v2.2.1: fix compaction orphaning tool outputs causing Crof incomplete

- Fix compaction cutting between function_call and function_call_output pairs
- Orphaned tool results confused Crof models causing finish_reason=length
- reasoning_effort=none now always sends enable_thinking=false too
- Added Crof upstream debug logging

CHANGELOG.md

@@ -1,5 +1,15 @@
 # Changelog
 
+## v2.2.1 (2026-05-20)
+
+- **Fixed compaction orphaning function_call_output items** — root cause of Crof `incomplete` responses
+  - Compaction cut between function_call and its function_call_output, creating dangling tool results
+  - Crof model received orphaned `tool` messages with empty `tool_call_id`, causing confusion and token exhaustion
+  - Compaction now expands tail boundary to include matching function_call/function_call_output pairs
+- **Fixed reasoning control**: `reasoning_effort=none` now always sends both `enable_thinking=false` AND `reasoning_effort=none`
+  - Crof API testing confirmed `reasoning_effort=none` is what actually suppresses reasoning, not `enable_thinking=false`
+- Added upstream debug logging to `~/.cache/codex-proxy/crof-upstream.jsonl`
+
 ## v2.2.0 (2026-05-20)
 
 - **Added per-provider Reasoning controls in endpoint editor**

src/codex-launcher-gui

@@ -24,6 +24,11 @@ model_catalog_json = ""
 """
 
 CHANGELOG = [
+    ("2.2.1", "2026-05-20", [
+        "Fixed compaction orphaning function_call_output items — root cause of Crof incomplete responses",
+        "Compaction now respects function_call/function_call_output pairs — no more dangling tool results",
+        "Fixed reasoning control: reasoning_effort=none now always sends enable_thinking=false too",
+    ]),
     ("2.2.0", "2026-05-20", [
         "Added per-provider Reasoning On/Off toggle in endpoint editor",
         "Added Reasoning Effort level per provider: None, Minimal, Low, Medium, High, Max",
@@ -543,7 +548,7 @@ class LauncherWin(Gtk.Window):
         # header row
         hdr = Gtk.Box(spacing=8)
         vbox.pack_start(hdr, False, False, 0)
-        lbl = Gtk.Label(label="<b>Codex Launcher v2.2.0</b>")
+        lbl = Gtk.Label(label="<b>Codex Launcher v2.2.1</b>")
         lbl.set_use_markup(True)
         hdr.pack_start(lbl, False, False, 0)
         changelog_btn = Gtk.Button(label="Changelog")

src/translate-proxy.py

@@ -238,8 +238,16 @@ def _compact_input(input_data):
             break
 
     head = input_data[:head_end]
-    tail = input_data[-_COMPACT_KEEP_RECENT:]
-    body = input_data[head_end:-_COMPACT_KEEP_RECENT]
+    tail_start = len(input_data) - _COMPACT_KEEP_RECENT
+    while tail_start > head_end:
+        if input_data[tail_start].get("type") == "function_call_output":
+            tail_start -= 1
+        elif input_data[tail_start].get("type") == "message" and input_data[tail_start].get("role") == "assistant":
+            tail_start -= 1
+        else:
+            break
+    tail = input_data[tail_start:]
+    body = input_data[head_end:tail_start]
 
     if not body:
         return head + tail
@@ -891,9 +899,11 @@ class Handler(http.server.BaseHTTPRequestHandler):
         if body.get("tool_choice"):
             chat_body["tool_choice"] = body["tool_choice"]
         chat_body["stream"] = stream
-        if not REASONING_ENABLED:
+        if not REASONING_ENABLED or REASONING_EFFORT == "none":
             chat_body["enable_thinking"] = False
-        chat_body["reasoning_effort"] = REASONING_EFFORT if REASONING_ENABLED else "none"
+            chat_body["reasoning_effort"] = "none"
+        else:
+            chat_body["reasoning_effort"] = REASONING_EFFORT
 
         target = upstream_target(TARGET_URL, "/chat/completions")
         fwd = forwarded_headers(self.headers, {
@@ -901,6 +911,16 @@ class Handler(http.server.BaseHTTPRequestHandler):
             "Authorization": f"Bearer {API_KEY}",
         }, browser_ua=True)
         print(f"[translate-proxy] POST {target} model={model} stream={stream} ua={fwd.get('User-Agent','')[:50]}", file=sys.stderr)
+        _crof_debug_path = os.path.join(_LOG_DIR, "crof-upstream.jsonl")
+        with open(_crof_debug_path, "a") as _cdf:
+            _cdf.write(json.dumps({
+                "ts": _ts, "model": model, "max_tokens": chat_body.get("max_tokens"),
+                "reasoning_enabled": REASONING_ENABLED, "reasoning_effort": chat_body.get("reasoning_effort"),
+                "enable_thinking": chat_body.get("enable_thinking", "NOT_SENT"),
+                "n_messages": len(chat_body.get("messages", [])),
+                "has_tools": bool(chat_body.get("tools")),
+                "messages_summary": [{"role": m.get("role"), "tc": len(m.get("tool_calls", [])), "content_len": len(str(m.get("content", "")))[:6], "tool_call_id": m.get("tool_call_id")} for m in chat_body.get("messages", [])],
+            }) + "\n")
         req = urllib.request.Request(
             target,
             data=json.dumps(chat_body).encode(),