Merge pull request ChesterRa#41 from dweb-channel/main

perf: optimize group switching and fix blob 404 errors

Author: waterbang

README.ja.md

@@ -69,6 +69,23 @@ cccc
 >
 > エラーが発生した場合は、診断と解決を手伝ってください。
 
+### 旧バージョンからのアップグレード
+
+古いバージョンの cccc-pair（例：0.3.x）がインストールされている場合は、先にアンインストールが必要です：
+
+```bash
+# pipx ユーザー
+pipx uninstall cccc-pair
+
+# pip ユーザー
+pip uninstall cccc-pair
+
+# 残留ファイルがある場合は手動で削除
+rm -f ~/.local/bin/cccc ~/.local/bin/ccccd
+```
+
+> **注意**：0.4.x のコマンド構造は 0.3.x と完全に異なります。旧版の `init`、`run`、`bridge` コマンドは `attach`、`daemon`、`mcp` などに置き換えられました。
+
 ### TestPyPI からインストール（推奨）
 
 ```bash

README.md

@@ -69,6 +69,23 @@ Copy this prompt to your AI assistant (Claude, ChatGPT, etc.):
 >
 > If you encounter any errors, please help me diagnose and resolve them.
 
+### Upgrading from older versions
+
+If you have an older version of cccc-pair installed (e.g., 0.3.x), you must uninstall it first:
+
+```bash
+# For pipx users
+pipx uninstall cccc-pair
+
+# For pip users
+pip uninstall cccc-pair
+
+# Remove any leftover binaries if needed
+rm -f ~/.local/bin/cccc ~/.local/bin/ccccd
+```
+
+> **Note**: Version 0.4.x has a completely different command structure from 0.3.x. The old `init`, `run`, `bridge` commands are replaced with `attach`, `daemon`, `mcp`, etc.
+
 ### From TestPyPI (recommended)
 
 ```bash

README.zh-CN.md

@@ -69,6 +69,23 @@ cccc
 >
 > 如果遇到任何错误，请帮我诊断并解决。
 
+### 从旧版本升级
+
+如果你已安装旧版本的 cccc-pair（如 0.3.x），必须先卸载：
+
+```bash
+# pipx 用户
+pipx uninstall cccc-pair
+
+# pip 用户
+pip uninstall cccc-pair
+
+# 如有残留，手动删除
+rm -f ~/.local/bin/cccc ~/.local/bin/ccccd
+```
+
+> **注意**：0.4.x 版本的命令结构与 0.3.x 完全不同。旧版的 `init`、`run`、`bridge` 命令已被 `attach`、`daemon`、`mcp` 等替代。
+
 ### 从 TestPyPI 安装（推荐）
 
 ```bash

docs/guide/faq.md

@@ -10,14 +10,31 @@ Frequently asked questions about CCCC.
 # From TestPyPI (RC version)
 python -m pip install --index-url https://pypi.org/simple \
   --extra-index-url https://test.pypi.org/simple \
-  cccc-pair==0.4.0rc16
+  cccc-pair==0.4.0rc17
 
 # From source
-git clone https://github.com/ChesterRa/cccc
+git clone https://github.com/dweb-channel/cccc
 cd cccc
 pip install -e .
 ```
 
+### How do I upgrade from an older version (0.3.x)?
+
+You must uninstall the old version first:
+
+```bash
+# For pipx users
+pipx uninstall cccc-pair
+
+# For pip users
+pip uninstall cccc-pair
+
+# Remove any leftover binaries
+rm -f ~/.local/bin/cccc ~/.local/bin/ccccd
+```
+
+Then install the new version. Note that 0.4.x has a completely different command structure from 0.3.x.
+
 ### What are the system requirements?
 
 - Python 3.9+

docs/guide/getting-started/index.md

@@ -41,18 +41,37 @@ Both approaches require:
 
 ## Installation
 
+### Upgrading from older versions
+
+If you have an older version of cccc-pair installed (e.g., 0.3.x), you must uninstall it first:
+
+```bash
+# For pipx users
+pipx uninstall cccc-pair
+
+# For pip users
+pip uninstall cccc-pair
+
+# Remove any leftover binaries if needed
+rm -f ~/.local/bin/cccc ~/.local/bin/ccccd
+```
+
+::: warning Version 0.4.x Breaking Changes
+Version 0.4.x has a completely different command structure from 0.3.x. The old `init`, `run`, `bridge` commands are replaced with `attach`, `daemon`, `mcp`, etc.
+:::
+
 ### From TestPyPI (recommended for RC)
 
 ```bash
 python -m pip install --index-url https://pypi.org/simple \
   --extra-index-url https://test.pypi.org/simple \
-  cccc-pair==0.4.0rc16
+  cccc-pair==0.4.0rc17
 ```
 
 ### From Source
 
 ```bash
-git clone https://github.com/ChesterRa/cccc
+git clone https://github.com/dweb-channel/cccc
 cd cccc
 pip install -e .
 ```

docs/vnext/RELEASE.md

@@ -40,5 +40,5 @@ The release workflow is tag-driven (`v*`) and enforces that the git tag matches
 ```bash
 python -m pip install --index-url https://pypi.org/simple \
   --extra-index-url https://test.pypi.org/simple \
-  cccc-pair==0.4.0rc16
+  cccc-pair==0.4.0rc17
 ```

src/cccc/cli.py

@@ -3,6 +3,7 @@
 import argparse
 import json
 import os
+import socket
 import subprocess
 import sys
 import time
@@ -465,8 +466,25 @@ def _stop_daemon() -> None:
     )
     server = uvicorn.Server(config)
 
+    # Get LAN IP for display
+    def _get_lan_ip() -> str:
+        try:
+            # Create a socket to an external address to determine the local IP
+            s = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+            s.settimeout(0.1)
+            s.connect(("8.8.8.8", 80))
+            ip = s.getsockname()[0]
+            s.close()
+            return ip
+        except Exception:
+            return ""
+
     try:
         print("[cccc] Starting web server...", file=sys.stderr)
+        print(f"[cccc]   Local:   http://{host}:{port}", file=sys.stderr)
+        lan_ip = _get_lan_ip()
+        if lan_ip and lan_ip != host and lan_ip != "127.0.0.1":
+            print(f"[cccc]   Network: http://{lan_ip}:{port}", file=sys.stderr)
         server.run()
     except (SystemExit, KeyboardInterrupt):
         pass

src/cccc/daemon/automation.py

@@ -519,7 +519,14 @@ def _base_dt(ts: str) -> Optional[datetime]:
             _queue_notify_to_pty(group, actor_id=aid, runner_kind=runner_kind, ev=ev, notify=notify_data)
 
     def _check_actor_idle(self, group: Group, cfg: AutomationConfig, now: datetime) -> None:
-        """Check for idle actors and notify foreman."""
+        """Check for idle actors and notify foreman.
+
+        Idle detection uses multiple signals:
+        1. PTY output activity (for pty runners) - most accurate for CLI agents
+        2. Ledger activity (last event by actor) - fallback for headless runners
+
+        An actor is considered idle only if BOTH signals indicate inactivity.
+        """
         if cfg.actor_idle_timeout_seconds <= 0:
             return
 
@@ -552,12 +559,31 @@ def _check_actor_idle(self, group: Group, cfg: AutomationConfig, now: datetime)
                     if not pty_runner.SUPERVISOR.actor_running(group.group_id, aid):
                         continue
 
-                # Get last activity
+                # Get idle time from PTY (if applicable) - this is the most accurate signal
+                # for CLI-based agents that produce terminal output while working
+                pty_idle_seconds: Optional[float] = None
+                if runner_kind != "headless":
+                    pty_idle_seconds = pty_runner.SUPERVISOR.idle_seconds(
+                        group_id=group.group_id, actor_id=aid
+                    )
+
+                # Get last activity from ledger (fallback)
                 last_activity = _get_last_actor_activity(group, aid)
-                if last_activity is None:
-                    continue  # No activity yet, skip
-                
-                idle_seconds = (now - last_activity).total_seconds()
+                ledger_idle_seconds: Optional[float] = None
+                if last_activity is not None:
+                    ledger_idle_seconds = (now - last_activity).total_seconds()
+
+                # Determine effective idle time:
+                # - For PTY actors: use PTY idle time (more accurate)
+                # - For headless: use ledger idle time
+                # - If PTY shows recent activity, actor is NOT idle even if ledger is old
+                if pty_idle_seconds is not None:
+                    idle_seconds = pty_idle_seconds
+                elif ledger_idle_seconds is not None:
+                    idle_seconds = ledger_idle_seconds
+                else:
+                    continue  # No activity data, skip
+
                 if idle_seconds < float(cfg.actor_idle_timeout_seconds):
                     continue
 
@@ -569,7 +595,7 @@ def _check_actor_idle(self, group: Group, cfg: AutomationConfig, now: datetime)
                         # Don't notify again within the timeout period
                         if (now - last_notify_dt).total_seconds() < float(cfg.actor_idle_timeout_seconds):
                             continue
-                
+
                 st["last_idle_notify_at"] = utc_now_iso()
                 to_notify.append((aid, idle_seconds))
 

src/cccc/daemon/server.py

@@ -38,6 +38,7 @@
 from ..util.conv import coerce_bool
 from ..util.obslog import setup_root_json_logging
 from ..util.fs import atomic_write_json, atomic_write_text, read_json
+from ..util.file_lock import acquire_lockfile, release_lockfile, LockUnavailableError
 from ..util.time import utc_now_iso
 from .automation import AutomationManager
 from .delivery import (
@@ -965,37 +966,30 @@ def _maybe_autostart_enabled_im_bridges() -> None:
             pass
         log_path = state_dir / "im_bridge.log"
 
-        log_file = None
         try:
-            log_file = log_path.open("a", encoding="utf-8")
-            env = os.environ.copy()
-            env["CCCC_HOME"] = str(home)
-            proc = subprocess.Popen(
-                [sys.executable, "-m", "cccc.ports.im.bridge", gid, platform],
-                env=env,
-                stdout=log_file,
-                stderr=log_file,
-                stdin=subprocess.DEVNULL,
-                start_new_session=True,
-                cwd=str(home),
-            )
-            time.sleep(0.25)
-            rc = proc.poll()
-            if rc is not None:
-                logger.warning("IM bridge autostart failed for %s (platform=%s, code=%s). See log: %s", gid, platform, rc, log_path)
-                continue
-            try:
-                pid_path.write_text(str(proc.pid), encoding="utf-8")
-            except Exception:
-                pass
+            with log_path.open("a", encoding="utf-8") as log_file:
+                env = os.environ.copy()
+                env["CCCC_HOME"] = str(home)
+                proc = subprocess.Popen(
+                    [sys.executable, "-m", "cccc.ports.im.bridge", gid, platform],
+                    env=env,
+                    stdout=log_file,
+                    stderr=log_file,
+                    stdin=subprocess.DEVNULL,
+                    start_new_session=True,
+                    cwd=str(home),
+                )
+                time.sleep(0.25)
+                rc = proc.poll()
+                if rc is not None:
+                    logger.warning("IM bridge autostart failed for %s (platform=%s, code=%s). See log: %s", gid, platform, rc, log_path)
+                    continue
+                try:
+                    pid_path.write_text(str(proc.pid), encoding="utf-8")
+                except Exception:
+                    pass
         except Exception as e:
             logger.warning("IM bridge autostart failed for %s (platform=%s): %s", gid, platform, e)
-        finally:
-            try:
-                if log_file:
-                    log_file.close()
-            except Exception:
-                pass
 
 
 def _maybe_autostart_running_groups() -> None:
@@ -3857,6 +3851,15 @@ def serve_forever(paths: Optional[DaemonPaths] = None) -> int:
     p = paths or default_paths()
     p.daemon_dir.mkdir(parents=True, exist_ok=True)
 
+    # Acquire exclusive lock to prevent multiple daemon instances (race condition fix).
+    # The lock is held for the lifetime of the daemon process.
+    lock_path = p.daemon_dir / "ccccd.lock"
+    try:
+        lock_handle = acquire_lockfile(lock_path, blocking=False)
+    except LockUnavailableError:
+        # Another daemon already holds the lock
+        return 0
+
     # Apply global observability settings early (logging + developer mode gating).
     try:
         _apply_observability_settings(p.home, get_observability_settings())
@@ -3865,6 +3868,7 @@ def serve_forever(paths: Optional[DaemonPaths] = None) -> int:
 
     _cleanup_stale_daemon_endpoints(p)
     if _is_daemon_alive(p):
+        release_lockfile(lock_handle)
         return 0
 
     # Cleanup stale IM bridge state from previous runs/crashes.
@@ -4231,6 +4235,13 @@ def _bootstrap_after_listen() -> None:
             p.pid_path.unlink()
     except Exception:
         pass
+
+    # Release the daemon lock
+    try:
+        release_lockfile(lock_handle)
+    except Exception:
+        pass
+
     return 0
 
 

src/cccc/daemon/streaming.py

@@ -1,6 +1,7 @@
 from __future__ import annotations
 
 import json
+import logging
 import queue
 import socket
 import threading
@@ -12,6 +13,8 @@
 from ..kernel.inbox import is_message_for_actor
 from ..util.time import parse_utc_iso, utc_now_iso
 
+logger = logging.getLogger(__name__)
+
 
 STREAMABLE_KINDS_V1: Set[str] = {
     "chat.message",
@@ -178,7 +181,26 @@ def publish(self, event: Dict[str, Any]) -> None:
             try:
                 sub.q.put_nowait(event)
             except queue.Full:
-                self.close(sub)
+                # Backpressure: drop oldest events instead of closing the connection
+                dropped = 0
+                try:
+                    # Drop up to 10% of queue capacity to make room
+                    drop_count = max(1, sub.q.maxsize // 10)
+                    for _ in range(drop_count):
+                        try:
+                            sub.q.get_nowait()
+                            dropped += 1
+                        except queue.Empty:
+                            break
+                    sub.q.put_nowait(event)
+                except queue.Full:
+                    # Still full after dropping, give up on this event
+                    dropped += 1
+                if dropped > 0:
+                    logger.warning(
+                        "Event stream backpressure: dropped %d event(s) for sub %s (group=%s, by=%s)",
+                        dropped, sub.sub_id, sub.group_id, sub.by
+                    )
 
 
 EVENT_BROADCASTER = EventBroadcaster()