feat(token-viz): Phase 2 — trim warning, Gemini/Anthropic cache status display

gui_2.py

@@ -2765,6 +2765,41 @@ class App:
     imgui.table_set_column_index(1); imgui.text(f"{tok:,}")
     imgui.table_set_column_index(2); imgui.text(f"{tok / total_tok * 100:.0f}%")
    imgui.end_table()
+  if stats.get("would_trim"):
+   imgui.text_colored(imgui.ImVec4(1.0, 0.3, 0.0, 1.0), "WARNING: Next call will trim history")
+   trimmable = stats.get("trimmable_turns", 0)
+   if trimmable:
+    imgui.text_disabled(f"Trimmable turns: {trimmable}")
+   msgs = stats.get("messages")
+   if msgs:
+    shown = 0
+    for msg in msgs:
+     if shown >= 3:
+      break
+     if msg.get("trimmable"):
+      role = msg.get("role", "?")
+      toks = msg.get("tokens", 0)
+      imgui.text_disabled(f"  [{role}] ~{toks:,} tokens")
+      shown += 1
+
+  imgui.separator()
+  if ai_client._provider == "gemini":
+   if ai_client._gemini_cache is not None:
+    age = time.time() - (ai_client._gemini_cache_created_at or time.time())
+    ttl = ai_client._GEMINI_CACHE_TTL
+    imgui.text_colored(C_LBL, f"Gemini Cache: ACTIVE | Age: {age:.0f}s / {ttl}s | Renews at: {ttl * 0.9:.0f}s")
+   else:
+    imgui.text_disabled("Gemini Cache: INACTIVE")
+  elif ai_client._provider == "anthropic":
+   with ai_client._anthropic_history_lock:
+    turns = len(ai_client._anthropic_history)
+   cache_reads = 0
+   for entry in reversed(ai_client.get_comms_log()):
+    if entry.get("kind") == "response":
+     cache_reads = (entry.get("payload") or {}).get("usage", {}).get("cache_read_input_tokens") or 0
+     break
+   imgui.text_disabled("Anthropic: 4-breakpoint ephemeral caching (auto-managed)")
+   imgui.text_disabled(f"  {turns} history turns | Cache reads last call: {cache_reads:,}")
 
  def _render_message_panel(self) -> None:
  # LIVE indicator