2964 lines
122 KiB
Python
2964 lines
122 KiB
Python
#!/usr/bin/env python3
|
||
"""
|
||
GenericAgent Web2 Bridge.
|
||
|
||
Clear split:
|
||
1) AgentManager: owns GenericAgent instances, sessions and histories.
|
||
2) Transport: HTTP is the command/data channel; WebSocket only pushes small
|
||
session-state notifications.
|
||
|
||
HTTP API:
|
||
GET /status
|
||
GET /config
|
||
POST /config
|
||
GET /model-profiles (+ POST / PUT / DELETE by id)
|
||
GET /sessions
|
||
POST /session/new
|
||
GET /session/{sid}
|
||
DELETE /session/{sid}
|
||
POST /session/{sid}/prompt
|
||
GET /session/{sid}/messages?after=0&limit=200
|
||
POST /session/{sid}/cancel
|
||
POST /services/start body: {"id":"frontends/qqapp.py"}
|
||
POST /services/stop body: {"id":"frontends/qqapp.py"}
|
||
GET /services/logs?id=frontends/qqapp.py&tail=200
|
||
GET /services/panel
|
||
GET /services/capabilities
|
||
GET /services/mykey
|
||
POST /services/mykey body: {"content":"..."}
|
||
POST /services/stop-extras stop conductor + scheduler (127.0.0.1 only)
|
||
POST /services/start-extras start conductor + scheduler (127.0.0.1 only)
|
||
POST /services/bridge/exit stop managed services, then exit bridge (127.0.0.1 only)
|
||
POST /memory/import/inspect validate a data backup or legacy folder
|
||
POST /memory/import safely merge memory and sessions
|
||
POST /memory/export write a point-in-time data backup ZIP
|
||
|
||
WS API (state sync):
|
||
GET /ws -> on connect sends services.snapshot; service.changed on updates
|
||
{"type":"services.snapshot","services":[...]}
|
||
{"type":"service.changed","service":{...}}
|
||
"""
|
||
from __future__ import annotations
|
||
|
||
import asyncio, atexit, contextlib, copy, importlib, json, os, re, shutil, subprocess, sys
|
||
from collections import Counter, deque
|
||
import threading, time, traceback, uuid
|
||
from dataclasses import dataclass, field
|
||
from pathlib import Path
|
||
from typing import Any, Dict, List, Optional, Set
|
||
from aiohttp import web, WSMsgType
|
||
from data_backup import (
|
||
BackupFormatError,
|
||
canonical_session_record,
|
||
export_data_backup,
|
||
inspect_import_source,
|
||
materialize_import_source,
|
||
merge_data_files,
|
||
)
|
||
from desktop_settings import DesktopSettingsError, read_settings, update_settings
|
||
|
||
APP_DIR = Path(__file__).resolve().parent
|
||
DEFAULT_CONDUCTOR_PORT = 8900
|
||
E2E_CONDUCTOR_PORT_ENV = "GA_DESKTOP_E2E_CONDUCTOR_PORT"
|
||
E2E_REPORT_DIR_ENV = "GA_DESKTOP_E2E_REPORT_DIR"
|
||
|
||
|
||
def _configured_conductor_port() -> int:
|
||
"""Return the package-journey conductor port without changing production :8900.
|
||
|
||
Desktop's compiled renderer and production CSP intentionally use :8900. The
|
||
alternate port exists only to isolate real-package evidence from an installed
|
||
user's conductor; both E2E variables are required before it is honored.
|
||
"""
|
||
if not os.environ.get(E2E_REPORT_DIR_ENV):
|
||
return DEFAULT_CONDUCTOR_PORT
|
||
raw = os.environ.get(E2E_CONDUCTOR_PORT_ENV)
|
||
if raw is None:
|
||
return DEFAULT_CONDUCTOR_PORT
|
||
value = raw.strip()
|
||
if not value.isascii() or not value.isdigit() or not 1 <= int(value) <= 65535:
|
||
raise RuntimeError(f"{E2E_CONDUCTOR_PORT_ENV} must be an integer between 1 and 65535")
|
||
return int(value)
|
||
|
||
# ─── Bridge self-log ring buffer ───
|
||
import datetime as _dt, io as _io
|
||
|
||
_bridge_log: deque = deque(maxlen=500)
|
||
|
||
|
||
def bridge_print(*args, **kwargs):
|
||
"""Print to stderr AND capture into bridge ring buffer (timestamped)."""
|
||
kwargs.pop("file", None)
|
||
buf = _io.StringIO()
|
||
print(*args, file=buf, **kwargs)
|
||
text = buf.getvalue().rstrip("\n")
|
||
ts = _dt.datetime.now().strftime("%H:%M:%S")
|
||
for line in text.split("\n"):
|
||
_bridge_log.append(f"[{ts}] {line}")
|
||
print(*args, file=sys.stderr, **kwargs)
|
||
|
||
|
||
def _ga_root_override() -> Optional[Path]:
|
||
"""Resolve an external core while keeping this bundled bridge as the executable surface."""
|
||
value = ""
|
||
for index, argument in enumerate(sys.argv):
|
||
if argument == "--ga-root" and index + 1 < len(sys.argv):
|
||
value = sys.argv[index + 1]
|
||
elif argument.startswith("--ga-root="):
|
||
value = argument.split("=", 1)[1]
|
||
if not value:
|
||
value = os.environ.get("GA_ROOT", "")
|
||
value = (value or "").strip()
|
||
if not value:
|
||
return None
|
||
root = Path(value).expanduser().resolve()
|
||
return root if (root / "agentmain.py").exists() else None
|
||
|
||
|
||
def find_default_ga_root() -> Path:
|
||
override = _ga_root_override()
|
||
if override is not None:
|
||
return override
|
||
candidates = [
|
||
APP_DIR / "..",
|
||
APP_DIR / ".." / "..",
|
||
APP_DIR / ".." / "GenericAgent",
|
||
APP_DIR / ".." / ".." / "GenericAgent",
|
||
]
|
||
for p in candidates:
|
||
root = p.resolve()
|
||
if (root / "agentmain.py").exists():
|
||
return root
|
||
return APP_DIR.parent.parent.resolve()
|
||
|
||
|
||
DEFAULT_GA_ROOT = find_default_ga_root()
|
||
|
||
_FINAL_INFO_RE = re.compile(r'\n*`{5}\n*\[Info\] Final response to user\.\n*`{5}\s*$')
|
||
|
||
|
||
def strip_final_info_marker(text: Any) -> str:
|
||
return _FINAL_INFO_RE.sub('', str(text or ''))
|
||
|
||
|
||
def normalize_final_turn_segs(full: str, outputs: Any) -> Optional[List[str]]:
|
||
if not outputs or not isinstance(outputs, (list, tuple)):
|
||
return None
|
||
segs = [strip_final_info_marker(s) for s in outputs]
|
||
full_text = strip_final_info_marker(full)
|
||
if not segs:
|
||
return None
|
||
joined = "".join(segs)
|
||
if full_text.strip() == joined.strip():
|
||
return segs
|
||
if joined and full_text.startswith(joined):
|
||
suffix = full_text[len(joined):]
|
||
if suffix.strip():
|
||
segs[-1] = segs[-1] + suffix
|
||
return segs
|
||
return None
|
||
|
||
|
||
# ─── Empty-turn microcopy (i18n) ───
|
||
|
||
_EMPTY_TURN_MICROCOPY = {
|
||
"zh": '⚠️ 这一轮结束了，但没有产出可见回复。你可以发送"继续"重试。',
|
||
"en": '⚠️ This turn ended without a visible response. You can send "continue" to retry.',
|
||
}
|
||
|
||
|
||
def _get_ui_lang() -> str:
|
||
try:
|
||
p = Path.home() / ".ga_desktop_settings.json"
|
||
doc = json.loads(p.read_text(encoding="utf-8")) if p.is_file() else {}
|
||
if not isinstance(doc, dict):
|
||
return "zh"
|
||
ui = doc.get("ui")
|
||
candidates = (
|
||
ui.get("lang") if isinstance(ui, dict) else None,
|
||
doc.get("lang"), # legacy Desktop settings
|
||
)
|
||
return next(
|
||
(lang for lang in candidates
|
||
if isinstance(lang, str) and lang in _EMPTY_TURN_MICROCOPY),
|
||
"zh",
|
||
)
|
||
except Exception:
|
||
return "zh"
|
||
|
||
|
||
def empty_turn_fallback() -> str:
|
||
lang = _get_ui_lang()
|
||
return _EMPTY_TURN_MICROCOPY.get(lang, _EMPTY_TURN_MICROCOPY["zh"])
|
||
|
||
|
||
# ─── Test-only control plane ────────────────────────────────────────────────
|
||
|
||
_E2E_CONTROL_LOCK = threading.Lock()
|
||
_E2E_NEXT_TURN: Optional[str] = None
|
||
|
||
|
||
def _e2e_control_token() -> Optional[str]:
|
||
if os.environ.get("GA_E2E") != "1":
|
||
return None
|
||
token = os.environ.get("GA_E2E_CONTROL_TOKEN", "").strip()
|
||
return token or None
|
||
|
||
|
||
def _set_e2e_next_turn(mode: str) -> None:
|
||
if mode != "empty":
|
||
raise ValueError("mode must be empty")
|
||
global _E2E_NEXT_TURN
|
||
with _E2E_CONTROL_LOCK:
|
||
_E2E_NEXT_TURN = mode
|
||
|
||
|
||
def _consume_e2e_next_turn() -> Optional[str]:
|
||
global _E2E_NEXT_TURN
|
||
with _E2E_CONTROL_LOCK:
|
||
mode = _E2E_NEXT_TURN
|
||
_E2E_NEXT_TURN = None
|
||
return mode
|
||
|
||
|
||
for _s in (sys.stdout, sys.stderr):
|
||
with contextlib.suppress(Exception):
|
||
_s.reconfigure(encoding="utf-8", errors="replace")
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Agent management layer
|
||
# ---------------------------------------------------------------------------
|
||
|
||
@dataclass
|
||
class Session:
|
||
id: str
|
||
title: str = "New chat"
|
||
cwd: str = ""
|
||
created_at: float = field(default_factory=time.time)
|
||
updated_at: float = field(default_factory=time.time)
|
||
messages: List[dict] = field(default_factory=list)
|
||
msg_seq: int = 0
|
||
partial: Optional[dict] = None
|
||
status: str = "idle" # idle|running|error|cancelled
|
||
agent: Any = None
|
||
thread: Optional[threading.Thread] = None
|
||
cancel_requested: bool = False
|
||
active_turn_id: str = ""
|
||
last_error: str = ""
|
||
pinned: bool = False
|
||
untitled: bool = True
|
||
plan_scan_baseline: int = 0
|
||
plan_path: str = ""
|
||
llm_history: Optional[List[dict]] = None
|
||
# 该会话绑定的模型下标(mykey.py 配置块顺序,== agent.llmclients 下标)。
|
||
# None = 未绑定,发消息时回退到全局默认 ui.llmNo,保持旧会话平滑迁移。
|
||
llm_no: Optional[int] = None
|
||
# 当前正在执行的 turn 使用的模型。仅运行期存在，不写入 session JSON。
|
||
running_llm_no: Optional[int] = None
|
||
running_model: Optional[str] = None
|
||
|
||
|
||
class MaintenanceConflict(RuntimeError):
|
||
def __init__(
|
||
self,
|
||
message: str,
|
||
*,
|
||
code: str = "maintenance_conflict",
|
||
running_sessions: Optional[List[str]] = None,
|
||
running_extras: Optional[List[str]] = None,
|
||
):
|
||
super().__init__(message)
|
||
self.code = code
|
||
self.running_sessions = sorted(set(running_sessions or []))
|
||
self.running_extras = sorted(set(running_extras or []))
|
||
|
||
def payload(self) -> dict:
|
||
return {
|
||
"ok": False,
|
||
"error": str(self),
|
||
"code": self.code,
|
||
"runningSessions": self.running_sessions,
|
||
"runningExtras": self.running_extras,
|
||
}
|
||
|
||
|
||
def _is_desktop_session_id(session_id: Any) -> bool:
|
||
"""Keep internal TUI/Conductor worker artifacts out of Desktop sessions."""
|
||
if not isinstance(session_id, str):
|
||
return False
|
||
value = session_id
|
||
return (
|
||
not value.startswith("tui_")
|
||
and re.fullmatch(r"[A-Za-z0-9][A-Za-z0-9_-]{0,127}", value) is not None
|
||
)
|
||
|
||
|
||
def _load_plan_baseline(item: dict, msgs: list) -> int:
|
||
"""Persisted per-session baseline (tuiapp_v2: set on /continue, not on preset text)."""
|
||
base = int(item.get("plan_scan_baseline", 0) or 0)
|
||
if base >= len(msgs):
|
||
return 0
|
||
return max(0, base)
|
||
|
||
|
||
def _sanitize_desktop_plan_path(session_id: str, plan_path: str) -> str:
|
||
"""Keep only real plan-mode paths; never invent a placeholder path on load."""
|
||
import plan_state
|
||
p = (plan_path or "").strip()
|
||
if not p:
|
||
return ""
|
||
if plan_state.is_plan_mode_path(p):
|
||
return p.lstrip("./\\")
|
||
return ""
|
||
|
||
|
||
class AgentManager:
|
||
def __init__(self):
|
||
self.lock = threading.RLock()
|
||
self.ga_root = str(DEFAULT_GA_ROOT)
|
||
self.config: Dict[str, Any] = {}
|
||
self.sessions: Dict[str, Session] = {}
|
||
self._retired_sessions: Dict[str, Session] = {}
|
||
self._maintenance_token: Optional[str] = None
|
||
self._maintenance_kind: Optional[str] = None
|
||
self._shutdown_requested = False
|
||
self.active_session_id: Optional[str] = None
|
||
self._sessions_dir = Path(self.ga_root) / "temp" / "desktop_sessions"
|
||
# Legacy monolithic store; migrated into _sessions_dir on first load, then retired.
|
||
self._sessions_file = Path(self.ga_root) / "temp" / "desktop_sessions.json"
|
||
self._load_sessions()
|
||
|
||
@property
|
||
def mykey_path(self) -> str:
|
||
return str(Path(self.ga_root) / "mykey.py")
|
||
|
||
def _session_dict(self, s: "Session") -> dict:
|
||
llm_hist = None
|
||
if s.agent and hasattr(s.agent, 'llmclient'):
|
||
try: llm_hist = s.agent.llmclient.backend.history
|
||
except Exception: pass
|
||
if llm_hist is None:
|
||
llm_hist = s.llm_history
|
||
return {"id": s.id, "title": s.title, "cwd": s.cwd,
|
||
"created_at": s.created_at, "updated_at": s.updated_at,
|
||
"messages": s.messages, "msg_seq": s.msg_seq,
|
||
"pinned": s.pinned, "untitled": s.untitled,
|
||
"plan_scan_baseline": s.plan_scan_baseline,
|
||
"plan_path": s.plan_path or "",
|
||
"llm_no": s.llm_no,
|
||
"llm_history": llm_hist}
|
||
|
||
def _session_file(self, sid: str) -> Path:
|
||
return self._sessions_dir / f"{sid}.json"
|
||
|
||
@staticmethod
|
||
def _session_has_unfinished_work(session: "Session") -> bool:
|
||
agent = session.agent
|
||
task_queue = getattr(agent, "task_queue", None)
|
||
unfinished = int(getattr(task_queue, "unfinished_tasks", 0) or 0)
|
||
agent_running = bool(getattr(agent, "is_running", False))
|
||
thread_running = bool(session.thread and session.thread.is_alive())
|
||
return session.status == "running" or unfinished > 0 or agent_running or thread_running
|
||
|
||
def _running_session_ids_locked(self) -> List[str]:
|
||
retired_sessions = getattr(self, "_retired_sessions", {})
|
||
self._retired_sessions = retired_sessions
|
||
for sid, retired in list(retired_sessions.items()):
|
||
if not self._session_has_unfinished_work(retired):
|
||
retired_sessions.pop(sid, None)
|
||
running = [
|
||
sid for sid, session in self.sessions.items()
|
||
if self._session_has_unfinished_work(session)
|
||
]
|
||
running.extend(
|
||
sid for sid, session in retired_sessions.items()
|
||
if self._session_has_unfinished_work(session)
|
||
)
|
||
return sorted(set(running))
|
||
|
||
def _assert_mutation_allowed_locked(self) -> None:
|
||
if getattr(self, "_shutdown_requested", False):
|
||
raise MaintenanceConflict(
|
||
"Desktop bridge shutdown is in progress",
|
||
code="shutdown_in_progress",
|
||
)
|
||
if getattr(self, "_maintenance_token", None) is not None:
|
||
raise MaintenanceConflict(
|
||
f"data {getattr(self, '_maintenance_kind', None) or 'maintenance'} is in progress"
|
||
)
|
||
|
||
@contextlib.contextmanager
|
||
def mutation(self):
|
||
with self.lock:
|
||
self._assert_mutation_allowed_locked()
|
||
yield
|
||
|
||
def begin_maintenance(self, kind: str, running_extras_fn) -> str:
|
||
with self.lock:
|
||
self._assert_mutation_allowed_locked()
|
||
running_sessions = self._running_session_ids_locked()
|
||
running_extras = list(running_extras_fn())
|
||
if running_sessions or running_extras:
|
||
raise MaintenanceConflict(
|
||
"stop running sessions and managed services before data maintenance",
|
||
running_sessions=running_sessions,
|
||
running_extras=running_extras,
|
||
)
|
||
token = uuid.uuid4().hex
|
||
self._maintenance_token = token
|
||
self._maintenance_kind = kind
|
||
return token
|
||
|
||
def end_maintenance(self, token: str) -> None:
|
||
with self.lock:
|
||
if self._maintenance_token != token:
|
||
raise RuntimeError("maintenance token does not own the active gate")
|
||
self._maintenance_token = None
|
||
self._maintenance_kind = None
|
||
|
||
def _persist_session(self, s: "Session", *, strict: bool = False):
|
||
"""Write a single session file. Cost is O(one session), independent of how many
|
||
sessions exist — this is the fix for the monolithic-file scaling problem."""
|
||
tmp = self._sessions_dir / f"{s.id}.json.tmp"
|
||
try:
|
||
self._sessions_dir.mkdir(parents=True, exist_ok=True)
|
||
with self.lock:
|
||
data = self._session_dict(s)
|
||
data["messages"] = copy.deepcopy(data["messages"])
|
||
if data.get("llm_history"):
|
||
data["llm_history"] = copy.deepcopy(data["llm_history"])
|
||
tmp.write_text(json.dumps(data, ensure_ascii=False, default=str), encoding="utf-8")
|
||
os.replace(tmp, self._session_file(s.id))
|
||
except Exception as e:
|
||
with contextlib.suppress(OSError):
|
||
tmp.unlink()
|
||
bridge_print(f"[bridge] persist session {s.id} failed: {e}")
|
||
if strict:
|
||
raise OSError(f"could not persist session {s.id}: {e}") from e
|
||
|
||
def _delete_session_file(self, sid: str):
|
||
try:
|
||
f = self._session_file(sid)
|
||
if f.exists():
|
||
f.unlink()
|
||
except Exception as e:
|
||
bridge_print(f"[bridge] delete session file {sid} failed: {e}")
|
||
|
||
def _persist(self, *, strict: bool = False):
|
||
"""Write every session (one file each). Used for bulk ops (import) / full flush."""
|
||
with self.lock:
|
||
sessions = list(self.sessions.values())
|
||
for s in sessions:
|
||
self._persist_session(s, strict=strict)
|
||
|
||
def _session_from_item(self, item: dict) -> "Session":
|
||
item = canonical_session_record(item, default_cwd=self.ga_root)
|
||
msgs = item["messages"]
|
||
return Session(id=item["id"], title=item["title"],
|
||
cwd=item["cwd"],
|
||
created_at=item["created_at"],
|
||
updated_at=item["updated_at"],
|
||
messages=msgs,
|
||
msg_seq=item["msg_seq"],
|
||
pinned=item["pinned"],
|
||
untitled=item["untitled"],
|
||
plan_scan_baseline=_load_plan_baseline(item, msgs),
|
||
plan_path=_sanitize_desktop_plan_path(item["id"], item["plan_path"]),
|
||
status="idle", agent=None,
|
||
llm_history=item["llm_history"],
|
||
llm_no=item["llm_no"])
|
||
|
||
def _load_sessions(self):
|
||
# New format: one file per session under temp/desktop_sessions/.
|
||
try:
|
||
if self._sessions_dir.is_dir():
|
||
for f in self._sessions_dir.glob("*.json"):
|
||
try:
|
||
item = json.loads(f.read_text(encoding="utf-8"))
|
||
if not _is_desktop_session_id(item.get("id")):
|
||
continue
|
||
sess = self._session_from_item(item)
|
||
self.sessions[sess.id] = sess
|
||
except Exception as e:
|
||
bridge_print(f"[bridge] load session {f.name} failed: {e}")
|
||
except Exception as e:
|
||
bridge_print(f"[bridge] load sessions dir failed: {e}")
|
||
|
||
# One-time migration from the legacy monolithic desktop_sessions.json.
|
||
try:
|
||
if self._sessions_file.exists():
|
||
arr = json.loads(self._sessions_file.read_text(encoding="utf-8"))
|
||
for item in arr:
|
||
if (not isinstance(item, dict)
|
||
or not _is_desktop_session_id(item.get("id"))
|
||
or item.get("id") in self.sessions):
|
||
continue
|
||
try:
|
||
sess = self._session_from_item(item)
|
||
self.sessions[sess.id] = sess
|
||
self._persist_session(sess)
|
||
except Exception as e:
|
||
bridge_print(f"[bridge] migrate session failed: {e}")
|
||
# Retire the legacy file so we do not migrate again next launch.
|
||
with contextlib.suppress(Exception):
|
||
self._sessions_file.rename(
|
||
self._sessions_file.parent / (self._sessions_file.name + ".migrated"))
|
||
except Exception as e:
|
||
bridge_print(f"[bridge] migrate sessions failed: {e}")
|
||
|
||
if self.sessions:
|
||
self.active_session_id = max(self.sessions.values(), key=lambda s: s.updated_at).id
|
||
|
||
def import_sessions(self, source_dir: str) -> dict:
|
||
"""把 source 的桌面会话合并进当前列表(按 id 去重)。
|
||
|
||
兼容两种源格式:新版 temp/desktop_sessions/<id>.json,以及旧版单文件
|
||
temp/desktop_sessions.json(含已退休的 .migrated)。只落盘新增的会话。
|
||
"""
|
||
src = Path(source_dir).expanduser().resolve()
|
||
items: List[dict] = []
|
||
found = False
|
||
|
||
# New per-session format.
|
||
src_dir = src / "temp" / "desktop_sessions"
|
||
if src_dir.is_dir():
|
||
for f in src_dir.glob("*.json"):
|
||
try:
|
||
items.append(json.loads(f.read_text(encoding="utf-8")))
|
||
found = True
|
||
except Exception:
|
||
continue
|
||
|
||
# Legacy monolithic format (live or already retired).
|
||
for legacy in (src / "temp" / "desktop_sessions.json",
|
||
src / "temp" / "desktop_sessions.json.migrated"):
|
||
if legacy.is_file():
|
||
found = True
|
||
try:
|
||
arr = json.loads(legacy.read_text(encoding="utf-8"))
|
||
if isinstance(arr, list):
|
||
items.extend(x for x in arr if isinstance(x, dict))
|
||
except Exception:
|
||
continue
|
||
|
||
if not found:
|
||
return {"sessionsAdded": 0, "sessionsSkipped": 0, "sessionsFileFound": False}
|
||
|
||
added = 0
|
||
skipped = 0
|
||
new_sessions: List["Session"] = []
|
||
with self.mutation():
|
||
for item in items:
|
||
sid = item.get("id")
|
||
if not _is_desktop_session_id(sid) or sid in self.sessions:
|
||
skipped += 1
|
||
continue
|
||
try:
|
||
sess = self._session_from_item(item)
|
||
except ValueError:
|
||
skipped += 1
|
||
continue
|
||
self.sessions[sid] = sess
|
||
new_sessions.append(sess)
|
||
added += 1
|
||
for sess in new_sessions:
|
||
self._persist_session(sess)
|
||
return {"sessionsAdded": added, "sessionsSkipped": skipped, "sessionsFileFound": True}
|
||
|
||
def _mykey_file(self) -> Path:
|
||
p = Path(self.ga_root) / "mykey.py"
|
||
if not p.exists():
|
||
tpl = Path(self.ga_root) / "mykey_template.py"
|
||
p.write_text(tpl.read_text(encoding="utf-8") if tpl.exists() else "", encoding="utf-8")
|
||
return p
|
||
|
||
@staticmethod
|
||
def _next_native_var(text: str, protocol: str) -> str:
|
||
# 协议必选(由前端下拉强制),不再用 apibase 兜底瞎猜
|
||
proto = str(protocol or "").strip().lower()
|
||
if proto == "claude":
|
||
prefix = "native_claude_config"
|
||
elif proto in ("oai", "openai"):
|
||
prefix = "native_oai_config"
|
||
else:
|
||
raise ValueError("protocol is required: choose 'oai' or 'claude'")
|
||
nums = [0]
|
||
if re.search(rf"^{prefix}\s*=", text, re.M):
|
||
nums.append(0)
|
||
nums.extend(int(m.group(1)) for m in re.finditer(rf"^{prefix}(\d+)\s*=", text, re.M))
|
||
n = max(nums) + 1
|
||
return prefix if n == 1 and not re.search(rf"^{prefix}\s*=", text, re.M) else f"{prefix}{n}"
|
||
|
||
@staticmethod
|
||
def _format_py_dict(d: dict) -> str:
|
||
lines = [f" '{k}': {json.dumps(v, ensure_ascii=False)}," if isinstance(v, str) else f" '{k}': {v}," for k, v in d.items()]
|
||
return "{\n" + "\n".join(lines) + "\n}"
|
||
|
||
def _profile_keys(self) -> List[str]:
|
||
self.ensure_ga_import_path()
|
||
from llmcore import reload_mykeys
|
||
return [k for k in reload_mykeys()[0] if any(x in k for x in ("api", "config", "cookie"))]
|
||
|
||
def _profile_at(self, profile_id: int) -> tuple[str, dict]:
|
||
keys = self._profile_keys()
|
||
if profile_id < 0 or profile_id >= len(keys):
|
||
raise ValueError("profile not found")
|
||
var = keys[profile_id]
|
||
if "mixin" in var:
|
||
raise ValueError("mixin profiles not supported here")
|
||
from llmcore import reload_mykeys
|
||
cfg = reload_mykeys()[0].get(var)
|
||
if not isinstance(cfg, dict):
|
||
raise ValueError("profile not editable")
|
||
return var, dict(cfg)
|
||
|
||
@staticmethod
|
||
def _find_var_block_span(text: str, var_name: str) -> Optional[tuple[int, int]]:
|
||
m = re.search(rf"^{re.escape(var_name)}\s*=\s*\{{", text, re.M)
|
||
if not m:
|
||
return None
|
||
start, i, depth = m.start(), m.end() - 1, 0
|
||
while i < len(text):
|
||
if text[i] == "{":
|
||
depth += 1
|
||
elif text[i] == "}":
|
||
depth -= 1
|
||
if depth == 0:
|
||
end = i + 1
|
||
while end < len(text) and text[end] in "\r\n":
|
||
end += 1
|
||
return start, end
|
||
i += 1
|
||
return None
|
||
|
||
def _patch_var_block(self, text: str, var: str, cfg: Optional[dict] = None) -> str:
|
||
if not (span := self._find_var_block_span(text, var)):
|
||
raise ValueError(f"config block not found: {var}")
|
||
s, e = span
|
||
if cfg is None:
|
||
return text[:s].rstrip() + "\n" + text[e:].lstrip("\n")
|
||
return text[:s] + f"{var} = {self._format_py_dict(cfg)}\n" + text[e:]
|
||
|
||
def _build_cfg(self, data: dict, existing: Optional[dict] = None, *, require_key: bool = True) -> dict:
|
||
apibase, model = str(data.get("apibase") or "").strip(), str(data.get("model") or "").strip()
|
||
if not apibase or not model:
|
||
raise ValueError("apibase and model are required")
|
||
apikey = str(data.get("apikey") or "").strip() or str((existing or {}).get("apikey") or "").strip()
|
||
if require_key and not apikey:
|
||
raise ValueError("apikey is required")
|
||
# 从 existing 起步：保留表单未覆盖的高级字段（proxy / temperature / api_mode /
|
||
# reasoning_effort / fake_cc_system_prompt / thinking_type …），避免 GUI 编辑时丢失
|
||
cfg: Dict[str, Any] = dict(existing or {})
|
||
cfg.update({"apikey": apikey, "apibase": apibase, "model": model})
|
||
if "name" in data:
|
||
name = str(data.get("name") or "").strip()
|
||
if name:
|
||
cfg["name"] = name
|
||
else:
|
||
cfg.pop("name", None)
|
||
for k in ("max_retries", "connect_timeout", "read_timeout"):
|
||
if data.get(k) is not None and str(data.get(k)).strip() != "":
|
||
cfg[k] = int(data[k])
|
||
# 流式开关：默认 True 不写（保持 mykey 干净），仅显式非流式才落 'stream': False
|
||
if "stream" in data:
|
||
s = data["stream"]
|
||
stream = s if isinstance(s, bool) else str(s).strip().lower() not in ("false", "0", "no", "off")
|
||
if stream:
|
||
cfg.pop("stream", None)
|
||
else:
|
||
cfg["stream"] = False
|
||
return cfg
|
||
|
||
def _save_mykey_text(self, text: str) -> list:
|
||
self._mykey_file().write_text(text, encoding="utf-8")
|
||
with self.lock:
|
||
for sess in self.sessions.values():
|
||
if sess.agent is not None:
|
||
try:
|
||
sess.agent.list_llms()
|
||
emit_session_state(sess, sess.status)
|
||
except Exception as e:
|
||
bridge_print(f"[model reload] {sess.id}: {e}")
|
||
return self.list_model_profiles()
|
||
|
||
def add_model_profile(self, data: dict) -> dict:
|
||
cfg = self._build_cfg(data)
|
||
text = self._mykey_file().read_text(encoding="utf-8")
|
||
var = self._next_native_var(text, data.get("protocol", ""))
|
||
profiles = self._save_mykey_text(text.rstrip() + f"\n{var} = {self._format_py_dict(cfg)}\n")
|
||
return {"varName": var, "profileId": profiles[-1]["id"] if profiles else 0, "profiles": profiles}
|
||
|
||
def get_model_profile(self, profile_id: int) -> dict:
|
||
var, cfg = self._profile_at(profile_id)
|
||
ks = ("model", "apibase", "apikey", "name", "max_retries", "connect_timeout", "read_timeout")
|
||
out = {"id": profile_id, "varName": var, **{k: cfg.get(k, d) for k, d in zip(ks, ("", "", "", "", 5, 15, 300))}}
|
||
out["stream"] = cfg.get("stream", True)
|
||
return out
|
||
|
||
def update_model_profile(self, profile_id: int, data: dict) -> dict:
|
||
var, existing = self._profile_at(profile_id)
|
||
text = self._mykey_file().read_text(encoding="utf-8")
|
||
|
||
# Sync mixin llm_nos when name changes to avoid orphan references
|
||
old_name = str(existing.get("name") or existing.get("model") or "").strip()
|
||
new_cfg = self._build_cfg(data, existing, require_key=False)
|
||
new_name = str(new_cfg.get("name") or new_cfg.get("model") or "").strip()
|
||
|
||
if old_name != new_name:
|
||
keys, mk = self._mykey_vars()
|
||
mvar, mcfg = self._mixin_entry(keys, mk)
|
||
if mcfg and mvar is not None and old_name in [str(m) for m in (mcfg.get("llm_nos") or [])]:
|
||
# Replace old name with new name in mixin's llm_nos
|
||
updated_nos = [new_name if str(m) == old_name else str(m) for m in (mcfg.get("llm_nos") or [])]
|
||
mcfg = {**mcfg, "llm_nos": updated_nos}
|
||
text = self._patch_var_block(text, mvar, mcfg)
|
||
|
||
profiles = self._save_mykey_text(self._patch_var_block(text, var, new_cfg))
|
||
return {"varName": var, "profileId": profile_id, "profiles": profiles}
|
||
|
||
def delete_model_profile(self, profile_id: int) -> dict:
|
||
if len(self._profile_keys()) <= 1:
|
||
raise ValueError("cannot delete the last profile")
|
||
var, cfg = self._profile_at(profile_id)
|
||
text = self._patch_var_block(self._mykey_file().read_text(encoding="utf-8"), var).rstrip() + "\n"
|
||
# 顺手把它从聚合渠道里摘掉，避免 llm_nos 残留指向已删除的模型（会让 Mixin 构建失败）
|
||
name = str(cfg.get("name") or cfg.get("model") or "").strip()
|
||
keys, mk = self._mykey_vars()
|
||
mvar, mcfg = self._mixin_entry(keys, mk)
|
||
if mcfg and mvar is not None and name in [str(m) for m in (mcfg.get("llm_nos") or [])]:
|
||
mcfg = {**mcfg, "llm_nos": [str(m) for m in (mcfg.get("llm_nos") or []) if str(m) != name]}
|
||
if self._find_var_block_span(text, mvar):
|
||
text = self._patch_var_block(text, mvar, mcfg)
|
||
profiles = self._save_mykey_text(text)
|
||
return {"profileId": profile_id, "profiles": profiles}
|
||
|
||
def ensure_ga_import_path(self) -> Path:
|
||
root = Path(self.ga_root).resolve()
|
||
if str(root) not in sys.path:
|
||
sys.path.insert(0, str(root))
|
||
return root
|
||
|
||
def make_agent(self, sess: Session):
|
||
root = self.ensure_ga_import_path()
|
||
try: import cost_tracker; cost_tracker.install()
|
||
except Exception: pass
|
||
old_cwd = os.getcwd()
|
||
try:
|
||
os.chdir(sess.cwd or str(root))
|
||
agentmain = importlib.import_module("agentmain")
|
||
GA = getattr(agentmain, "GenericAgent")
|
||
agent = GA()
|
||
no = sess.llm_no if sess.llm_no is not None else _global_default_llm_no()
|
||
if no is not None:
|
||
agent.next_llm(int(no))
|
||
agent.inc_out = True
|
||
agent.verbose = True
|
||
threading.Thread(target=agent.run, daemon=True, name=f"GA-{sess.id}").start()
|
||
return agent
|
||
finally:
|
||
with contextlib.suppress(Exception):
|
||
os.chdir(old_cwd)
|
||
|
||
@staticmethod
|
||
def _base_display_name(var: str, cfg: Optional[dict]) -> str:
|
||
c = cfg or {}
|
||
return str(c.get("name") or c.get("model") or var)
|
||
|
||
def _mykey_vars(self):
|
||
"""(keys, mk)：mykey 里的模型变量名（按定义顺序，与 agentmain.llmclients 索引
|
||
一一对齐）和原始 dict。过滤规则与 _profile_keys / load_llm_sessions 完全一致，
|
||
因此 id == llmclients 下标，前端选中 llmNo 能正确激活对应 client。"""
|
||
self._mykey_file() # 确保 mykey.py 存在（首次从模板生成空配置），否则全新安装时
|
||
# reload_mykeys 找不到 mykey 会返回空，空聚合渠道就不显示了
|
||
self.ensure_ga_import_path()
|
||
from llmcore import reload_mykeys
|
||
mk = reload_mykeys()[0]
|
||
keys = [k for k in mk if any(x in k for x in ("api", "config", "cookie"))]
|
||
return keys, mk
|
||
|
||
def _mixin_entry(self, keys, mk):
|
||
"""返回 (mixin_var, mixin_cfg_dict) 或 (None, None)。单一主聚合渠道，只取第一个。"""
|
||
for k in keys:
|
||
if "mixin" in k and isinstance(mk.get(k), dict):
|
||
return k, dict(mk[k])
|
||
return None, None
|
||
|
||
def list_model_profiles(self):
|
||
"""直接读 mykey.py 结构（不依赖能否成功构建出 client），这样空聚合渠道、
|
||
未填 key 的模型也能如实展示。聚合渠道(kind=mixin)带 members；基本模型
|
||
(kind=native)带 inMixin/group。"""
|
||
try:
|
||
keys, mk = self._mykey_vars()
|
||
except Exception as e:
|
||
bridge_print(f"get model profiles failed: {e}")
|
||
return []
|
||
# A profile can be referenced by any mixin channel, not only the first one.
|
||
# Keep each mixin row's own members for display, but mark native profiles as
|
||
# inMixin when they appear in any mixin.
|
||
all_mixin_members: set[str] = set()
|
||
for k in keys:
|
||
cfg = mk.get(k) if isinstance(mk.get(k), dict) else {}
|
||
if "mixin" in k:
|
||
all_mixin_members.update(str(m) for m in (cfg.get("llm_nos") or []))
|
||
active = self.config.get("llmNo", 0)
|
||
out = []
|
||
for i, k in enumerate(keys):
|
||
cfg = mk.get(k) if isinstance(mk.get(k), dict) else {}
|
||
if "mixin" in k:
|
||
members = [str(m) for m in (cfg.get("llm_nos") or [])]
|
||
out.append({"id": i, "varName": k, "kind": "mixin", "name": str(cfg.get("name") or k),
|
||
"members": members, "active": i == active})
|
||
else:
|
||
name = self._base_display_name(k, cfg)
|
||
out.append({"id": i, "varName": k, "kind": "native", "name": name,
|
||
"model": cfg.get("model", ""),
|
||
"group": "native" if "native" in k else "std",
|
||
"inMixin": name in all_mixin_members, "active": i == active})
|
||
return out
|
||
|
||
def add_to_mixin(self, profile_id: int) -> dict:
|
||
"""把一个基本模型加入主聚合渠道：把它的 name 追加进 mixin_config['llm_nos']。
|
||
坑1：校验 Native 一致性（聚合内必须全 Native 或全非 Native）。
|
||
坑2：加入前若该模型没有显式 name，先把 name 写进它的配置块（保证引用稳定）。"""
|
||
var, cfg = self._profile_at(profile_id) # 对 mixin 会抛错（只接受 native）
|
||
name = str(cfg.get("name") or cfg.get("model") or "").strip()
|
||
if not name:
|
||
raise ValueError("this model needs a name or model before joining the channel")
|
||
keys, mk = self._mykey_vars()
|
||
mvar, mcfg = self._mixin_entry(keys, mk)
|
||
new_is_native = "native" in var
|
||
name2var = {self._base_display_name(k, mk.get(k) if isinstance(mk.get(k), dict) else {}): k
|
||
for k in keys if "mixin" not in k}
|
||
existing = [str(m) for m in (mcfg.get("llm_nos") or [])] if mcfg else []
|
||
for m in existing:
|
||
mv = name2var.get(m)
|
||
if mv is not None and ("native" in mv) != new_is_native:
|
||
raise ValueError("aggregation channel requires all-Native or all-non-Native models")
|
||
text = self._mykey_file().read_text(encoding="utf-8")
|
||
if not cfg.get("name"):
|
||
text = self._patch_var_block(text, var, {**cfg, "name": name})
|
||
if mcfg is None:
|
||
mcfg, mvar, existing = {"llm_nos": [], "max_retries": 10, "base_delay": 0.5}, "mixin_config", []
|
||
if name not in existing:
|
||
existing.append(name)
|
||
mcfg = {**mcfg, "llm_nos": existing}
|
||
if self._find_var_block_span(text, mvar):
|
||
text = self._patch_var_block(text, mvar, mcfg)
|
||
else:
|
||
text = text.rstrip() + f"\n{mvar} = {self._format_py_dict(mcfg)}\n"
|
||
return {"profiles": self._save_mykey_text(text)}
|
||
|
||
def remove_from_mixin(self, profile_id: int) -> dict:
|
||
"""把一个基本模型移出主聚合渠道。"""
|
||
var, cfg = self._profile_at(profile_id)
|
||
name = str(cfg.get("name") or cfg.get("model") or "").strip()
|
||
keys, mk = self._mykey_vars()
|
||
mvar, mcfg = self._mixin_entry(keys, mk)
|
||
if not mcfg or mvar is None:
|
||
return {"profiles": self.list_model_profiles()}
|
||
members = [str(m) for m in (mcfg.get("llm_nos") or []) if str(m) != name]
|
||
mcfg = {**mcfg, "llm_nos": members}
|
||
text = self._patch_var_block(self._mykey_file().read_text(encoding="utf-8"), mvar, mcfg)
|
||
return {"profiles": self._save_mykey_text(text)}
|
||
|
||
def reorder_mixin(self, members: list) -> dict:
|
||
"""按前端拖拽后的顺序重写主渠道组 llm_nos。只接受当前成员的重排，不增删。"""
|
||
keys, mk = self._mykey_vars()
|
||
mvar, mcfg = self._mixin_entry(keys, mk)
|
||
if not mcfg or mvar is None:
|
||
raise ValueError("mixin channel not found")
|
||
old = [str(m) for m in (mcfg.get("llm_nos") or [])]
|
||
new = [str(m) for m in (members or [])]
|
||
if len(new) != len(old) or Counter(new) != Counter(old):
|
||
raise ValueError("reorder must contain the same channel members")
|
||
if new == old:
|
||
return {"profiles": self.list_model_profiles()}
|
||
mcfg = {**mcfg, "llm_nos": new}
|
||
text = self._patch_var_block(self._mykey_file().read_text(encoding="utf-8"), mvar, mcfg)
|
||
return {"profiles": self._save_mykey_text(text)}
|
||
|
||
@staticmethod
|
||
def _live_model(sess: Session) -> Optional[dict]:
|
||
"""该会话 agent 当前真正在用的模型(渠道组会随故障转移变化)。
|
||
agent 还没建(没跑过 turn)时返回静态绑定信息,前端据 llmNo 回显选择器。
|
||
已创建的 agent 是模型选择的唯一来源。"""
|
||
ag = getattr(sess, "agent", None)
|
||
if ag is None:
|
||
return {"current": None, "isMixin": False, "llmNo": sess.llm_no,
|
||
"runningLlmNo": sess.running_llm_no, "runningModel": sess.running_model}
|
||
try:
|
||
back = ag.llmclient.backend
|
||
live_no = sess.llm_no = ag.llm_no
|
||
if "Mixin" in type(back).__name__:
|
||
return {"current": back.current_name, "isMixin": True, "llmNo": live_no,
|
||
"runningLlmNo": sess.running_llm_no, "runningModel": sess.running_model}
|
||
return {"current": back.name, "isMixin": False, "llmNo": live_no,
|
||
"runningLlmNo": sess.running_llm_no, "runningModel": sess.running_model}
|
||
except Exception:
|
||
return {"current": None, "isMixin": False, "llmNo": sess.llm_no,
|
||
"runningLlmNo": sess.running_llm_no, "runningModel": sess.running_model}
|
||
|
||
def snapshot(self, sess: Session, include_messages: bool = True) -> dict:
|
||
out = {
|
||
"sessionId": sess.id,
|
||
"id": sess.id,
|
||
"title": sess.title,
|
||
"cwd": sess.cwd,
|
||
"status": sess.status,
|
||
"createdAt": sess.created_at,
|
||
"updatedAt": sess.updated_at,
|
||
"lastError": sess.last_error,
|
||
"msgSeq": sess.msg_seq,
|
||
"pinned": sess.pinned,
|
||
"untitled": sess.untitled,
|
||
"model": self._live_model(sess),
|
||
}
|
||
if include_messages:
|
||
out["messages"] = list(sess.messages)
|
||
out["partial"] = dict(sess.partial) if sess.partial else None
|
||
return out
|
||
|
||
def add_message(self, sess: Session, role: str, content: str, **extra) -> dict:
|
||
sess.msg_seq += 1
|
||
msg = {"id": sess.msg_seq, "role": role, "content": content, "ts": time.time()}
|
||
msg.update(extra)
|
||
sess.messages.append(msg)
|
||
sess.updated_at = time.time()
|
||
if role == "user" and content.strip() and sess.title == "New chat":
|
||
sess.title = content.strip().replace("\n", " ")[:40]
|
||
self._persist_session(sess)
|
||
return msg
|
||
|
||
def create_session(self, cwd: Optional[str] = None) -> Session:
|
||
sid = "sess-" + uuid.uuid4().hex[:12]
|
||
sess = Session(id=sid, cwd=str(cwd or self.ga_root), llm_no=_global_default_llm_no())
|
||
with self.mutation():
|
||
self.sessions[sid] = sess
|
||
self.active_session_id = sid
|
||
emit_session_state(sess, "created")
|
||
self._persist_session(sess)
|
||
return sess
|
||
|
||
def get_session(self, sid: str) -> Session:
|
||
with self.lock:
|
||
sess = self.sessions.get(sid)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
return sess
|
||
|
||
def delete_session(self, sid: str) -> dict:
|
||
with self.mutation():
|
||
sess = self.sessions.pop(sid, None)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
if self.active_session_id == sid:
|
||
self.active_session_id = next(iter(self.sessions), None)
|
||
# Retire the turn identity before aborting so a late runner cannot
|
||
# persist and resurrect the deleted session file.
|
||
sess.cancel_requested = True
|
||
sess.active_turn_id = ""
|
||
sess.status = "cancelled"
|
||
sess.partial = None
|
||
if sess.agent and hasattr(sess.agent, "abort"):
|
||
with contextlib.suppress(Exception):
|
||
sess.agent.abort()
|
||
if self._session_has_unfinished_work(sess):
|
||
self._retired_sessions[sid] = sess
|
||
emit_session_state(sess, "closed")
|
||
self._delete_session_file(sid)
|
||
_purge_session_uploads(sid)
|
||
return {"ok": True, "sessionId": sid}
|
||
|
||
def submit_prompt(self, sid: str, prompt: Any, images: Optional[list] = None, display: Optional[str] = None, files_meta: Optional[list] = None, image_metas: Optional[list] = None) -> dict:
|
||
prompt, image_ids = normalize_prompt(prompt, images)
|
||
# Build agent_prompt with file paths prepended (agent sees paths, UI sees clean text)
|
||
agent_prompt = prompt
|
||
if files_meta:
|
||
paths = [m["path"] for m in files_meta if m.get("path")]
|
||
if paths:
|
||
agent_prompt = " ".join(paths) + "\n" + prompt
|
||
with self.mutation():
|
||
sess = self.sessions.get(sid)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
if sess.status == "running":
|
||
raise web.HTTPConflict(text=json.dumps({"error": "session is already running"}, ensure_ascii=False), content_type="application/json")
|
||
extra = {}
|
||
if image_ids:
|
||
extra["image_ids"] = image_ids
|
||
if isinstance(display, str) and display.strip() and display != prompt:
|
||
extra["display"] = display
|
||
if files_meta:
|
||
extra["files"] = files_meta
|
||
if image_metas:
|
||
extra["images"] = image_metas
|
||
user_msg = self.add_message(sess, "user", prompt, **extra)
|
||
turn_id = uuid.uuid4().hex
|
||
sess.status = "running"
|
||
sess.cancel_requested = False
|
||
sess.active_turn_id = turn_id
|
||
sess.last_error = ""
|
||
_turn_start = time.time()
|
||
sess.partial = {"id": sess.msg_seq + 1, "role": "assistant", "content": "", "ts": _turn_start, "partial": True,
|
||
"turn_started": _turn_start, # stable turn-start clock (ts gets bumped on each stream chunk)
|
||
"curr_turn": 0, "turn_segs": []} # turn_segs[i]=第i轮全文(权威结构化,前端按轮渲染);content保留双轮兜底
|
||
image_paths = [m["path"] for m in (image_metas or []) if m.get("path")]
|
||
t = threading.Thread(
|
||
target=self.run_agent_turn,
|
||
args=(sess, agent_prompt, image_paths or None, turn_id),
|
||
daemon=True,
|
||
name=f"Turn-{sid}",
|
||
)
|
||
sess.thread = t
|
||
t.start()
|
||
seq = sess.msg_seq
|
||
emit_session_state(sess, "running")
|
||
return {"ok": True, "sessionId": sid, "accepted": True, "userMessageId": user_msg["id"], "seq": seq}
|
||
|
||
@staticmethod
|
||
def _patch_chat_for_images(client, image_paths):
|
||
"""Monkey-patch backend.ask to inject base64 image blocks on the first LLM call."""
|
||
import base64 as b64, mimetypes
|
||
try:
|
||
from llmcore import NativeToolClient
|
||
except ImportError:
|
||
return
|
||
if not isinstance(client, NativeToolClient):
|
||
return
|
||
backend = client.backend
|
||
original_ask = backend.ask
|
||
|
||
_VISION_MIMES = {'image/png', 'image/jpeg', 'image/gif', 'image/webp'}
|
||
|
||
def patched_ask(msg):
|
||
try:
|
||
del backend.ask
|
||
except AttributeError:
|
||
backend.ask = original_ask
|
||
if isinstance(msg, dict) and isinstance(msg.get("content"), list):
|
||
for p in image_paths:
|
||
try:
|
||
mime = mimetypes.guess_type(p)[0] or 'image/png'
|
||
if mime not in _VISION_MIMES:
|
||
# Unsupported image format (e.g. SVG) — inject as text path reference
|
||
msg["content"].append({"type": "text", "text": f"[attached file: {p}]"})
|
||
continue
|
||
with open(p, 'rb') as f:
|
||
raw = f.read()
|
||
data = b64.b64encode(raw).decode()
|
||
msg["content"].append({"type": "image", "source": {"type": "base64", "media_type": mime, "data": data}})
|
||
except Exception:
|
||
pass
|
||
resp = yield from original_ask(msg)
|
||
return resp
|
||
|
||
backend.ask = patched_ask
|
||
|
||
def run_agent_turn(
|
||
self,
|
||
sess: Session,
|
||
prompt: str,
|
||
images: Optional[list] = None,
|
||
turn_id: str = "",
|
||
):
|
||
def turn_state() -> tuple[bool, bool]:
|
||
with self.lock:
|
||
return sess.active_turn_id == turn_id, sess.cancel_requested
|
||
|
||
try:
|
||
is_current, _ = turn_state()
|
||
if not is_current:
|
||
return
|
||
if _consume_e2e_next_turn() == "empty":
|
||
with self.lock:
|
||
if sess.active_turn_id != turn_id:
|
||
return
|
||
sess.partial = None
|
||
self.add_message(sess, "assistant", empty_turn_fallback())
|
||
sess.running_llm_no = None
|
||
sess.running_model = None
|
||
sess.status = "idle"
|
||
sess.active_turn_id = ""
|
||
sess.last_error = ""
|
||
emit_session_state(sess, "idle")
|
||
return
|
||
if sess.agent is None:
|
||
sess.agent = self.make_agent(sess)
|
||
if sess.llm_history:
|
||
try:
|
||
sess.agent.llmclient.backend.history = sess.llm_history
|
||
except Exception:
|
||
pass
|
||
agent = sess.agent
|
||
with self.lock:
|
||
sess.running_llm_no = agent.llm_no
|
||
try:
|
||
running_backend = agent.llmclient.backend
|
||
sess.running_model = str(
|
||
getattr(running_backend, "current_name", None)
|
||
or getattr(running_backend, "name", None)
|
||
or getattr(running_backend, "model", None)
|
||
or ""
|
||
) or None
|
||
except Exception:
|
||
sess.running_model = None
|
||
if images:
|
||
self._patch_chat_for_images(agent.llmclient, images)
|
||
full = ""
|
||
done_outputs = None # done时agent给的全量轮文本(turn_resps.copy())
|
||
if hasattr(agent, "put_task"):
|
||
display_q = agent.put_task(prompt, images=images or [])
|
||
pieces = []
|
||
import queue as _queue
|
||
while True:
|
||
is_current, is_cancelled = turn_state()
|
||
if not is_current:
|
||
return
|
||
if is_cancelled:
|
||
break
|
||
try:
|
||
item = display_q.get(timeout=1.0)
|
||
except _queue.Empty:
|
||
continue
|
||
if isinstance(item, dict):
|
||
if item.get("next"):
|
||
text = str(item["next"])
|
||
pieces.append(text)
|
||
with self.lock:
|
||
if sess.partial is not None and sess.active_turn_id == turn_id:
|
||
sess.partial["content"] = "".join(pieces) if getattr(agent, "inc_out", False) else text
|
||
sess.partial["ts"] = time.time()
|
||
sess.updated_at = time.time()
|
||
# 轨道2: bridge 归一化为前端直接可渲染的 0 基 turn_segs；outputs=turn_resps[-2:]
|
||
_t = int(item.get("turn", 0) or 0)
|
||
_outs = item.get("outputs") or []
|
||
_idx = max(0, _t - 1)
|
||
sess.partial["curr_turn"] = _idx
|
||
_segs = sess.partial["turn_segs"]
|
||
while len(_segs) <= _idx:
|
||
_segs.append("")
|
||
if _outs:
|
||
_segs[_idx] = str(_outs[-1])
|
||
if len(_outs) >= 2 and _idx >= 1:
|
||
_segs[_idx - 1] = str(_outs[-2])
|
||
# Push partial to frontend via WS for real-time streaming
|
||
hub.emit({"type": "partial-update", "sessionId": sess.id,
|
||
"content": sess.partial["content"],
|
||
"turn_segs": list(_segs), "curr_turn": _idx})
|
||
if "done" in item:
|
||
full = strip_final_info_marker(item.get("done") or "")
|
||
done_outputs = normalize_final_turn_segs(full, item.get("outputs")) # done时=turn_resps.copy()全量轮
|
||
if done_outputs:
|
||
with self.lock:
|
||
if sess.partial is not None and sess.active_turn_id == turn_id:
|
||
sess.partial["content"] = full
|
||
sess.partial["ts"] = time.time()
|
||
sess.partial["updatedAt"] = sess.partial["ts"] if "updatedAt" in sess.partial else sess.partial.get("updatedAt")
|
||
sess.partial["curr_turn"] = max(0, len(done_outputs) - 1)
|
||
sess.partial["turn_segs"] = list(done_outputs)
|
||
sess.updated_at = time.time()
|
||
break
|
||
else:
|
||
pieces.append(str(item))
|
||
if not full and pieces:
|
||
full = pieces[-1] if not getattr(agent, "inc_out", False) else "".join(pieces)
|
||
else:
|
||
full = "GenericAgent object has no put_task method"
|
||
is_current, is_cancelled = turn_state()
|
||
if not is_current:
|
||
return
|
||
if not full:
|
||
full = empty_turn_fallback()
|
||
if is_cancelled:
|
||
with self.lock:
|
||
sess.partial = None
|
||
sess.running_llm_no = None
|
||
sess.running_model = None
|
||
if sess.active_turn_id == turn_id:
|
||
sess.active_turn_id = ""
|
||
# Ensure status stays cancelled (don't overwrite)
|
||
if sess.status != "cancelled":
|
||
sess.status = "cancelled"
|
||
sess.updated_at = time.time()
|
||
emit_session_state(sess, "cancelled")
|
||
return
|
||
with self.lock:
|
||
if sess.active_turn_id != turn_id:
|
||
return
|
||
turn_started = sess.partial.get("turn_started") if sess.partial else None
|
||
sess.partial = None
|
||
full = strip_final_info_marker(full)
|
||
import plan_state
|
||
plan_state.sync_plan_path_from_text(sess, full, sess.cwd or self.ga_root)
|
||
# 轨道2: 落库时带结构化全量轮(权威turn_segs),前端按轮渲染;content保留兜底
|
||
_final_segs = normalize_final_turn_segs(full, done_outputs)
|
||
msg_extra = {}
|
||
if _final_segs:
|
||
msg_extra["turn_segs"] = _final_segs
|
||
if turn_started:
|
||
msg_extra["executionMs"] = round((time.time() - turn_started) * 1000)
|
||
self.add_message(sess, "assistant", full, **msg_extra)
|
||
try: sess.llm_history = json.loads(json.dumps(agent.llmclient.backend.history, ensure_ascii=False, default=str))
|
||
except Exception: pass
|
||
sess.running_llm_no = None
|
||
sess.running_model = None
|
||
sess.status = "idle"
|
||
sess.active_turn_id = ""
|
||
sess.last_error = ""
|
||
emit_session_state(sess, "idle")
|
||
except Exception as e:
|
||
tb = traceback.format_exc()
|
||
with self.lock:
|
||
if sess.active_turn_id != turn_id:
|
||
return
|
||
sess.partial = None
|
||
sess.running_llm_no = None
|
||
sess.running_model = None
|
||
sess.status = "error"
|
||
sess.active_turn_id = ""
|
||
sess.last_error = str(e)
|
||
self.add_message(sess, "error", str(e))
|
||
bridge_print(f"[turn] error: {e}")
|
||
print(tb, file=sys.stderr)
|
||
emit_session_state(sess, "error")
|
||
|
||
def messages(self, sid: str, after: int = 0, limit: int = 200) -> dict:
|
||
with self.lock:
|
||
sess = self.sessions.get(sid)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
msgs = [m for m in sess.messages if int(m.get("id", 0)) > after]
|
||
if limit > 0:
|
||
msgs = msgs[-limit:]
|
||
import plan_state
|
||
return {
|
||
"sessionId": sid,
|
||
"status": sess.status,
|
||
"hasUnfinishedWork": self._session_has_unfinished_work(sess),
|
||
"messages": msgs,
|
||
"partial": dict(sess.partial) if sess.partial else None,
|
||
"plan": plan_state.desktop_plan_payload_from_session(sess, self.ga_root),
|
||
"msgSeq": sess.msg_seq,
|
||
"updatedAt": sess.updated_at,
|
||
"lastError": sess.last_error,
|
||
"model": self._live_model(sess),
|
||
}
|
||
|
||
def plan_snapshot(self, sid: str) -> dict:
|
||
with self.lock:
|
||
sess = self.sessions.get(sid)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
import plan_state
|
||
return {
|
||
"sessionId": sid,
|
||
"plan": plan_state.desktop_plan_payload_from_session(sess, self.ga_root),
|
||
}
|
||
|
||
def cancel(self, sid: str) -> dict:
|
||
with self.mutation():
|
||
sess = self.sessions.get(sid)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
sess.cancel_requested = True
|
||
if sess.agent and hasattr(sess.agent, "abort"):
|
||
with contextlib.suppress(Exception):
|
||
sess.agent.abort()
|
||
partial_text = ""
|
||
if sess.partial:
|
||
partial_text = (sess.partial.get("content") or "").strip()
|
||
if partial_text:
|
||
self.add_message(sess, "assistant", partial_text, stopped=True)
|
||
sess.status = "cancelled"
|
||
sess.active_turn_id = ""
|
||
sess.partial = None
|
||
sess.updated_at = time.time()
|
||
emit_session_state(sess, "cancelled")
|
||
return {"ok": True, "sessionId": sid}
|
||
|
||
def restore_context(self, sid: str) -> dict:
|
||
with self.lock:
|
||
sess = self.sessions.get(sid)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
if sess.agent is not None:
|
||
return {"ok": True, "sessionId": sid, "restored": False, "reason": "agent already alive"}
|
||
agent = self.make_agent(sess)
|
||
# 恢复 agent 时按会话绑定 seed 模型(未绑定则全局默认),保持显示/使用一致。
|
||
no = sess.llm_no if sess.llm_no is not None else _global_default_llm_no()
|
||
if no is not None and hasattr(agent, "next_llm"):
|
||
with contextlib.suppress(Exception):
|
||
agent.next_llm(int(no))
|
||
if sess.llm_history:
|
||
try:
|
||
agent.llmclient.backend.history = sess.llm_history
|
||
except Exception as e:
|
||
bridge_print(f"[bridge] restore llm_history failed: {e}")
|
||
else:
|
||
history = []
|
||
for m in sess.messages:
|
||
role = m.get("role")
|
||
content = m.get("content", "")
|
||
if role == "user":
|
||
history.append({"role": "user", "content": [{"type": "text", "text": content}]})
|
||
elif role == "assistant":
|
||
history.append({"role": "assistant", "content": [{"type": "text", "text": content}]})
|
||
if history:
|
||
try:
|
||
agent.llmclient.backend.history = history
|
||
except Exception as e:
|
||
bridge_print(f"[bridge] inject history failed: {e}")
|
||
with self.lock:
|
||
sess.agent = agent
|
||
sess.status = "idle"
|
||
return {"ok": True, "sessionId": sid, "restored": True, "messageCount": len(sess.llm_history or sess.messages)}
|
||
|
||
def set_session_model(self, sid: str, llm_no: int) -> dict:
|
||
"""Bind a model to a session.
|
||
|
||
Existing agents own selection; running turns ignore switch requests.
|
||
"""
|
||
with self.mutation():
|
||
sess = self.sessions.get(sid)
|
||
if not sess:
|
||
raise web.HTTPNotFound(text=json.dumps({"error": f"session not found: {sid}"}, ensure_ascii=False), content_type="application/json")
|
||
if sess.status != "running":
|
||
if sess.agent is None:
|
||
sess.llm_no = int(llm_no)
|
||
else:
|
||
sess.agent.next_llm(int(llm_no))
|
||
sess.llm_no = sess.agent.llm_no
|
||
sess.updated_at = time.time()
|
||
self._persist_session(sess)
|
||
return {"ok": True, "sessionId": sid, "llmNo": sess.llm_no, "model": self._live_model(sess)}
|
||
|
||
|
||
import base64
|
||
|
||
|
||
def normalize_prompt(prompt: Any, images: Optional[list] = None):
|
||
"""Flatten a prompt (str or content-part list) to plain text.
|
||
|
||
Image/file attachments are handled by the frontend, which inlines the
|
||
uploaded file path into the prompt text (see expandFilePlaceholders) and
|
||
sends path-only metadata via files/imageMetas — so no per-prompt image
|
||
persistence happens here. The `images` arg is accepted for backward compat
|
||
and ignored; the returned image-id list is always empty.
|
||
"""
|
||
if isinstance(prompt, list):
|
||
text_parts = []
|
||
for part in prompt:
|
||
if isinstance(part, str):
|
||
text_parts.append(part)
|
||
elif isinstance(part, dict) and part.get("type") in ("text", "input_text"):
|
||
text_parts.append(str(part.get("text") or part.get("content") or ""))
|
||
prompt = "\n".join([p for p in text_parts if p])
|
||
|
||
return str(prompt or ""), []
|
||
|
||
|
||
manager = AgentManager()
|
||
|
||
# Initialize per-call token ledger
|
||
try:
|
||
import cost_tracker
|
||
cost_tracker.init_ledger(manager.ga_root)
|
||
except Exception:
|
||
pass
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Transport layer: WS state push
|
||
# ---------------------------------------------------------------------------
|
||
|
||
class WsHub:
|
||
def __init__(self):
|
||
self.websockets: Set[web.WebSocketResponse] = set()
|
||
self.loop: Optional[asyncio.AbstractEventLoop] = None
|
||
|
||
def emit(self, obj: dict):
|
||
if self.loop and self.loop.is_running():
|
||
asyncio.run_coroutine_threadsafe(self._broadcast(obj), self.loop)
|
||
|
||
async def _broadcast(self, obj: dict):
|
||
data = json.dumps(obj, ensure_ascii=False, default=str)
|
||
dead = set()
|
||
for ws in list(self.websockets):
|
||
try:
|
||
await ws.send_str(data)
|
||
except Exception:
|
||
dead.add(ws)
|
||
self.websockets.difference_update(dead)
|
||
|
||
|
||
hub = WsHub()
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Service management (hub.pyw core + WS notify)
|
||
# ---------------------------------------------------------------------------
|
||
|
||
_SKIP = frozenset({"goal_mode.py", "chatapp_common.py", "tuiapp.py", "qtapp.py"})
|
||
BRIDGE_ID = "__bridge__"
|
||
|
||
_SERVICE_KEYS: Dict[str, tuple] = {
|
||
"frontends/qqapp.py": ("qq_app_id", "qq_app_secret"),
|
||
"frontends/dcapp.py": ("discord_bot_token",),
|
||
"frontends/dingtalkapp.py": ("dingtalk_client_id", "dingtalk_client_secret"),
|
||
"frontends/fsapp.py": ("fs_app_id", "fs_app_secret"),
|
||
"frontends/tgapp.py": ("tg_bot_token",),
|
||
"frontends/wecomapp.py": ("wecom_bot_id", "wecom_secret"),
|
||
}
|
||
|
||
|
||
def _load_mykeys(ga_root: Path) -> dict:
|
||
if not (ga_root / "mykey.py").exists():
|
||
return {}
|
||
root = str(ga_root.resolve())
|
||
if root not in sys.path:
|
||
sys.path.insert(0, root)
|
||
import mykey as mk
|
||
importlib.reload(mk)
|
||
return {k: v for k, v in vars(mk).items() if not k.startswith("_")}
|
||
|
||
|
||
def discover_im_services(ga_root: Path) -> List[dict]:
|
||
out: List[dict] = []
|
||
d = ga_root / "frontends"
|
||
if not d.is_dir():
|
||
return out
|
||
for f in sorted(os.listdir(d)):
|
||
if "app" not in f or not f.endswith(".py") or f in _SKIP or "stapp" in f or "tuiapp" in f:
|
||
continue
|
||
rel = f"frontends/{f}"
|
||
out.append({"id": rel, "cmd": [sys.executable, str(d / f)]})
|
||
return out
|
||
|
||
|
||
def discover_extra_services(ga_root: Path) -> List[dict]:
|
||
out: List[dict] = []
|
||
sched = ga_root / "reflect" / "scheduler.py"
|
||
if sched.is_file():
|
||
out.append({
|
||
"id": "reflect/scheduler.py",
|
||
"cmd": [sys.executable, "agentmain.py", "--reflect", "reflect/scheduler.py"],
|
||
})
|
||
# conductor 跟 scheduler 一样,bridge 启动时自动拉起。--no-browser 是关键:
|
||
# conductor.py 默认会用 webbrowser.open 在用户浏览器弹一个 8900 端口 UI,
|
||
# 桌面版自启时不需要这个独立 UI(用户从「指挥家」页直接访问)。
|
||
# The desktop conductor evolves with this bridge. Keep both package-owned and inject the
|
||
# external core through GA_ROOT instead of executing arbitrary UI glue from that checkout.
|
||
conductor = APP_DIR / "conductor.py"
|
||
if conductor.is_file():
|
||
conductor_port = _configured_conductor_port()
|
||
out.append({
|
||
"id": "frontends/conductor.py",
|
||
"cmd": [
|
||
sys.executable,
|
||
str(conductor),
|
||
"--no-browser",
|
||
"--port",
|
||
str(conductor_port),
|
||
],
|
||
"port": conductor_port,
|
||
})
|
||
return out
|
||
|
||
|
||
def _port_alive(port: Optional[int]) -> bool:
|
||
"""Check if something is listening on localhost:port."""
|
||
if not port:
|
||
return False
|
||
import socket
|
||
with socket.socket(socket.AF_INET, socket.SOCK_STREAM) as s:
|
||
s.settimeout(0.3)
|
||
return s.connect_ex(("127.0.0.1", port)) == 0
|
||
|
||
|
||
def _mem_mb(pid: Optional[int]) -> Optional[int]:
|
||
if not pid:
|
||
return None
|
||
if sys.platform == "win32":
|
||
import ctypes
|
||
from ctypes import wintypes
|
||
class PROCESS_MEMORY_COUNTERS(ctypes.Structure):
|
||
_fields_ = [
|
||
("cb", wintypes.DWORD), ("PageFaultCount", wintypes.DWORD),
|
||
("PeakWorkingSetSize", ctypes.c_size_t), ("WorkingSetSize", ctypes.c_size_t),
|
||
("QuotaPeakPagedPoolUsage", ctypes.c_size_t), ("QuotaPagedPoolUsage", ctypes.c_size_t),
|
||
("QuotaPeakNonPagedPoolUsage", ctypes.c_size_t), ("QuotaNonPagedPoolUsage", ctypes.c_size_t),
|
||
("PagefileUsage", ctypes.c_size_t), ("PeakPagefileUsage", ctypes.c_size_t),
|
||
]
|
||
counters = PROCESS_MEMORY_COUNTERS()
|
||
counters.cb = ctypes.sizeof(PROCESS_MEMORY_COUNTERS)
|
||
h = ctypes.windll.kernel32.OpenProcess(0x1000, False, pid)
|
||
if not h:
|
||
return None
|
||
ok = ctypes.windll.psapi.GetProcessMemoryInfo(h, ctypes.byref(counters), counters.cb)
|
||
ctypes.windll.kernel32.CloseHandle(h)
|
||
return round(counters.WorkingSetSize / 1024 / 1024) if ok else None
|
||
status = Path(f"/proc/{pid}/status")
|
||
if status.is_file():
|
||
for line in status.read_text(encoding="utf-8", errors="replace").splitlines():
|
||
if line.startswith("VmRSS:"):
|
||
return round(int(line.split()[1]) / 1024)
|
||
return None
|
||
|
||
|
||
def _cpu_pct(pid: Optional[int]) -> Optional[float]:
|
||
if not pid:
|
||
return None
|
||
try:
|
||
import psutil
|
||
return round(psutil.Process(pid).cpu_percent(0) or 0, 1)
|
||
except Exception:
|
||
return None
|
||
|
||
|
||
_ERROR_PATTERNS: list[tuple[re.Pattern, str, str]] = [
|
||
(re.compile(r"errno 48|address already in use", re.I), "transient", "err.portBusy"),
|
||
(re.compile(r"Exception in thread conductor", re.I), "fatal", "err.conductorCrash"),
|
||
(re.compile(r"ModuleNotFoundError|ImportError", re.I), "fatal", "err.missingModule"),
|
||
(re.compile(r"ConnectionRefusedError|Connection refused", re.I), "warning", "err.connRefused"),
|
||
(re.compile(r"TimeoutError|timed out", re.I), "warning", "err.timeout"),
|
||
]
|
||
|
||
|
||
def _classify_log_error(line: str) -> tuple[str, str] | None:
|
||
"""Classify a log line by severity. Returns (severity, i18n_key) or None."""
|
||
for pattern, severity, key in _ERROR_PATTERNS:
|
||
if pattern.search(line):
|
||
return (severity, key)
|
||
return None
|
||
|
||
|
||
class ServiceManager:
|
||
"""hub.pyw ServiceManager + HTTP/WS glue."""
|
||
|
||
def __init__(self, ga_root: str, emit_fn):
|
||
self.lock = threading.RLock()
|
||
self.ga_root = Path(ga_root)
|
||
self.procs: Dict[str, subprocess.Popen] = {}
|
||
self.buffers: Dict[str, deque] = {}
|
||
self._emit = emit_fn
|
||
im = discover_im_services(self.ga_root)
|
||
extra = discover_extra_services(self.ga_root)
|
||
self._im_catalog = {s["id"]: s for s in im}
|
||
self._catalog = {**self._im_catalog, **{s["id"]: s for s in extra}}
|
||
self._stopping: Set[str] = set()
|
||
|
||
def _is_configured(self, sid: str) -> bool:
|
||
keys = _SERVICE_KEYS.get(sid)
|
||
if not keys:
|
||
return True
|
||
mykeys = _load_mykeys(self.ga_root)
|
||
return all(str(mykeys.get(k) or "").strip() for k in keys)
|
||
|
||
def _scan_errors(self, sid: str, n: int = 20) -> dict:
|
||
"""Scan last N buffer lines, classify by severity.
|
||
Returns {"fatal": (line, key) | None, "warning": (line, key) | None}.
|
||
Transient errors (e.g. port-busy retry) are discarded.
|
||
"""
|
||
buf = self.buffers.get(sid)
|
||
if not buf:
|
||
return {"fatal": None, "warning": None}
|
||
lines = [ln.strip() for ln in list(buf)[-n:] if ln.strip()]
|
||
if not lines:
|
||
return {"fatal": None, "warning": None}
|
||
result: dict = {"fatal": None, "warning": None}
|
||
for line in lines:
|
||
classified = _classify_log_error(line)
|
||
if classified is None:
|
||
continue
|
||
severity, key = classified
|
||
if severity == "fatal":
|
||
result["fatal"] = (line[:300], key)
|
||
elif severity == "warning":
|
||
result["warning"] = (line[:300], key)
|
||
return result
|
||
|
||
def _log_tail(self, sid: str, n: int = 3) -> str:
|
||
"""Legacy: return last fatal error line for lastError field."""
|
||
scan = self._scan_errors(sid, n=20)
|
||
if scan["fatal"]:
|
||
return scan["fatal"][0]
|
||
buf = self.buffers.get(sid)
|
||
if not buf:
|
||
return ""
|
||
lines = [ln.strip() for ln in list(buf)[-n:] if ln.strip()]
|
||
return lines[-1][:300] if lines else ""
|
||
|
||
def _state(self, sid: str, *, err: str = "") -> dict:
|
||
proc = self.procs.get(sid)
|
||
owned = proc is not None and proc.poll() is None
|
||
running = owned
|
||
status = "running" if owned else "offline"
|
||
last_error = err
|
||
error_key = ""
|
||
last_warning = ""
|
||
warning_key = ""
|
||
scan = self._scan_errors(sid)
|
||
catalog_port = self._catalog.get(sid, {}).get("port")
|
||
port_alive = _port_alive(catalog_port)
|
||
external = bool(catalog_port and port_alive and not owned)
|
||
if proc is not None and not owned:
|
||
if sid in self._stopping:
|
||
status, last_error = "offline", ""
|
||
elif external:
|
||
status = "error"
|
||
last_error = err or f"port {catalog_port} is occupied by an untracked process"
|
||
error_key = "err.portBusy"
|
||
else:
|
||
status = "error"
|
||
if scan["fatal"]:
|
||
last_error = scan["fatal"][0]
|
||
error_key = scan["fatal"][1]
|
||
elif err:
|
||
last_error = err
|
||
else:
|
||
last_error = f"exit code {proc.returncode}"
|
||
elif external:
|
||
status = "error"
|
||
last_error = err or f"port {catalog_port} is occupied by an untracked process"
|
||
error_key = "err.portBusy"
|
||
elif owned:
|
||
if scan["warning"]:
|
||
last_warning = scan["warning"][0]
|
||
warning_key = scan["warning"][1]
|
||
elif err:
|
||
status, running = "error", False
|
||
last_error = err
|
||
if scan["fatal"]:
|
||
error_key = scan["fatal"][1]
|
||
return {
|
||
"id": sid,
|
||
"status": status,
|
||
"running": running,
|
||
"owned": owned,
|
||
"external": external,
|
||
"portConflict": external,
|
||
"port": catalog_port,
|
||
"pid": proc.pid if owned and proc is not None else None,
|
||
"lastError": last_error,
|
||
"errorKey": error_key,
|
||
"lastWarning": last_warning,
|
||
"warningKey": warning_key,
|
||
}
|
||
|
||
def list_state(self) -> List[dict]:
|
||
with self.lock:
|
||
return [self._state(sid) for sid in sorted(self._im_catalog)]
|
||
|
||
def running_managed_ids(self) -> List[str]:
|
||
with self.lock:
|
||
return [
|
||
sid for sid in sorted(self._catalog)
|
||
if self._state(sid).get("owned") is True
|
||
]
|
||
|
||
def _bridge_state(self) -> dict:
|
||
pid = os.getpid()
|
||
port = int(os.environ.get("BRIDGE_PORT", "14168"))
|
||
return {
|
||
"id": BRIDGE_ID,
|
||
"name": f"bridge (:{port})",
|
||
"status": "running",
|
||
"running": True,
|
||
"pid": pid,
|
||
"memMb": _mem_mb(pid),
|
||
"cpuPct": _cpu_pct(pid),
|
||
"managed": False,
|
||
"lastError": "",
|
||
}
|
||
|
||
def _managed_state(self, sid: str, *, err: str = "") -> dict:
|
||
item = self._state(sid, err=err)
|
||
item["name"] = sid
|
||
item["memMb"] = _mem_mb(item.get("pid"))
|
||
item["cpuPct"] = _cpu_pct(item.get("pid"))
|
||
item["managed"] = True
|
||
return item
|
||
|
||
def list_panel_state(self) -> List[dict]:
|
||
with self.lock:
|
||
out = [self._bridge_state()]
|
||
for sid in sorted(self._catalog, key=lambda s: (s in self._im_catalog, s)):
|
||
out.append(self._managed_state(sid))
|
||
return out
|
||
|
||
def _notify(self, sid: str, *, err: str = "") -> None:
|
||
self._emit({"type": "service.changed", "service": self._managed_state(sid, err=err)})
|
||
|
||
def _wait_started(self, proc: subprocess.Popen, timeout: float = 2.0) -> None:
|
||
deadline = time.time() + timeout
|
||
while time.time() < deadline:
|
||
if proc.poll() is not None:
|
||
return
|
||
time.sleep(0.1)
|
||
|
||
def _reader(self, sid: str, proc: subprocess.Popen) -> None:
|
||
assert proc.stdout is not None
|
||
for line in proc.stdout:
|
||
buf = self.buffers.get(sid)
|
||
if buf is not None:
|
||
buf.append(line)
|
||
self._notify(sid)
|
||
|
||
def start_service(self, sid: str) -> dict:
|
||
# Lock order is always AgentManager -> ServiceManager. This makes a
|
||
# service start atomic with maintenance-gate admission.
|
||
with manager.mutation(), self.lock:
|
||
svc = self._catalog.get(sid)
|
||
if not svc:
|
||
raise KeyError(sid)
|
||
proc = self.procs.get(sid)
|
||
if proc is not None and proc.poll() is None:
|
||
return {"ok": True, "service": self._managed_state(sid)}
|
||
catalog_port = svc.get("port")
|
||
if catalog_port and _port_alive(catalog_port):
|
||
err = f"port {catalog_port} is occupied by an untracked process"
|
||
item = self._managed_state(sid, err=err)
|
||
self._notify(sid, err=err)
|
||
return {"ok": False, "error": "port_conflict", "service": item}
|
||
if not self._is_configured(sid):
|
||
keys = ", ".join(_SERVICE_KEYS.get(sid, ()))
|
||
err = f"not configured in mykey.py ({keys})"
|
||
self._notify(sid, err=err)
|
||
return {
|
||
"ok": False,
|
||
"error": "not_configured",
|
||
"service": self._managed_state(sid, err=err),
|
||
}
|
||
self.buffers[sid] = deque(maxlen=500)
|
||
env = {
|
||
**os.environ,
|
||
"PYTHONUNBUFFERED": "1",
|
||
"PYTHONIOENCODING": "utf-8",
|
||
"GA_ROOT": str(self.ga_root),
|
||
}
|
||
kw: Dict[str, Any] = dict(
|
||
cwd=str(self.ga_root), stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
|
||
text=True, encoding="utf-8", errors="replace", bufsize=1, env=env,
|
||
)
|
||
if sys.platform == "win32":
|
||
kw["creationflags"] = subprocess.CREATE_NO_WINDOW
|
||
proc = subprocess.Popen(svc["cmd"], **kw)
|
||
self.procs[sid] = proc
|
||
threading.Thread(target=self._reader, args=(sid, proc), daemon=True).start()
|
||
self._wait_started(proc)
|
||
item = self._managed_state(sid)
|
||
self._notify(sid)
|
||
if item["running"] is not True:
|
||
return {"ok": False, "error": item["lastError"] or "start_failed", "service": item}
|
||
return {"ok": True, "service": item}
|
||
|
||
def autostart_extras(self) -> None:
|
||
"""Auto-start non-IM services on bridge boot. Currently:
|
||
- reflect/scheduler.py (drives L4 archive cron every 12h).
|
||
IM services stay manual (need explicit mykey.py config + user opt-in)."""
|
||
for sid in sorted(set(self._catalog) - set(self._im_catalog)):
|
||
try:
|
||
res = self.start_service(sid)
|
||
tag = "ok" if res.get("ok") else f"fail: {res.get('error')}"
|
||
except Exception as e:
|
||
tag = f"exception {type(e).__name__}: {e}"
|
||
bridge_print(f"[autostart] {sid}: {tag}")
|
||
|
||
def stop_all_extras(self) -> None:
|
||
for sid in sorted(set(self._catalog) - set(self._im_catalog)):
|
||
with contextlib.suppress(Exception):
|
||
self.stop_service(sid)
|
||
|
||
def _extra_is_broken(self, sid: str) -> bool:
|
||
"""判断一个 extra 是否「已经坏掉、需要重启才能恢复」:
|
||
- 进程异常退出(非用户主动停)→ 坏(覆盖 scheduler 那种进程级崩溃);
|
||
- 进程还活着,但捕获日志里出现 `Exception in thread conductor-agent`
|
||
→ 内部 agent 线程已崩死,uvicorn 还在跑但再也处理不了任务 → 坏。
|
||
健康运行中的进程返回 False:它会在下个任务靠自身 mtime 热重载读到新 mykey,
|
||
不该被打断。每次 start_service 都会换新缓冲,故缓冲里的崩溃签名只反映当前进程。"""
|
||
proc = self.procs.get(sid)
|
||
if proc is None:
|
||
return False # 没起过 / 用户主动停掉 → 不复活
|
||
if proc.poll() is not None:
|
||
return sid not in self._stopping # 意外退出 = 坏
|
||
buf = self.buffers.get(sid)
|
||
return bool(buf) and any("Exception in thread conductor-agent" in ln for ln in buf)
|
||
|
||
def restart_broken_extras(self) -> None:
|
||
"""mykey 被整体重写(导入密钥/编辑渠道配置)后,只重启「已经坏掉」的
|
||
conductor/scheduler。健康运行中的进程不动——它们会在下个任务靠自身 mtime
|
||
热重载新 key,强行重启反而会打断正在跑的任务。"""
|
||
for sid in sorted(set(self._catalog) - set(self._im_catalog)):
|
||
if not self._extra_is_broken(sid):
|
||
continue
|
||
with contextlib.suppress(Exception):
|
||
self.stop_service(sid)
|
||
try:
|
||
res = self.start_service(sid)
|
||
tag = "ok" if res.get("ok") else f"fail: {res.get('error')}"
|
||
except Exception as e:
|
||
tag = f"exception {type(e).__name__}: {e}"
|
||
bridge_print(f"[restart-broken] {sid}: {tag}")
|
||
|
||
def stop_service(self, sid: str) -> dict:
|
||
with self.lock:
|
||
if sid not in self._catalog:
|
||
raise KeyError(sid)
|
||
self._stopping.add(sid)
|
||
proc = self.procs.get(sid)
|
||
stop_error = ""
|
||
if proc and proc.poll() is None:
|
||
try:
|
||
proc.terminate()
|
||
try:
|
||
proc.wait(timeout=5.0)
|
||
except subprocess.TimeoutExpired:
|
||
proc.kill()
|
||
proc.wait(timeout=2.0)
|
||
except Exception as exc:
|
||
stop_error = f"failed to stop managed process: {type(exc).__name__}: {exc}"
|
||
if proc is None or proc.poll() is not None:
|
||
self.procs.pop(sid, None)
|
||
self._stopping.discard(sid)
|
||
item = self._managed_state(sid, err=stop_error)
|
||
self._notify(sid, err=stop_error)
|
||
if item.get("external") is True:
|
||
return {"ok": False, "error": "not_owned", "service": item}
|
||
return {
|
||
"ok": not stop_error,
|
||
"error": "stop_failed" if stop_error else "",
|
||
"service": item,
|
||
}
|
||
|
||
def read_logs(self, sid: str, tail: int = 200) -> dict:
|
||
if sid == BRIDGE_ID:
|
||
tail = max(1, min(int(tail or 200), 2000))
|
||
lines = list(_bridge_log)[-tail:]
|
||
if not lines:
|
||
lines = [f"GenericAgent bridge pid={os.getpid()}"]
|
||
return {"ok": True, "lines": lines}
|
||
if sid not in self._catalog:
|
||
raise KeyError(sid)
|
||
tail = max(1, min(int(tail or 200), 2000))
|
||
buf = self.buffers.get(sid)
|
||
lines = [ln.rstrip("\n") for ln in list(buf or [])[-tail:]]
|
||
return {"ok": True, "lines": lines}
|
||
|
||
|
||
services = ServiceManager(str(DEFAULT_GA_ROOT), hub.emit)
|
||
|
||
|
||
def _bridge_shutdown_services() -> None:
|
||
with contextlib.suppress(Exception):
|
||
services.stop_all_extras()
|
||
|
||
|
||
atexit.register(_bridge_shutdown_services)
|
||
|
||
|
||
def emit_session_state(sess: Session, state_name: str):
|
||
if state_name != "created":
|
||
title = (sess.title or sess.id)[:20]
|
||
bridge_print(f"[session] {title}: {state_name}")
|
||
hub.emit({
|
||
"type": "session-state",
|
||
"sessionId": sess.id,
|
||
"state": state_name,
|
||
"status": sess.status,
|
||
"seq": sess.msg_seq,
|
||
"updatedAt": sess.updated_at,
|
||
"title": sess.title,
|
||
})
|
||
|
||
|
||
async def ws_handler(request):
|
||
origin_error = _request_origin_error(request)
|
||
if origin_error:
|
||
return web.json_response(
|
||
{"ok": False, "error": origin_error, "code": "origin_forbidden"},
|
||
status=403,
|
||
)
|
||
ws = web.WebSocketResponse(heartbeat=30)
|
||
await ws.prepare(request)
|
||
hub.websockets.add(ws)
|
||
await ws.send_str(json.dumps({
|
||
"type": "bridge-ready",
|
||
"gaRoot": manager.ga_root,
|
||
"mykeyPath": manager.mykey_path,
|
||
"http": True,
|
||
"wsEventsOnly": True,
|
||
}, ensure_ascii=False))
|
||
await ws.send_str(json.dumps({
|
||
"type": "services.snapshot",
|
||
"services": services.list_panel_state(),
|
||
}, ensure_ascii=False, default=str))
|
||
async for msg in ws:
|
||
if msg.type == WSMsgType.TEXT:
|
||
# WS is intentionally not a data/command channel anymore.
|
||
with contextlib.suppress(Exception):
|
||
data = json.loads(msg.data)
|
||
if data.get("action") == "ping":
|
||
await ws.send_str(json.dumps({"type": "pong", "ts": time.time()}, ensure_ascii=False))
|
||
hub.websockets.discard(ws)
|
||
return ws
|
||
|
||
|
||
# ---------------------------------------------------------------------------
|
||
# Transport layer: HTTP command/data API
|
||
# ---------------------------------------------------------------------------
|
||
|
||
def _valid_port(value: str, default: int) -> int:
|
||
if not re.fullmatch(r"[0-9]{1,5}", value or ""):
|
||
return default
|
||
port = int(value)
|
||
return port if 1 <= port <= 65535 else default
|
||
|
||
|
||
def _allowed_request_origins() -> Set[str]:
|
||
bridge_port = _valid_port(os.environ.get("BRIDGE_PORT", "14168"), 14168)
|
||
origins = {
|
||
"tauri://localhost",
|
||
"http://tauri.localhost",
|
||
"http://localhost:5173",
|
||
f"http://127.0.0.1:{bridge_port}",
|
||
f"http://localhost:{bridge_port}",
|
||
f"http://[::1]:{bridge_port}",
|
||
}
|
||
if os.environ.get("GA_E2E") == "1":
|
||
vite_port = os.environ.get("VITE_PORT", "")
|
||
if re.fullmatch(r"[0-9]{1,5}", vite_port or ""):
|
||
parsed_port = int(vite_port)
|
||
if 1 <= parsed_port <= 65535:
|
||
origins.add(f"http://127.0.0.1:{parsed_port}")
|
||
return origins
|
||
|
||
|
||
# Read-only GET endpoints that are meant to be loaded as browser subresources
|
||
# (<img>/<video>/<audio>/font). Cross-origin resource loads never carry an
|
||
# Origin header but do set Sec-Fetch-Site: cross-site, so the generic CSRF guard
|
||
# below would 403 them. These endpoints have no side effects and are confined to
|
||
# a whitelisted directory, so a cross-site *resource* GET/HEAD is safe to allow.
|
||
_RESOURCE_GET_PATHS = frozenset({"/upload/raw"})
|
||
_RESOURCE_FETCH_DESTS = frozenset({"image", "video", "audio", "font"})
|
||
|
||
|
||
def _is_safe_cross_site_resource_get(request) -> bool:
|
||
"""True for a no-side-effect subresource GET the desktop webview issues for
|
||
a message-bubble thumbnail (bridge origin differs from the tauri app origin,
|
||
so the <img> request is cross-site and carries no Origin header)."""
|
||
if request.method not in ("GET", "HEAD"):
|
||
return False
|
||
if request.path not in _RESOURCE_GET_PATHS:
|
||
return False
|
||
dest = request.headers.get("Sec-Fetch-Dest", "").strip().lower()
|
||
# Only genuine subresource loads — never a cross-site top-level navigation
|
||
# (Sec-Fetch-Dest: document) — are exempted.
|
||
return dest in _RESOURCE_FETCH_DESTS
|
||
|
||
|
||
def _request_origin_error(request) -> Optional[str]:
|
||
origin = request.headers.get("Origin")
|
||
if origin is not None:
|
||
if origin not in _allowed_request_origins():
|
||
return "request origin is not allowed"
|
||
return None
|
||
if request.headers.get("Sec-Fetch-Site", "").strip().lower() == "cross-site":
|
||
if _is_safe_cross_site_resource_get(request):
|
||
return None
|
||
return "cross-site request without an Origin header is not allowed"
|
||
return None
|
||
|
||
|
||
def _add_cors_response_headers(response: web.StreamResponse, origin: Optional[str]) -> None:
|
||
if origin is None:
|
||
return
|
||
response.headers["Access-Control-Allow-Origin"] = origin
|
||
vary = [part.strip() for part in response.headers.get("Vary", "").split(",") if part.strip()]
|
||
if not any(part.lower() == "origin" for part in vary):
|
||
vary.append("Origin")
|
||
response.headers["Vary"] = ", ".join(vary)
|
||
response.headers["Access-Control-Allow-Methods"] = "GET,POST,PUT,PATCH,DELETE,OPTIONS,HEAD"
|
||
response.headers["Access-Control-Allow-Headers"] = "Content-Type, X-GA-E2E-Token"
|
||
|
||
|
||
@web.middleware
|
||
async def cors_middleware(request, handler):
|
||
origin = request.headers.get("Origin")
|
||
origin_error = _request_origin_error(request)
|
||
if origin_error:
|
||
return web.json_response(
|
||
{"ok": False, "error": origin_error, "code": "origin_forbidden"},
|
||
status=403,
|
||
)
|
||
if request.method == "OPTIONS":
|
||
response: web.StreamResponse = web.Response(status=204)
|
||
else:
|
||
try:
|
||
response = await handler(request)
|
||
except MaintenanceConflict as error:
|
||
response = web.json_response(error.payload(), status=409)
|
||
except web.HTTPException as error:
|
||
response = web.Response(
|
||
status=error.status,
|
||
reason=error.reason,
|
||
body=error.body,
|
||
headers=error.headers,
|
||
)
|
||
except asyncio.CancelledError:
|
||
raise
|
||
except Exception as error:
|
||
bridge_print(f"[bridge] unhandled request error: {type(error).__name__}: {error}")
|
||
response = web.json_response(
|
||
{"ok": False, "error": "internal server error"}, status=500
|
||
)
|
||
_add_cors_response_headers(response, origin)
|
||
return response
|
||
|
||
|
||
def json_ok(data: dict, status: int = 200):
|
||
return web.json_response(
|
||
data,
|
||
status=status,
|
||
dumps=lambda x: json.dumps(x, ensure_ascii=False, default=str),
|
||
)
|
||
|
||
|
||
async def read_json(request) -> dict:
|
||
if request.can_read_body:
|
||
try:
|
||
data = await request.json()
|
||
return data if isinstance(data, dict) else {}
|
||
except Exception:
|
||
return {}
|
||
return {}
|
||
|
||
|
||
async def status_handler(request):
|
||
return json_ok({
|
||
"ok": True,
|
||
"running": True,
|
||
"ready": True,
|
||
"gaRoot": manager.ga_root,
|
||
"mykeyPath": manager.mykey_path,
|
||
"sessionCount": len(manager.sessions),
|
||
"activeSessionId": manager.active_session_id,
|
||
"maintenance": getattr(manager, "_maintenance_kind", None),
|
||
"ws": "/ws",
|
||
"transport": {"http": True, "wsEventsOnly": True},
|
||
})
|
||
|
||
|
||
_SETTINGS = Path.home() / ".ga_desktop_settings.json"
|
||
_UI_KEYS = ("lang", "theme", "appearance", "plain", "llmNo", "fontSize")
|
||
|
||
|
||
def _settings_doc() -> dict:
|
||
return read_settings(_SETTINGS, strict=False)
|
||
|
||
|
||
def _update_settings_doc(mutate) -> dict:
|
||
return update_settings(_SETTINGS, mutate)
|
||
|
||
|
||
def _desktop_ui() -> dict:
|
||
ui = _settings_doc().get("ui")
|
||
return dict(ui) if isinstance(ui, dict) else {}
|
||
|
||
|
||
def _conductor_settings() -> dict:
|
||
conductor = _settings_doc().get("conductor")
|
||
return dict(conductor) if isinstance(conductor, dict) else {}
|
||
|
||
|
||
def _global_default_llm_no() -> int:
|
||
"""全局默认模型下标。会话未绑定(sess.llm_no is None)时回退到它。"""
|
||
no = _desktop_ui().get("llmNo")
|
||
try:
|
||
return int(no) if no is not None else 0
|
||
except (TypeError, ValueError):
|
||
return 0
|
||
|
||
|
||
def _parse_model_no(value: Any) -> Optional[int]:
|
||
if value is None or isinstance(value, bool):
|
||
return None
|
||
try:
|
||
return int(value)
|
||
except (TypeError, ValueError):
|
||
return None
|
||
|
||
|
||
def _resolve_conductor_model_state(doc: dict, profile_count: int) -> dict:
|
||
"""Resolve persisted Conductor config without relying on modulo selection."""
|
||
count = max(0, int(profile_count or 0))
|
||
conductor = doc.get("conductor") if isinstance(doc, dict) else None
|
||
ui = doc.get("ui") if isinstance(doc, dict) else None
|
||
raw_configured = conductor.get("llmNo") if isinstance(conductor, dict) else None
|
||
configured = _parse_model_no(raw_configured)
|
||
ui_default = _parse_model_no(ui.get("llmNo")) if isinstance(ui, dict) else None
|
||
|
||
if count <= 0:
|
||
return {"configured": configured, "effective": None, "fallbackReason": "no_models"}
|
||
if configured is not None and 0 <= configured < count:
|
||
return {"configured": configured, "effective": configured, "fallbackReason": None}
|
||
if ui_default is not None and 0 <= ui_default < count:
|
||
reason = "invalid_configured" if raw_configured is not None else "ui_default"
|
||
return {"configured": configured, "effective": ui_default, "fallbackReason": reason}
|
||
return {"configured": configured, "effective": 0, "fallbackReason": "first_available"}
|
||
|
||
|
||
async def get_config_handler(request):
|
||
profiles = manager.list_model_profiles()
|
||
active = next((p["id"] for p in profiles if p.get("active")), manager.config.get("llmNo", 0))
|
||
cfg = dict(manager.config)
|
||
if "llmNo" not in cfg:
|
||
cfg["llmNo"] = active
|
||
cfg.update(_desktop_ui())
|
||
cfg["conductor"] = _conductor_settings()
|
||
return json_ok({"gaRoot": manager.ga_root, "mykeyPath": manager.mykey_path, "config": cfg})
|
||
|
||
|
||
async def save_config_handler(request):
|
||
data = await read_json(request)
|
||
cfg = data.get("config", data)
|
||
if isinstance(cfg, dict):
|
||
with manager.mutation():
|
||
patch = {k: cfg[k] for k in _UI_KEYS if k in cfg}
|
||
if patch:
|
||
try:
|
||
def update_ui(doc):
|
||
ui = doc["ui"] if isinstance(doc.get("ui"), dict) else {}
|
||
ui.update(patch)
|
||
doc["ui"] = ui
|
||
|
||
_update_settings_doc(update_ui)
|
||
except DesktopSettingsError as error:
|
||
bridge_print(f"[bridge] save ui prefs failed: {error}")
|
||
return json_ok({"ok": False, "error": str(error)}, status=500)
|
||
manager.config.update(cfg)
|
||
return json_ok({"ok": True, "gaRoot": manager.ga_root, "mykeyPath": manager.mykey_path, "config": manager.config})
|
||
|
||
|
||
async def model_profiles_handler(request):
|
||
try:
|
||
pid = request.match_info.get("id")
|
||
if pid is not None:
|
||
profile_id = int(pid)
|
||
if request.method == "GET":
|
||
return json_ok({"profile": manager.get_model_profile(profile_id)})
|
||
if request.method == "PUT":
|
||
data = await read_json(request)
|
||
with manager.mutation():
|
||
return json_ok({"ok": True, **manager.update_model_profile(profile_id, data)})
|
||
if request.method == "DELETE":
|
||
with manager.mutation():
|
||
return json_ok({"ok": True, **manager.delete_model_profile(profile_id)})
|
||
return json_ok({"ok": False, "error": "method not allowed"}, status=405)
|
||
if request.method == "POST":
|
||
data = await read_json(request)
|
||
with manager.mutation():
|
||
return json_ok({"ok": True, **manager.add_model_profile(data)})
|
||
return json_ok({"profiles": manager.list_model_profiles()})
|
||
except MaintenanceConflict:
|
||
raise
|
||
except ValueError as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=400)
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=500)
|
||
|
||
|
||
async def mixin_handler(request):
|
||
"""聚合渠道成员管理：POST 加入 / DELETE 移出 主聚合渠道。"""
|
||
try:
|
||
profile_id = int(request.match_info.get("id"))
|
||
with manager.mutation():
|
||
if request.method == "POST":
|
||
return json_ok({"ok": True, **manager.add_to_mixin(profile_id)})
|
||
if request.method == "DELETE":
|
||
return json_ok({"ok": True, **manager.remove_from_mixin(profile_id)})
|
||
return json_ok({"ok": False, "error": "method not allowed"}, status=405)
|
||
except MaintenanceConflict:
|
||
raise
|
||
except ValueError as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=400)
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=500)
|
||
|
||
|
||
async def mixin_order_handler(request):
|
||
"""渠道组成员拖拽排序：PUT {members:[name,...]}。"""
|
||
try:
|
||
data = await read_json(request)
|
||
with manager.mutation():
|
||
return json_ok({"ok": True, **manager.reorder_mixin(data.get("members") or [])})
|
||
except MaintenanceConflict:
|
||
raise
|
||
except ValueError as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=400)
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=500)
|
||
|
||
|
||
async def list_sessions_handler(request):
|
||
with manager.lock:
|
||
sessions = [manager.snapshot(s, include_messages=False)
|
||
for s in manager.sessions.values() if _is_desktop_session_id(s.id)]
|
||
return json_ok({"sessions": sessions, "activeSessionId": manager.active_session_id})
|
||
|
||
|
||
async def new_session_handler(request):
|
||
data = await read_json(request)
|
||
sess = manager.create_session(cwd=data.get("cwd") or data.get("path"))
|
||
return json_ok({"ok": True, "sessionId": sess.id, "session": manager.snapshot(sess)}, status=201)
|
||
|
||
|
||
async def get_session_handler(request):
|
||
sid = request.match_info["sid"]
|
||
sess = manager.get_session(sid)
|
||
return json_ok({"sessionId": sid, "session": manager.snapshot(sess), "messages": list(sess.messages), "partial": sess.partial})
|
||
|
||
|
||
async def delete_session_handler(request):
|
||
sid = request.match_info["sid"]
|
||
return json_ok(manager.delete_session(sid))
|
||
|
||
|
||
async def patch_session_handler(request):
|
||
sid = request.match_info["sid"]
|
||
data = await read_json(request)
|
||
with manager.mutation():
|
||
sess = manager.get_session(sid)
|
||
if "title" in data:
|
||
sess.title = data["title"]
|
||
sess.untitled = False
|
||
if "pinned" in data:
|
||
sess.pinned = bool(data["pinned"])
|
||
if "untitled" in data:
|
||
sess.untitled = bool(data["untitled"])
|
||
if "plan_scan_baseline" in data:
|
||
sess.plan_scan_baseline = int(data["plan_scan_baseline"])
|
||
sess.updated_at = time.time()
|
||
manager._persist_session(sess)
|
||
return json_ok({"ok": True, "session": manager.snapshot(sess, include_messages=False)})
|
||
|
||
|
||
async def prompt_handler(request):
|
||
sid = request.match_info["sid"]
|
||
data = await read_json(request)
|
||
prompt = data.get("prompt", data.get("content", data.get("message", "")))
|
||
images = data.get("images") or []
|
||
display = data.get("display")
|
||
files_meta = data.get("files") or [] # 非图片附件 [{name, path}]
|
||
image_metas = data.get("imageMetas") or [] # 图片附件 [{name, path}]（不含 dataUrl）
|
||
# 模型不再随 prompt 携带:切换模型走 POST /session/{sid}/model 这一唯一入口,
|
||
# 发消息只使用会话已绑定的 sess.llm_no(未绑定则回退全局默认)。
|
||
return json_ok(manager.submit_prompt(sid, prompt, images, display=display,
|
||
files_meta=files_meta, image_metas=image_metas))
|
||
|
||
|
||
async def messages_handler(request):
|
||
sid = request.match_info["sid"]
|
||
after = int(request.query.get("after") or request.query.get("afterId") or 0)
|
||
limit = int(request.query.get("limit") or 200)
|
||
return json_ok(manager.messages(sid, after=after, limit=limit))
|
||
|
||
|
||
async def cancel_handler(request):
|
||
sid = request.match_info["sid"]
|
||
return json_ok(manager.cancel(sid))
|
||
|
||
|
||
async def restore_handler(request):
|
||
sid = request.match_info["sid"]
|
||
with manager.mutation():
|
||
return json_ok(manager.restore_context(sid))
|
||
|
||
|
||
async def session_model_handler(request):
|
||
sid = request.match_info["sid"]
|
||
data = await read_json(request)
|
||
no = data.get("llmNo", data.get("llm_no"))
|
||
if no is None:
|
||
return json_ok({"ok": False, "error": "missing llmNo"}, status=400)
|
||
try:
|
||
return json_ok(manager.set_session_model(sid, int(no)))
|
||
except (TypeError, ValueError):
|
||
return json_ok({"ok": False, "error": "invalid llmNo"}, status=400)
|
||
|
||
|
||
async def plan_handler(request):
|
||
sid = request.match_info["sid"]
|
||
return json_ok(manager.plan_snapshot(sid))
|
||
|
||
|
||
async def path_open_handler(request):
|
||
data = await read_json(request)
|
||
kind = data.get("kind", "")
|
||
mode = data.get("mode", "open")
|
||
if kind == "mykey":
|
||
target = Path(manager.ga_root) / "mykey.py"
|
||
if not target.exists():
|
||
template = Path(manager.ga_root) / "mykey_template.py"
|
||
target = template if template.exists() else target
|
||
elif kind == "mykeyTemplate":
|
||
target = Path(manager.ga_root) / "mykey_template.py"
|
||
elif kind == "upload":
|
||
raw = Path(data.get("path") or "")
|
||
try:
|
||
resolved = raw.resolve()
|
||
upload_root = _WEB_UPLOAD_DIR.resolve()
|
||
resolved.relative_to(upload_root)
|
||
except (ValueError, OSError):
|
||
return json_ok({"ok": False, "error": "path not in upload dir"}, status=403)
|
||
target = resolved
|
||
else:
|
||
target = Path(data.get("path") or data.get("target") or manager.ga_root)
|
||
target = target.resolve()
|
||
if not target.exists():
|
||
return json_ok({"ok": False, "error": f"File not found: {target}"}, status=404)
|
||
try:
|
||
if mode == "reveal":
|
||
_reveal_path_in_file_manager(target)
|
||
elif kind == "upload":
|
||
_open_path_default(target) # 用户文件用系统默认程序(open 动词),避免 edit 动词 fallback 记事本
|
||
else:
|
||
_open_path_in_editor(target) # mykey 等配置文件仍用编辑器(edit 动词)
|
||
except OSError as e:
|
||
return json_ok({"ok": False, "error": str(e), "path": str(target)}, status=500)
|
||
return json_ok({"ok": True, "path": str(target)})
|
||
|
||
|
||
# File attachments live under GA's own temp dir (gitignored), NOT the OS temp
|
||
# dir, so they survive bridge restarts. Instead of wiping everything on startup,
|
||
# we keep files for UPLOAD_RETENTION_DAYS and only sweep stale ones.
|
||
_WEB_UPLOAD_DIR = Path(DEFAULT_GA_ROOT) / "temp" / "desktop_uploads"
|
||
_WEB_UPLOAD_DIR.mkdir(parents=True, exist_ok=True)
|
||
|
||
UPLOAD_RETENTION_DAYS = 30
|
||
|
||
|
||
def _safe_session_dir(sid: str) -> str:
|
||
"""Sanitize a session id into a safe single-level folder name."""
|
||
s = re.sub(r"[^A-Za-z0-9_-]", "", str(sid or ""))
|
||
return s or "_misc"
|
||
|
||
|
||
def _session_upload_dir(sid: str) -> Path:
|
||
"""Per-session upload subdir under desktop_uploads/, created on demand."""
|
||
d = _WEB_UPLOAD_DIR / _safe_session_dir(sid)
|
||
d.mkdir(parents=True, exist_ok=True)
|
||
return d
|
||
|
||
|
||
def _purge_session_uploads(sid: str) -> None:
|
||
"""Best-effort: drop a session's whole upload subdir when the session is deleted."""
|
||
import shutil
|
||
with contextlib.suppress(Exception):
|
||
shutil.rmtree(_WEB_UPLOAD_DIR / _safe_session_dir(sid), ignore_errors=True)
|
||
|
||
|
||
def _sweep_stale_uploads(retention_days: int = UPLOAD_RETENTION_DAYS) -> None:
|
||
"""Best-effort: delete uploaded files older than retention_days (by mtime),
|
||
then drop empty session subdirs. Replaces the old wholesale rmtree-on-startup
|
||
so attachments persist across restarts while temp storage can't grow forever."""
|
||
cutoff = time.time() - retention_days * 86400
|
||
try:
|
||
for f in _WEB_UPLOAD_DIR.rglob("*"):
|
||
try:
|
||
if f.is_file() and f.stat().st_mtime < cutoff:
|
||
f.unlink()
|
||
except OSError:
|
||
pass
|
||
for d in _WEB_UPLOAD_DIR.iterdir():
|
||
try:
|
||
if d.is_dir() and not any(d.iterdir()):
|
||
d.rmdir()
|
||
except OSError:
|
||
pass
|
||
except OSError:
|
||
pass
|
||
|
||
|
||
_sweep_stale_uploads()
|
||
|
||
|
||
async def upload_handler(request):
|
||
"""Save a file uploaded by the web client and return its absolute path.
|
||
Body: {name: "<original filename>", dataUrl: "data:<mime>;base64,<...>", sid: "<session id>"}
|
||
Files are grouped per session under desktop_uploads/<sid>/ so deleting a
|
||
session can purge its attachments. Missing sid falls back to a _misc bucket.
|
||
Returns: {ok: true, path: "<abs path>"}
|
||
"""
|
||
try:
|
||
data = await request.json()
|
||
if not isinstance(data, dict):
|
||
data = {}
|
||
except web.HTTPRequestEntityTooLarge:
|
||
return json_ok({"ok": False, "error": "file too large for bridge body limit"})
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": f"invalid request: {e}"})
|
||
name = (data.get("name") or "file").strip().replace("/", "_").replace("\\", "_")
|
||
data_url = data.get("dataUrl") or ""
|
||
if "," in data_url:
|
||
b64 = data_url.split(",", 1)[1]
|
||
else:
|
||
b64 = data_url
|
||
try:
|
||
blob = base64.b64decode(b64)
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": f"decode failed: {e}"})
|
||
if not blob:
|
||
return json_ok({"ok": False, "error": "empty file"})
|
||
safe_name = name or "file"
|
||
with manager.mutation():
|
||
fpath = _session_upload_dir(data.get("sid") or "") / f"{uuid.uuid4().hex[:12]}__{safe_name}"
|
||
fpath.write_bytes(blob)
|
||
return json_ok({"ok": True, "path": str(fpath)})
|
||
|
||
|
||
# Max bytes we will base64 back to the webview for an image preview. Larger
|
||
# images (and all non-images) skip the read: the agent opens them by path.
|
||
_DROP_PREVIEW_MAX = 50 * 1024 * 1024
|
||
|
||
|
||
async def drop_stat_handler(request):
|
||
"""Inspect a path dropped onto the window via Tauri's native drag-drop.
|
||
|
||
Body: {path: "<abs path>", preview: <bool>}
|
||
Native drops give absolute paths (not File objects), so the client asks the
|
||
bridge what the path is. Returns is_dir + size for every path; when preview
|
||
is truthy and the target is a readable image under the size cap, also returns
|
||
a base64 data payload so the composer can render a thumbnail. Files and
|
||
folders otherwise travel to the agent by path alone (it reads via file_read
|
||
/ os.walk), so no bytes cross the wire for them.
|
||
"""
|
||
import mimetypes
|
||
data = await read_json(request)
|
||
raw = (data.get("path") or "").strip()
|
||
want_preview = bool(data.get("preview"))
|
||
if not raw:
|
||
return json_ok({"ok": False, "error": "missing path"})
|
||
try:
|
||
target = Path(raw)
|
||
st = target.stat()
|
||
except FileNotFoundError:
|
||
return json_ok({"ok": False, "error": "not found"})
|
||
except OSError as e:
|
||
return json_ok({"ok": False, "error": str(e)})
|
||
is_dir = target.is_dir()
|
||
size = 0 if is_dir else st.st_size
|
||
result = {"ok": True, "is_dir": is_dir, "size": size, "name": target.name}
|
||
if want_preview and not is_dir and size <= _DROP_PREVIEW_MAX:
|
||
ctype = mimetypes.guess_type(target.name)[0] or ""
|
||
if ctype.startswith("image/") and ctype != "image/svg+xml":
|
||
try:
|
||
encoded = base64.b64encode(target.read_bytes()).decode("ascii")
|
||
result["preview"] = f"data:{ctype};base64,{encoded}"
|
||
except OSError:
|
||
pass
|
||
return json_ok(result)
|
||
|
||
|
||
async def upload_delete_handler(request):
|
||
"""Delete a previously-uploaded file. Path must live under _WEB_UPLOAD_DIR."""
|
||
data = await read_json(request)
|
||
raw = data.get("path") or ""
|
||
try:
|
||
with manager.mutation():
|
||
target = Path(raw).resolve()
|
||
upload_root = _WEB_UPLOAD_DIR.resolve()
|
||
if upload_root not in target.parents:
|
||
return json_ok({"ok": False, "error": "path outside upload dir"})
|
||
if target.exists():
|
||
target.unlink()
|
||
return json_ok({"ok": True})
|
||
except MaintenanceConflict:
|
||
raise
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": str(e)})
|
||
|
||
|
||
async def upload_raw_handler(request):
|
||
"""Stream an uploaded file. inline by default (browser preview / <img>),
|
||
?download=1 forces a download. Path must live under _WEB_UPLOAD_DIR
|
||
(whitelist — prevents path traversal). Works for remote browsers too,
|
||
so it covers both 'preview after refresh' and 'download from remote'."""
|
||
import mimetypes
|
||
from urllib.parse import quote
|
||
raw = request.query.get("path", "")
|
||
try:
|
||
target = Path(raw).resolve()
|
||
target.relative_to(_WEB_UPLOAD_DIR.resolve())
|
||
except (ValueError, OSError):
|
||
return web.Response(status=403, text="path not in upload dir")
|
||
if not target.is_file():
|
||
return web.Response(status=404, text="file not found")
|
||
ctype = mimetypes.guess_type(target.name)[0] or "application/octet-stream"
|
||
disp = "attachment" if request.query.get("download") in ("1", "true") else "inline"
|
||
orig_name = target.name.split("__", 1)[-1] # 去掉 <uuid>__ 前缀，还原原始文件名
|
||
return web.Response(
|
||
body=target.read_bytes(),
|
||
content_type=ctype,
|
||
headers={
|
||
"Content-Disposition": f"{disp}; filename*=UTF-8''{quote(orig_name)}",
|
||
"Cache-Control": "no-cache",
|
||
},
|
||
)
|
||
|
||
|
||
def _open_path_in_editor(target: Path) -> None:
|
||
"""Open a file in the user's editor; Windows .py often has no default association."""
|
||
import platform
|
||
path = str(target.resolve())
|
||
if platform.system() == "Windows":
|
||
try:
|
||
os.startfile(path, "edit")
|
||
return
|
||
except OSError:
|
||
pass
|
||
for cmd in (["notepad.exe", path], ["cursor.cmd", path], ["code.cmd", path], ["cursor", path], ["code", path]):
|
||
try:
|
||
subprocess.Popen(cmd, close_fds=True)
|
||
return
|
||
except (FileNotFoundError, OSError):
|
||
continue
|
||
raise OSError(f"No editor available to open: {path}")
|
||
if platform.system() == "Darwin":
|
||
subprocess.Popen(["open", path])
|
||
return
|
||
subprocess.Popen(["xdg-open", path])
|
||
|
||
|
||
def _reveal_path_in_file_manager(target: Path) -> None:
|
||
"""Open the system file manager and select/highlight the target file."""
|
||
import platform
|
||
path = str(target.resolve())
|
||
if platform.system() == "Windows":
|
||
subprocess.Popen(["explorer", "/select,", path])
|
||
return
|
||
if platform.system() == "Darwin":
|
||
subprocess.Popen(["open", "-R", path])
|
||
return
|
||
# Linux: no universal "select file" command; fall back to opening parent dir
|
||
subprocess.Popen(["xdg-open", str(target.parent)])
|
||
|
||
|
||
def _open_path_default(target: Path) -> None:
|
||
"""Open a file with the OS default app (default 'open' verb).
|
||
|
||
For user uploads. Unlike _open_path_in_editor (which uses Windows' 'edit'
|
||
verb and falls back to Notepad), this respects each file type's registered
|
||
default app — PDF viewer, Word, archive tool, etc. — so binaries like pdf
|
||
or docx no longer land in Notepad as garbage."""
|
||
import platform
|
||
path = str(target.resolve())
|
||
if platform.system() == "Windows":
|
||
os.startfile(path) # default "open" verb = double-click behavior
|
||
return
|
||
if platform.system() == "Darwin":
|
||
subprocess.Popen(["open", path])
|
||
return
|
||
subprocess.Popen(["xdg-open", path])
|
||
|
||
|
||
def _mykey_file() -> Path:
|
||
root = Path(manager.ga_root)
|
||
target = root / "mykey.py"
|
||
if not target.is_file():
|
||
template = root / "mykey_template.py"
|
||
if template.is_file():
|
||
target.write_text(template.read_text(encoding="utf-8"), encoding="utf-8")
|
||
return target
|
||
|
||
|
||
async def mykey_get_handler(request):
|
||
target = _mykey_file()
|
||
content = target.read_text(encoding="utf-8") if target.is_file() else ""
|
||
return json_ok({"content": content, "path": str(target)})
|
||
|
||
|
||
async def mykey_save_handler(request):
|
||
data = await read_json(request)
|
||
content = data.get("content")
|
||
if content is None:
|
||
return json_ok({"ok": False, "error": "missing_content"}, status=400)
|
||
try:
|
||
with manager.mutation():
|
||
profiles = manager._save_mykey_text(str(content))
|
||
# Importing/rewriting mykey may recover only extras that are already
|
||
# broken. Healthy tasks keep their current process/model snapshot.
|
||
services.restart_broken_extras()
|
||
except MaintenanceConflict:
|
||
raise
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=400)
|
||
return json_ok({"ok": True, "path": str(manager._mykey_file()), "profiles": profiles})
|
||
|
||
|
||
def _import_memory_from(source_dir: str, ga_root: str) -> dict:
|
||
"""Compatibility wrapper for the transactional Desktop data import."""
|
||
return merge_data_files(source_dir, ga_root)
|
||
|
||
|
||
async def _run_worker_to_completion(function, *args):
|
||
"""A cancelled HTTP task must not release a worker-owned maintenance gate."""
|
||
worker = asyncio.create_task(asyncio.to_thread(function, *args))
|
||
try:
|
||
return await asyncio.shield(worker)
|
||
except asyncio.CancelledError as cancellation:
|
||
while not worker.done():
|
||
try:
|
||
await asyncio.shield(worker)
|
||
except asyncio.CancelledError:
|
||
continue
|
||
except Exception:
|
||
break
|
||
with contextlib.suppress(Exception):
|
||
worker.result()
|
||
raise cancellation
|
||
|
||
|
||
async def memory_import_inspect_handler(request):
|
||
data = await read_json(request)
|
||
source_path = str(data.get("sourcePath") or data.get("sourceDir") or "").strip()
|
||
if not source_path:
|
||
return json_ok({"ok": False, "error": "missing_sourcePath"}, status=400)
|
||
try:
|
||
result = await asyncio.to_thread(inspect_import_source, source_path)
|
||
except (BackupFormatError, OSError, ValueError) as error:
|
||
return json_ok({"ok": False, "error": str(error)}, status=400)
|
||
return json_ok(result)
|
||
|
||
|
||
def _import_data_source(source_path: str) -> dict:
|
||
with materialize_import_source(source_path) as source_root:
|
||
token = manager.begin_maintenance("import", services.running_managed_ids)
|
||
try:
|
||
with manager.lock:
|
||
existing_session_ids = set(manager.sessions)
|
||
result = merge_data_files(
|
||
str(source_root),
|
||
manager.ga_root,
|
||
existing_session_ids=existing_session_ids,
|
||
session_preparer=manager._session_from_item,
|
||
)
|
||
prepared_sessions = result.pop("_preparedSessions", [])
|
||
if prepared_sessions:
|
||
# Files are already committed atomically by merge_data_files. Adopt the
|
||
# prevalidated objects without a second persistence pass that could
|
||
# partially fail or overwrite an existing Desktop session.
|
||
with manager.lock:
|
||
for session in prepared_sessions:
|
||
manager.sessions.setdefault(session.id, session)
|
||
return result
|
||
finally:
|
||
manager.end_maintenance(token)
|
||
|
||
|
||
async def memory_import_handler(request):
|
||
data = await read_json(request)
|
||
source_path = str(data.get("sourcePath") or data.get("sourceDir") or "").strip()
|
||
if not source_path:
|
||
return json_ok({"ok": False, "error": "missing_sourceDir"}, status=400)
|
||
try:
|
||
result = await _run_worker_to_completion(_import_data_source, source_path)
|
||
except MaintenanceConflict as error:
|
||
return json_ok(error.payload(), status=409)
|
||
except (BackupFormatError, ValueError) as error:
|
||
return json_ok({"ok": False, "error": str(error)}, status=400)
|
||
except OSError as error:
|
||
return json_ok({"ok": False, "error": str(error)}, status=500)
|
||
return json_ok(result)
|
||
|
||
|
||
def _export_data_source(destination_path: str, source_mode: str) -> dict:
|
||
token = manager.begin_maintenance("export", services.running_managed_ids)
|
||
try:
|
||
manager._persist(strict=True)
|
||
return export_data_backup(
|
||
manager.ga_root,
|
||
destination_path,
|
||
source_mode,
|
||
forbidden_roots=(
|
||
_WEB_UPLOAD_DIR,
|
||
APP_DIR / "desktop" / "static",
|
||
),
|
||
)
|
||
finally:
|
||
manager.end_maintenance(token)
|
||
|
||
|
||
async def memory_export_handler(request):
|
||
data = await read_json(request)
|
||
destination_path = str(data.get("destinationPath") or "").strip()
|
||
source_mode = str(data.get("sourceMode") or "").strip()
|
||
if not destination_path:
|
||
return json_ok({"ok": False, "error": "missing_destinationPath"}, status=400)
|
||
try:
|
||
result = await _run_worker_to_completion(
|
||
_export_data_source, destination_path, source_mode
|
||
)
|
||
except MaintenanceConflict as error:
|
||
return json_ok(error.payload(), status=409)
|
||
except (BackupFormatError, ValueError) as error:
|
||
return json_ok({"ok": False, "error": str(error)}, status=400)
|
||
except OSError as error:
|
||
return json_ok({"ok": False, "error": str(error)}, status=500)
|
||
return json_ok(result)
|
||
|
||
|
||
async def conductor_model_get_handler(request):
|
||
state = _resolve_conductor_model_state(_settings_doc(), len(manager.list_model_profiles()))
|
||
return json_ok({"model": state})
|
||
|
||
|
||
async def conductor_model_save_handler(request):
|
||
data = await read_json(request)
|
||
try:
|
||
llm_no = int(data.get("llmNo"))
|
||
except (TypeError, ValueError):
|
||
return json_ok({"ok": False, "error": "invalid_llmNo"}, status=400)
|
||
profile_count = len(manager.list_model_profiles())
|
||
if llm_no < 0 or llm_no >= profile_count:
|
||
return json_ok({"ok": False, "error": "model_out_of_range"}, status=400)
|
||
try:
|
||
with manager.mutation():
|
||
def update_conductor(doc):
|
||
conductor = doc["conductor"] if isinstance(doc.get("conductor"), dict) else {}
|
||
conductor["llmNo"] = llm_no
|
||
doc["conductor"] = conductor
|
||
|
||
doc = _update_settings_doc(update_conductor)
|
||
except MaintenanceConflict:
|
||
raise
|
||
except Exception as e:
|
||
return json_ok({"ok": False, "error": str(e)}, status=500)
|
||
state = _resolve_conductor_model_state(doc, profile_count)
|
||
return json_ok({"ok": True, "model": state})
|
||
|
||
|
||
async def service_start_handler(request):
|
||
body = await read_json(request)
|
||
sid = body.get("id") or request.query.get("id")
|
||
if not sid:
|
||
return json_ok({"ok": False, "error": "missing_id"}, status=400)
|
||
result = services.start_service(sid)
|
||
if not result.get("ok"):
|
||
return json_ok(result, status=400)
|
||
return json_ok(result)
|
||
|
||
|
||
async def service_stop_handler(request):
|
||
body = await read_json(request)
|
||
sid = body.get("id") or request.query.get("id")
|
||
if not sid:
|
||
return json_ok({"ok": False, "error": "missing_id"}, status=400)
|
||
return json_ok(services.stop_service(sid))
|
||
|
||
|
||
async def service_logs_handler(request):
|
||
sid = request.query.get("id")
|
||
if not sid:
|
||
return json_ok({"ok": False, "error": "missing_id"}, status=400)
|
||
tail = int(request.query.get("tail") or 200)
|
||
return json_ok(services.read_logs(sid, tail=tail))
|
||
|
||
|
||
async def service_panel_handler(request):
|
||
return json_ok({"services": services.list_panel_state()})
|
||
|
||
|
||
def _is_local_peer(peer: str) -> bool:
|
||
p = (peer or "").strip()
|
||
return p in ("127.0.0.1", "::1") or p.startswith("::ffff:127.0.0.1")
|
||
|
||
|
||
async def stop_extras_handler(request):
|
||
if not _is_local_peer(request.remote or ""):
|
||
return json_ok({"ok": False, "error": "forbidden"}, status=403)
|
||
services.stop_all_extras()
|
||
return json_ok({"ok": True})
|
||
|
||
|
||
async def start_extras_handler(request):
|
||
if not _is_local_peer(request.remote or ""):
|
||
return json_ok({"ok": False, "error": "forbidden"}, status=403)
|
||
with manager.mutation():
|
||
services.autostart_extras()
|
||
return json_ok({"ok": True})
|
||
|
||
|
||
async def identity_handler(request):
|
||
return json_ok({"ga_root": str(DEFAULT_GA_ROOT), "app_dir": str(APP_DIR), "pid": os.getpid(),
|
||
"build_id": os.environ.get("GA_BUILD_ID", "")})
|
||
|
||
|
||
async def service_capabilities_handler(request):
|
||
return json_ok({"dataBackup": True})
|
||
|
||
|
||
def _exit_bridge() -> None:
|
||
with contextlib.suppress(Exception):
|
||
services.stop_all_extras()
|
||
threading.Timer(0.4, lambda: os._exit(0)).start()
|
||
|
||
|
||
async def bridge_exit_handler(request):
|
||
if not _is_local_peer(request.remote or ""):
|
||
return json_ok({"ok": False, "error": "forbidden"}, status=403)
|
||
with manager.mutation():
|
||
# Make graceful shutdown admission irreversible before scheduling the
|
||
# delayed process exit. Otherwise a new import/export could acquire the
|
||
# maintenance gate during the response-to-exit timer window.
|
||
manager._shutdown_requested = True
|
||
_exit_bridge()
|
||
return json_ok({"ok": True})
|
||
|
||
|
||
async def e2e_next_turn_handler(request):
|
||
token = _e2e_control_token()
|
||
if token is None:
|
||
raise web.HTTPNotFound()
|
||
if not _is_local_peer(request.remote or ""):
|
||
return json_ok({"ok": False, "error": "forbidden"}, status=403)
|
||
if request.headers.get("X-GA-E2E-Token", "") != token:
|
||
return json_ok({"ok": False, "error": "forbidden"}, status=403)
|
||
try:
|
||
body = await request.json()
|
||
with manager.mutation():
|
||
_set_e2e_next_turn(str(body.get("mode", "")))
|
||
except (ValueError, TypeError, json.JSONDecodeError):
|
||
return json_ok({"ok": False, "error": "mode must be empty"}, status=400)
|
||
return json_ok({"ok": True, "mode": "empty"})
|
||
|
||
|
||
async def token_stats_handler(request):
|
||
try:
|
||
sys.path.insert(0, str(APP_DIR)) if str(APP_DIR) not in sys.path else None
|
||
import cost_tracker
|
||
trackers = cost_tracker.all_trackers()
|
||
records = []
|
||
for k, v in trackers.items():
|
||
model = ''
|
||
sid = k.replace('GA-', '')
|
||
with manager.lock:
|
||
sess = manager.sessions.get(sid)
|
||
if sess and sess.agent:
|
||
try: model = sess.agent.get_llm_name(model=True) or ''
|
||
except Exception: pass
|
||
records.append({"thread": k, "input": v.input, "output": v.output,
|
||
"cacheCreate": v.cache_create, "cacheRead": v.cache_read, "model": model})
|
||
except Exception:
|
||
records = []
|
||
return json_ok({"records": records})
|
||
|
||
|
||
async def get_token_history_handler(request):
|
||
try:
|
||
import cost_tracker
|
||
data = cost_tracker.aggregate_ledger()
|
||
# Enrich session titles from manager
|
||
for entry in data.get("history", []):
|
||
sid = entry.get("sessionId", "")
|
||
with manager.lock:
|
||
sess = manager.sessions.get(sid)
|
||
if sess and sess.title:
|
||
entry["title"] = sess.title
|
||
if not entry.get("model") and sess and sess.agent:
|
||
try:
|
||
entry["model"] = sess.agent.get_llm_name(model=True) or ""
|
||
except Exception:
|
||
pass
|
||
return json_ok(data)
|
||
except Exception:
|
||
return json_ok({"history": [], "snap": {}})
|
||
|
||
|
||
async def subscription_portal_handler(request):
|
||
manager.ensure_ga_import_path()
|
||
try:
|
||
import agentmain as am
|
||
except Exception:
|
||
am = None
|
||
sp = getattr(am, "start_subscription_portal", None) if am else None
|
||
if request.method == "GET":
|
||
return json_ok({"available": bool(sp)})
|
||
if not sp:
|
||
return json_ok({"ok": False, "available": False}, status=404)
|
||
with manager.mutation():
|
||
sp()
|
||
return json_ok({"ok": True})
|
||
|
||
|
||
def create_app():
|
||
app = web.Application(middlewares=[cors_middleware], client_max_size=500 * 1024 * 1024)
|
||
app.router.add_get("/ws", ws_handler)
|
||
app.router.add_get("/status", status_handler)
|
||
app.router.add_get("/config", get_config_handler)
|
||
app.router.add_post("/config", save_config_handler)
|
||
app.router.add_get("/model-profiles", model_profiles_handler)
|
||
app.router.add_post("/model-profiles", model_profiles_handler)
|
||
app.router.add_put("/model-profiles/mixin/order", mixin_order_handler)
|
||
app.router.add_post("/model-profiles/{id}/mixin", mixin_handler)
|
||
app.router.add_delete("/model-profiles/{id}/mixin", mixin_handler)
|
||
app.router.add_get("/model-profiles/{id}", model_profiles_handler)
|
||
app.router.add_put("/model-profiles/{id}", model_profiles_handler)
|
||
app.router.add_delete("/model-profiles/{id}", model_profiles_handler)
|
||
app.router.add_get("/sessions", list_sessions_handler)
|
||
app.router.add_post("/session/new", new_session_handler)
|
||
app.router.add_get("/session/{sid}", get_session_handler)
|
||
app.router.add_delete("/session/{sid}", delete_session_handler)
|
||
app.router.add_patch("/session/{sid}", patch_session_handler)
|
||
app.router.add_post("/session/{sid}/prompt", prompt_handler)
|
||
app.router.add_get("/session/{sid}/messages", messages_handler)
|
||
app.router.add_get("/session/{sid}/plan", plan_handler)
|
||
app.router.add_post("/session/{sid}/cancel", cancel_handler)
|
||
app.router.add_post("/session/{sid}/restore", restore_handler)
|
||
app.router.add_post("/session/{sid}/model", session_model_handler)
|
||
app.router.add_post("/path/open", path_open_handler)
|
||
app.router.add_post("/upload", upload_handler)
|
||
app.router.add_delete("/upload", upload_delete_handler)
|
||
app.router.add_get("/upload/raw", upload_raw_handler)
|
||
app.router.add_post("/drop/stat", drop_stat_handler)
|
||
app.router.add_get("/token-stats", token_stats_handler)
|
||
app.router.add_get("/token-history", get_token_history_handler)
|
||
app.router.add_get("/subscription-portal", subscription_portal_handler)
|
||
app.router.add_post("/subscription-portal", subscription_portal_handler)
|
||
app.router.add_post("/services/start", service_start_handler)
|
||
app.router.add_post("/services/stop", service_stop_handler)
|
||
app.router.add_get("/services/logs", service_logs_handler)
|
||
app.router.add_get("/services/panel", service_panel_handler)
|
||
app.router.add_get("/services/mykey", mykey_get_handler)
|
||
app.router.add_post("/services/mykey", mykey_save_handler)
|
||
app.router.add_post("/memory/import/inspect", memory_import_inspect_handler)
|
||
app.router.add_post("/memory/import", memory_import_handler)
|
||
app.router.add_post("/memory/export", memory_export_handler)
|
||
app.router.add_get("/services/conductor/model", conductor_model_get_handler)
|
||
app.router.add_post("/services/conductor/model", conductor_model_save_handler)
|
||
app.router.add_post("/services/stop-extras", stop_extras_handler)
|
||
app.router.add_post("/services/start-extras", start_extras_handler)
|
||
app.router.add_get("/services/capabilities", service_capabilities_handler)
|
||
app.router.add_get("/services/identity", identity_handler)
|
||
app.router.add_post("/services/bridge/exit", bridge_exit_handler)
|
||
if _e2e_control_token() is not None:
|
||
app.router.add_post("/__e2e__/next-turn", e2e_next_turn_handler)
|
||
|
||
async def on_startup(app):
|
||
hub.loop = asyncio.get_running_loop()
|
||
services.autostart_extras()
|
||
|
||
async def on_shutdown(app):
|
||
services.stop_all_extras()
|
||
|
||
app.on_startup.append(on_startup)
|
||
app.on_shutdown.append(on_shutdown)
|
||
return app
|
||
|
||
|
||
if __name__ == "__main__":
|
||
host = os.environ.get("BRIDGE_HOST", "127.0.0.1")
|
||
port = int(os.environ.get("BRIDGE_PORT", "14168"))
|
||
bridge_print(f"GenericAgent Web2 bridge: http://{host}:{port} ws://{host}:{port}/ws")
|
||
web.run_app(create_app(), host=host, port=port, print=None)
|