score=0.5),
+        f"/users/{actor_id}/preferences": RetrievalConfig(top_k={{#if memoryRetrievalTopK}}{{memoryRetrievalTopK}}{{else}}3{{/if}}, relevance_score={{#if memoryRetrievalRelevanceScore}}{{memoryRetrievalRelevanceScore}}{{else}}0.5{{/if}}),
 {{/if}}
 {{#if (includes memoryStrategies "EPISODIC")}}
-        f"/episodes/{actor_id}/{session_id}": RetrievalConfig(top_k=5, relevance_score=0.5),
+        f"/episodes/{actor_id}/{session_id}": RetrievalConfig(top_k={{#if memoryRetrievalTopK}}{{memoryRetrievalTopK}}{{else}}5{{/if}}, relevance_score={{#if memoryRetrievalRelevanceScore}}{{memoryRetrievalRelevanceScore}}{{else}}0.5{{/if}}),
 {{/if}}
 {{#if (includes memoryStrategies "SUMMARIZATION")}}
-        f"/summaries/{actor_id}": RetrievalConfig(top_k=3, relevance_score=0.5),
+        f"/summaries/{actor_id}": RetrievalConfig(top_k={{#if memoryRetrievalTopK}}{{memoryRetrievalTopK}}{{else}}3{{/if}}, relevance_score={{#if memoryRetrievalRelevanceScore}}{{memoryRetrievalRelevanceScore}}{{else}}0.5{{/if}}),
 {{/if}}
     }
 {{/if}}

src/assets/templates/strands-http-python/model/load.py

@@ -1,6 +1,9 @@
 {{#if (eq modelProvider "Bedrock")}}
 {{#if bedrockMantle}}
 import os
+{{#if modelAdditionalParams}}
+import json
+{{/if}}
 
 from aws_bedrock_token_generator import provide_token
 {{#if (eq mantleApiFormat "chat_completions")}}
@@ -34,7 +37,7 @@ def load_model():
     {{/if}}
     client_args = {"api_key": token, "base_url": base_url}
 
-    params = {}
+    params = {{#if modelAdditionalParams}}json.loads({{pyJsonStr modelAdditionalParams}}){{else}}{}{{/if}}
     {{#if modelMaxTokens}}
     {{#if (eq mantleApiFormat "chat_completions")}}
     params["max_completion_tokens"] = {{modelMaxTokens}}
@@ -60,12 +63,22 @@ def load_model():
     {{/if}}
     {{/if}}
 {{else}}
+{{#if modelAdditionalParams}}
+import json
+{{/if}}
 from strands.models.bedrock import BedrockModel
 
 
 def load_model() -> BedrockModel:
     """Get Bedrock model client using IAM credentials."""
-    return BedrockModel(model_id="{{#if modelId}}{{modelId}}{{else}}global.anthropic.claude-sonnet-4-5-20250929-v1:0{{/if}}"{{#if modelMaxTokens}}, max_tokens={{modelMaxTokens}}{{/if}}{{#if modelTemperature}}, temperature={{modelTemperature}}{{/if}}{{#if modelTopP}}, top_p={{modelTopP}}{{/if}})
+    return BedrockModel(
+        model_id="{{#if modelId}}{{modelId}}{{else}}global.anthropic.claude-sonnet-4-5-20250929-v1:0{{/if}}",
+        {{#if modelMaxTokens}}max_tokens={{modelMaxTokens}},
+        {{/if}}{{#if modelTemperature}}temperature={{modelTemperature}},
+        {{/if}}{{#if modelTopP}}top_p={{modelTopP}},
+        {{/if}}{{#if modelAdditionalParams}}additional_request_fields=json.loads({{pyJsonStr modelAdditionalParams}}),
+        {{/if}}
+    )
 {{/if}}
 {{/if}}
 {{#if (eq modelProvider "Anthropic")}}
@@ -109,8 +122,15 @@ def load_model() -> AnthropicModel:
 {{/if}}
 {{#if (eq modelProvider "OpenAI")}}
 import os
+{{#if modelAdditionalParams}}
+import json
+{{/if}}
 
+{{#if (eq modelApiFormat "responses")}}
+from strands.models.openai_responses import OpenAIResponsesModel
+{{else}}
 from strands.models.openai import OpenAIModel
+{{/if}}
 from bedrock_agentcore.identity.auth import requires_api_key
 
 IDENTITY_PROVIDER_NAME = "{{identityProviders.[0].name}}"
@@ -138,15 +158,29 @@ def _get_api_key() -> str:
     return _agentcore_identity_api_key_provider()
 
 
-def load_model() -> OpenAIModel:
+def load_model():
     """Get authenticated OpenAI model client."""
-    return OpenAIModel(
+    params = {{#if modelAdditionalParams}}json.loads({{pyJsonStr modelAdditionalParams}}){{else}}{}{{/if}}
+    {{#if modelMaxTokens}}
+    params["{{#if (eq modelApiFormat "responses")}}max_output_tokens{{else}}max_completion_tokens{{/if}}"] = {{modelMaxTokens}}
+    {{/if}}
+    {{#if modelTemperature}}
+    params["temperature"] = {{modelTemperature}}
+    {{/if}}
+    {{#if modelTopP}}
+    params["top_p"] = {{modelTopP}}
+    {{/if}}
+    return {{#if (eq modelApiFormat "responses")}}OpenAIResponsesModel{{else}}OpenAIModel{{/if}}(
         client_args={"api_key": _get_api_key()},
         model_id="{{#if modelId}}{{modelId}}{{else}}gpt-4.1{{/if}}",
+        params=params,
     )
 {{/if}}
 {{#if (eq modelProvider "Gemini")}}
 import os
+{{#if modelAdditionalParams}}
+import json
+{{/if}}
 
 from strands.models.gemini import GeminiModel
 from bedrock_agentcore.identity.auth import requires_api_key
@@ -178,14 +212,28 @@ def _get_api_key() -> str:
 
 def load_model() -> GeminiModel:
     """Get authenticated Gemini model client."""
+    params = {{#if modelAdditionalParams}}json.loads({{pyJsonStr modelAdditionalParams}}){{else}}{}{{/if}}
+    {{#if modelMaxTokens}}
+    params["max_output_tokens"] = {{modelMaxTokens}}
+    {{/if}}
+    {{#if modelTemperature}}
+    params["temperature"] = {{modelTemperature}}
+    {{/if}}
+    {{#if modelTopP}}
+    params["top_p"] = {{modelTopP}}
+    {{/if}}
+    {{#if modelTopK}}
+    params["top_k"] = {{modelTopK}}
+    {{/if}}
     return GeminiModel(
         client_args={"api_key": _get_api_key()},
         model_id="{{#if modelId}}{{modelId}}{{else}}gemini-2.5-flash{{/if}}",
+        params=params,
     )
 {{/if}}
 {{#if (eq modelProvider "LiteLLM")}}
 import os
-{{#if litellmAdditionalParams}}
+{{#if modelAdditionalParams}}
 import json
 {{/if}}
 
@@ -230,7 +278,16 @@ def load_model() -> LiteLLMModel:
     {{#if litellmApiBase}}
     client_args["api_base"] = {{safeJson litellmApiBase}}
     {{/if}}
-    params = {{#if litellmAdditionalParams}}json.loads({{pyJsonStr litellmAdditionalParams}}){{else}}{}{{/if}}
+    params = {{#if modelAdditionalParams}}json.loads({{pyJsonStr modelAdditionalParams}}){{else}}{}{{/if}}
+    {{#if modelMaxTokens}}
+    params["max_tokens"] = {{modelMaxTokens}}
+    {{/if}}
+    {{#if modelTemperature}}
+    params["temperature"] = {{modelTemperature}}
+    {{/if}}
+    {{#if modelTopP}}
+    params["top_p"] = {{modelTopP}}
+    {{/if}}
     return LiteLLMModel(
         client_args=client_args,
         model_id="{{#if modelId}}{{modelId}}{{else}}bedrock/us.anthropic.claude-sonnet-4-5-20250514-v1:0{{/if}}",