fix(endpoint): improve runtime.kiro.dev migration

Author: jwadow

CONTRIBUTORS.md

@@ -6,7 +6,7 @@ Thank you to all the contributors who have helped improve this project!
 
 These contributors have made significant, sustained contributions to the project:
 
-- [@bhaskoro-muthohar](https://github.com/bhaskoro-muthohar) — MCP tool results analysis (#46, #50), message structure validation (#60), payload size guard (#73), tool_reference support (#89, #90), and ongoing project support
+- [@bhaskoro-muthohar](https://github.com/bhaskoro-muthohar) — MCP tool results analysis (#46, #50), message structure validation (#60), payload size guard (#73), tool_reference support (#89, #90), runtime.kiro.dev endpoint migration (#155), and ongoing project support
 
 ## Contributors
 

kiro/account_manager.py

@@ -64,6 +64,35 @@
 from kiro.http_client import KiroHttpClient
 
 
+def _is_runtime_endpoint(auth_manager: KiroAuthManager) -> bool:
+    """
+    Check if auth manager uses runtime endpoint that doesn't provide /ListAvailableModels.
+    
+    Runtime endpoint pattern: https://runtime.{region}.kiro.dev
+    Old endpoint pattern: https://q.{region}.amazonaws.com
+    
+    Runtime endpoint does not provide /ListAvailableModels API (AWS limitation).
+    
+    Args:
+        auth_manager: KiroAuthManager instance
+    
+    Returns:
+        True if using runtime endpoint, False otherwise
+    
+    Examples:
+        >>> auth_manager.api_host = "https://runtime.us-east-1.kiro.dev"
+        >>> _is_runtime_endpoint(auth_manager)
+        True
+        >>> auth_manager.api_host = "https://runtime.eu-central-1.kiro.dev"
+        >>> _is_runtime_endpoint(auth_manager)
+        True
+        >>> auth_manager.api_host = "https://q.us-east-1.amazonaws.com"
+        >>> _is_runtime_endpoint(auth_manager)
+        False
+    """
+    return "://runtime." in auth_manager.api_host
+
+
 def _format_duration(seconds: float) -> str:
     """
     Format duration in human-readable format.
@@ -468,40 +497,48 @@ async def _initialize_account(self, account_id: str) -> bool:
             # Get token to verify credentials
             token = await auth_manager.get_access_token()
 
-            # Fetch models list with retry + fallback
-            params = {"origin": "AI_EDITOR"}
-            if auth_manager.auth_type == AuthType.KIRO_DESKTOP and auth_manager.profile_arn:
-                params["profileArn"] = auth_manager.profile_arn
-            
-            list_models_url = f"{auth_manager.q_host}/ListAvailableModels"
-            
-            # Use KiroHttpClient for retry logic (3 attempts with exponential backoff)
-            http_client = KiroHttpClient(auth_manager, shared_client=None)
-            
-            try:
-                response = await http_client.request_with_retry(
-                    method="GET",
-                    url=list_models_url,
-                    json_data=None,
-                    params=params,
-                    stream=False
-                )
-                
-                if response.status_code == 200:
-                    data = response.json()
-                    models_list = data.get("models", [])
-                else:
-                    # Shouldn't happen (retry handles non-200), but keep for safety
-                    raise Exception(f"HTTP {response.status_code}")
-            
-            except Exception as e:
-                # All retries exhausted - use fallback
-                logger.error(f"Failed to fetch models for {account_id} after retries: {e}")
-                logger.warning("Using pre-configured fallback models. Models will be refreshed on next TTL cycle when network recovers.")
+            # Determine if we should fetch models or use static list
+            if _is_runtime_endpoint(auth_manager):
+                # New runtime endpoint does not provide /ListAvailableModels (AWS limitation)
+                # Use static list without attempting request
+                logger.debug(f"Account {account_id}: Using static model list for runtime.kiro.dev endpoint")
                 models_list = FALLBACK_MODELS
-            
-            finally:
-                await http_client.close()
+            else:
+                # Old endpoint - attempt to fetch dynamic model list
+                # Fetch models list with retry + fallback
+                params = {"origin": "AI_EDITOR"}
+                if auth_manager.auth_type == AuthType.KIRO_DESKTOP and auth_manager.profile_arn:
+                    params["profileArn"] = auth_manager.profile_arn
+                
+                list_models_url = f"{auth_manager.q_host}/ListAvailableModels"
+                
+                # Use KiroHttpClient for retry logic (3 attempts with exponential backoff)
+                http_client = KiroHttpClient(auth_manager, shared_client=None)
+                
+                try:
+                    response = await http_client.request_with_retry(
+                        method="GET",
+                        url=list_models_url,
+                        json_data=None,
+                        params=params,
+                        stream=False
+                    )
+                    
+                    if response.status_code == 200:
+                        data = response.json()
+                        models_list = data.get("models", [])
+                    else:
+                        # Shouldn't happen (retry handles non-200), but keep for safety
+                        raise Exception(f"HTTP {response.status_code}")
+                
+                except Exception as e:
+                    # All retries exhausted - use fallback
+                    logger.error(f"Failed to fetch models for {account_id} after retries: {e}")
+                    logger.warning("Using pre-configured fallback models. Models will be refreshed on next TTL cycle when network recovers.")
+                    models_list = FALLBACK_MODELS
+                
+                finally:
+                    await http_client.close()
 
             # Create model cache and update
             model_cache = ModelInfoCache()
@@ -552,6 +589,17 @@ async def _refresh_account_models(self, account_id: str) -> None:
         if not account or not account.auth_manager:
             return
 
+        # Check if using runtime endpoint (no dynamic model list available)
+        if _is_runtime_endpoint(account.auth_manager):
+            # Runtime endpoint does not provide /ListAvailableModels
+            # Use static list and update cache timestamp
+            logger.debug(f"Account {account_id}: Skipping model refresh for runtime.kiro.dev endpoint (using static list)")
+            await account.model_cache.update(FALLBACK_MODELS)
+            account.models_cached_at = time.time()
+            self._dirty = True
+            return
+        
+        # Old endpoint - attempt to fetch dynamic model list
         # Use KiroHttpClient for retry logic
         http_client = KiroHttpClient(account.auth_manager, shared_client=None)
 

kiro/config.py

@@ -276,9 +276,12 @@ def _get_raw_env_value(var_name: str, env_file: str = ".env") -> Optional[str]:
 FALLBACK_MODELS: List[Dict[str, str]] = [
     {"modelId": "auto"},
     {"modelId": "claude-sonnet-4"},
-    {"modelId": "claude-haiku-4.5"},
     {"modelId": "claude-sonnet-4.5"},
+    {"modelId": "claude-sonnet-4.6"},
+    {"modelId": "claude-haiku-4.5"},
     {"modelId": "claude-opus-4.5"},
+    {"modelId": "claude-opus-4.6"},
+    {"modelId": "claude-opus-4.7"},
 ]
 
 # ==================================================================================================

kiro/model_resolver.py

@@ -42,16 +42,10 @@
 
 
 # Valid model IDs accepted by runtime.{region}.kiro.dev
-VALID_RUNTIME_MODEL_IDS: set = {
-    "auto",
-    "claude-sonnet-4",
-    "claude-sonnet-4.5",
-    "claude-sonnet-4.6",
-    "claude-opus-4.5",
-    "claude-opus-4.6",
-    "claude-opus-4.7",
-    "claude-haiku-4.5",
-}
+# Generated from FALLBACK_MODELS to maintain single source of truth
+from kiro.config import FALLBACK_MODELS
+
+VALID_RUNTIME_MODEL_IDS: set = {model["modelId"] for model in FALLBACK_MODELS}
 
 
 def to_runtime_model_id(normalized: str) -> str:

kiro/routes_anthropic.py

@@ -34,7 +34,7 @@
 from fastapi.security import APIKeyHeader
 from loguru import logger
 
-from kiro.config import PROXY_API_KEY
+from kiro.config import PROXY_API_KEY, PROFILE_ARN
 from kiro.models_anthropic import (
     AnthropicMessagesRequest,
     AnthropicCountTokensRequest,
@@ -376,9 +376,8 @@ async def messages(
             conversation_id = generate_conversation_id()
 
             # Build payload for Kiro
-            profile_arn_for_payload = ""
-            if auth_manager.auth_type == AuthType.KIRO_DESKTOP and auth_manager.profile_arn:
-                profile_arn_for_payload = auth_manager.profile_arn
+            # profileArn is required by runtime.kiro.dev for all auth types
+            profile_arn_for_payload = auth_manager.profile_arn or PROFILE_ARN or ""
 
             try:
                 kiro_payload = anthropic_to_kiro(
@@ -686,7 +685,7 @@ async def make_retry_request():
 
     # Build payload for Kiro
     # profileArn is required by runtime.kiro.dev for all auth types
-    profile_arn_for_payload = auth_manager.profile_arn or ""
+    profile_arn_for_payload = auth_manager.profile_arn or PROFILE_ARN or ""
 
     try:
         kiro_payload = anthropic_to_kiro(

kiro/routes_openai.py

@@ -37,6 +37,7 @@
 from kiro.config import (
     PROXY_API_KEY,
     APP_VERSION,
+    PROFILE_ARN,
 )
 from kiro.models_openai import (
     OpenAIModel,
@@ -322,9 +323,8 @@ async def chat_completions(request: Request, request_data: ChatCompletionRequest
             conversation_id = generate_conversation_id()
 
             # Build payload for Kiro
-            profile_arn_for_payload = ""
-            if auth_manager.auth_type == AuthType.KIRO_DESKTOP and auth_manager.profile_arn:
-                profile_arn_for_payload = auth_manager.profile_arn
+            # profileArn is required by runtime.kiro.dev for all auth types
+            profile_arn_for_payload = auth_manager.profile_arn or PROFILE_ARN or ""
 
             try:
                 kiro_payload = build_kiro_payload(
@@ -572,7 +572,7 @@ async def make_retry_request():
 
     # Build payload for Kiro
     # profileArn is required by runtime.kiro.dev for all auth types
-    profile_arn_for_payload = auth_manager.profile_arn or ""
+    profile_arn_for_payload = auth_manager.profile_arn or PROFILE_ARN or ""
 
     try:
         kiro_payload = build_kiro_payload(