Fix _build_prompt receiving state dict instead of messages list

LangGraph passes the full agent state dict {"messages": [...]} to the
prompt callable, not just the messages list. Iterating a dict yields
its keys, so "messages" (the string) was getting appended to the LLM
input — causing the agent to respond as if the literal word "messages"
was the user's query.

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

blueprints/conversation/agents.py

@@ -1232,11 +1232,13 @@ if google_gmail_enabled:
 
 
 @RunnableLambda
-def _build_prompt(messages: list, config: RunnableConfig = None) -> list:
+def _build_prompt(state, config: RunnableConfig = None) -> list:
     """Inject the per-request system prompt, replacing any prior system messages."""
     system_prompt = (
         (config or {}).get("configurable", {}).get("system_prompt", SIMBA_SYSTEM_PROMPT)
     )
+    # LangGraph passes the full state dict {"messages": [...]}, not just the list
+    messages = state["messages"] if isinstance(state, dict) else state
     non_system = [m for m in messages if not isinstance(m, SystemMessage)]
     return [SystemMessage(content=system_prompt)] + non_system