Support multimodal planner scenes without OCR

runtime/planner_config.py

@@ -19,6 +19,7 @@ TIMEOUT_ENV = "AI_PLANNER_TIMEOUT_SECONDS"
 THINKING_BUDGET_ENV = "AI_PLANNER_THINKING_BUDGET_TOKENS"
 API_KEY_ENV = "AI_PLANNER_API_KEY"
 BASE_URL_ENV = "AI_PLANNER_BASE_URL"
+MULTIMODAL_ENV = "AI_PLANNER_MULTIMODAL"
 
 SUPPORTED_PROVIDERS = frozenset(DEFAULT_MODEL_BY_PROVIDER)
 
@@ -32,6 +33,7 @@ class PlannerConfig:
     thinking_budget_tokens: int | None = None
     api_key: str | None = None
     base_url: str | None = None
+    multimodal: bool = False
 
     def resolved_model(self) -> str:
         return self.model or DEFAULT_MODEL_BY_PROVIDER[self.provider]
@@ -47,6 +49,7 @@ def load_config(env: Mapping[str, str] | None = None) -> PlannerConfig:
         thinking_budget_tokens=_parse_thinking_budget(values.get(THINKING_BUDGET_ENV)),
         api_key=values.get(API_KEY_ENV) or _provider_key(values),
         base_url=values.get(BASE_URL_ENV) or None,
+        multimodal=_parse_bool(values.get(MULTIMODAL_ENV), default=False),
     )