feat(bot): every free-text message gets the agent's answer; every demo entry is a private copy

- Idle chat: all messages go to the agent (greetings, side-effect and substitute questions included); check-in
  signals are still recorded silently, quit intent keeps its rule reaction. Seen live: «какие будут побочки»
  was logged as a complaint («Спасибо, отметил»), «привет» got a canned phrase.
- Agent prompt: warm short replies to small talk; substitutes → only the doctor changes the drug.
- ЕМИАС demo persona button creates a private copy too (starting from consent + questionnaires, subject to
  the screening switch); the web app drops a stored session of the old shared demo account.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

app/dialog/flows/veteran_daily.py

@@ -62,6 +62,7 @@ from app.dialog.flows.meds import (
 )
 from app.dialog.fsm import Ctx, Step, abandonable, on_enter, state, trigger
 from app.dialog.scale_session import register_scale_intro
+from app.llm import agent
 
 PREFIX = "daily"
 STEPS = ("mood", "med", "sleep", "nightmares", "alcohol", "text")
@@ -429,6 +430,9 @@ async def _idle_text(ctx: Ctx, event: Event) -> Step:
     assert ctx.text is not None
     facts = await ctx.gateway.extract_checkin(ctx.text)
     if facts is None:
+        smart = await agent.answer_chat(ctx.token, ctx.text)
+        if smart is not None:
+            return Step(replies=[menu_reply(ctx, agent.render(smart))])
         text = await knowledge_answer(ctx, explicit=False) if looks_like_question(ctx.text) else None
         return Step(replies=[menu_reply(ctx, text or await ctx.gateway.phrase("support"))])
     fields: dict[str, Any] = {}
@@ -446,6 +450,14 @@ async def _idle_text(ctx: Ctx, event: Event) -> Step:
             text = (await side_effect_reply(ctx, codes)).text
     if fields:
         await save_checkin(ctx, **fields)
+    if not facts.quit_intent:
+        # Chat first: whatever the patient writes in idle gets the agent's answer; check-in signals above are
+        # recorded silently for the doctor. Quit intent keeps its rule reaction (explanation + doctor notified).
+        smart = await agent.answer_chat(ctx.token, ctx.text)
+        if smart is not None:
+            return Step(
+                replies=[menu_reply(ctx, agent.render(smart))], notify_monitoring=bool(fields or facts.side_effects)
+            )
     if text is None and facts.med_question and not facts.quit_intent:
         text = await answer_question(ctx, await current_prescription(ctx))
     # A question is answered even if it also mentioned alcohol/side effects (those are recorded above).