fix(tui): /compress shows a before/after summary (#46686)

The TUI /compress slash side-effect compressed the session, synced the
key, and emitted session.info — but returned an empty string, so the
user saw no 'Compressed: N → M messages / ~X → ~Y tokens' feedback. The
CLI (_manual_compress) and gateway (slash_commands) paths both already
call summarize_manual_compression; the TUI slash path was the lone gap.

Snapshot history + rough token estimate before and after compaction and
return the formatted summarize_manual_compression() feedback, mirroring
the session.compress RPC handler. The estimate uses the same
estimate_request_tokens_rough(system_prompt, tools) inputs as the RPC
path, re-reading the system prompt after compaction (it may be rebuilt).

Co-authored-by: liuhao1024 <sunsky.lau@gmail.com>

tests/test_tui_gateway_server.py

@@ -4968,7 +4968,8 @@ def test_mirror_slash_side_effects_allowed_when_idle(monkeypatch):
 def test_mirror_slash_compress_does_not_prelock_history(monkeypatch):
     """Regression guard: /compress side effect must not hold history_lock
     when calling _compress_session_history (the helper snapshots under
-    the same non-reentrant lock internally)."""
+    the same non-reentrant lock internally). It also returns a before/after
+    summary string (#46686)."""
     import types
 
     seen = {"compress": False, "sync": False}
@@ -4977,7 +4978,9 @@ def test_mirror_slash_compress_does_not_prelock_history(monkeypatch):
     def _fake_compress(session, focus_topic=None, **_kw):
         seen["compress"] = True
         assert not session["history_lock"].locked()
-        return (0, {"total": 0})
+        # Simulate a real compaction shrinking the transcript.
+        session["history"] = [{"role": "user", "content": "summary"}]
+        return (1, {"total": 0})
 
     def _fake_sync(_sid, _session):
         seen["sync"] = True
@@ -4988,14 +4991,20 @@ def test_mirror_slash_compress_does_not_prelock_history(monkeypatch):
     monkeypatch.setattr(server, "_emit", lambda *args: emitted.append(args))
 
     session = _session(running=False)
-    session["agent"] = types.SimpleNamespace(model="x")
+    session["history"] = [
+        {"role": "user", "content": f"m{i}"} for i in range(6)
+    ]
+    session["agent"] = types.SimpleNamespace(model="x", _cached_system_prompt="", tools=None)
 
     warning = server._mirror_slash_side_effects("sid", session, "/compress")
 
-    assert warning == ""
+    # Now returns a before/after summary (was "" before #46686).
     assert seen["compress"]
     assert seen["sync"]
     assert ("session.info", "sid", {"model": "x"}) in emitted
+    assert "Compressed:" in warning
+    assert "6 → 1 messages" in warning
+    assert "tokens" in warning
 
 
 # ---------------------------------------------------------------------------

tui_gateway/server.py

@@ -9806,9 +9806,49 @@ def _mirror_slash_side_effects(sid: str, session: dict, command: str) -> str:
             agent.ephemeral_system_prompt = new_prompt or None
             agent._cached_system_prompt = None
         elif name == "compress" and agent:
+            # Mirror the session.compress RPC: build a before/after summary so
+            # the user gets feedback (#46686). The slash path previously just
+            # compressed + emitted session.info and returned "", so the TUI
+            # showed no "compressed N → M messages / ~X → ~Y tokens" stats
+            # while CLI and gateway both did.
+            from agent.manual_compression_feedback import summarize_manual_compression
+            from agent.model_metadata import estimate_request_tokens_rough
+
+            with session["history_lock"]:
+                _before_messages = list(session.get("history", []))
+            _before_count = len(_before_messages)
+            _sys_prompt = getattr(agent, "_cached_system_prompt", "") or ""
+            _tools = getattr(agent, "tools", None) or None
+            _before_tokens = (
+                estimate_request_tokens_rough(
+                    _before_messages, system_prompt=_sys_prompt, tools=_tools
+                )
+                if _before_count
+                else 0
+            )
+
             _compress_session_history(session, arg)
             _sync_session_key_after_compress(sid, session)
+
+            with session["history_lock"]:
+                _after_messages = list(session.get("history", []))
+            _sys_prompt_after = getattr(agent, "_cached_system_prompt", "") or _sys_prompt
+            _tools_after = getattr(agent, "tools", None) or _tools
+            _after_tokens = (
+                estimate_request_tokens_rough(
+                    _after_messages, system_prompt=_sys_prompt_after, tools=_tools_after
+                )
+                if _after_messages
+                else 0
+            )
             _emit("session.info", sid, _session_info(agent, session))
+            _fb = summarize_manual_compression(
+                _before_messages, _after_messages, _before_tokens, _after_tokens
+            )
+            _lines = [_fb["headline"], _fb["token_line"]]
+            if _fb.get("note"):
+                _lines.append(_fb["note"])
+            return "\n".join(_lines)
         elif name == "fast" and agent:
             mode = arg.lower()
             if mode in {"fast", "on"}: