feat(memory): add memory.recall_indicator to suppress recall status line

agent/agent_init.py

@@ -1771,6 +1771,17 @@ def init_agent(
 
     # Memory provider plugin (external — one at a time, alongside built-in)
     # Reads memory.provider from config to select which plugin to activate.
+    # memory.recall_indicator gates the deterministic recall status line
+    # ("👁️ Hindsight — recalled 3 memories") emitted when external memory is
+    # auto-recalled before a turn. Default true preserves current behaviour;
+    # false keeps recall internal (context still injected, status suppressed).
+    agent._recall_indicator_enabled = True
+    try:
+        agent._recall_indicator_enabled = bool(
+            ((_agent_cfg or {}).get("memory") or {}).get("recall_indicator", True)
+        )
+    except Exception:
+        agent._recall_indicator_enabled = True
     agent._memory_manager = None
     if not skip_memory:
         try:

agent/turn_context.py

@@ -1287,11 +1287,16 @@ def build_turn_context(
         # Deterministic, model-independent recall indicator: when memory was
         # actually injected this turn, tell the user — don't rely on the model
         # to surface it. Rendered by Hermes (via _emit_status), so it always
-        # shows and can't be silently dropped by the model.
+        # shows and can't be silently dropped by the model. Gate: the
+        # memory.recall_indicator config flag (agent._recall_indicator_enabled,
+        # set at init, default true) lets users keep recall internal — memory
+        # still injects into context, but the status line is suppressed.
         if ext_prefetch_cache:
             try:
                 _recall_indicator = agent._memory_manager.describe_recall()
-                if _recall_indicator:
+                if _recall_indicator and getattr(
+                    agent, "_recall_indicator_enabled", True
+                ):
                     agent._emit_status(_recall_indicator)
             except Exception:
                 pass

cli-config.yaml.example

@@ -827,6 +827,13 @@ memory:
   # For exit/reset, only fires if the session had at least this many user turns.
   flush_min_turns: 6        # Min user turns to trigger flush on exit/reset (0 = disabled)
 
+  # Deterministic recall indicator: when an external memory provider
+  # (memory.provider) auto-recalls memory before a turn, Hermes shows a status
+  # line (e.g. "👁️ Hindsight — recalled 3 memories") on the CLI and messaging
+  # platforms so you SEE memory was used. Set false to keep recall internal —
+  # memory still injects into context, but the status line is suppressed.
+  recall_indicator: true
+
 # =============================================================================
 # Session Reset Policy (Messaging Platforms)
 # =============================================================================

hermes_cli/config_defaults.py

@@ -1778,6 +1778,12 @@
         # "hindsight", "holographic", "retaindb", "byterover".
         # Only ONE external provider is allowed at a time.
         "provider": "",
+        # Deterministic recall indicator: when an external provider auto-recalls
+        # memory before a turn, Hermes shows a status line (e.g.
+        # "👁️ Hindsight — recalled 3 memories") so you SEE memory was used.
+        # Set false to keep recall internal: memory still injects into context,
+        # but the status line is suppressed on every surface (CLI + gateway).
+        "recall_indicator": True,
     },
 
     # Subagent delegation — override the provider:model used by delegate_task

tests/agent/test_turn_context.py

@@ -365,6 +365,27 @@ def test_recall_indicator_skipped_when_nothing_injected():
         assert "👁️" not in str(call)
 
 
+def test_recall_indicator_suppressed_when_config_disabled():
+    """memory.recall_indicator: false keeps recall internal — no status line.
+
+    Memory still injects into context (prefetch_all runs and returns content),
+    but the deterministic indicator must not be emitted on any surface.
+    """
+    agent = _FakeAgent()
+    agent._emit_status = MagicMock()
+    agent._recall_indicator_enabled = False
+    mm = MagicMock()
+    mm.prefetch_all.return_value = "- recalled fact"
+    mm.describe_recall.return_value = "👁️ Hindsight — recalled 2 memories"
+    agent._memory_manager = mm
+
+    _build(agent, user_message="what did we decide about the deploy pipeline?")
+
+    mm.describe_recall.assert_called_once()
+    for call in agent._emit_status.call_args_list:
+        assert "👁️" not in str(call)
+
+
 def test_ensure_db_session_runs_after_system_prompt_restore():
     """Regression for #45499.