autodevloop 0.1.0__py3-none-any.whl

autodevloop/webapp.py

@@ -0,0 +1,1184 @@
+"""A dependency-free local web dashboard for AutoDevLoop.
+
+Bind to localhost only. It can create, start, and stop runs, exposes live run
+state (progress, per-agent timers, token usage, per-agent output), and an
+editable settings page (config + prompt templates) that is locked while a run
+is active. It never asks for API keys; provider switching is a CLI command swap.
+"""
+
+from __future__ import annotations
+
+import json
+import os
+import subprocess
+import sys
+import threading
+from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
+from pathlib import Path
+from typing import Any
+from urllib.parse import parse_qs, urlparse
+
+from . import prompts, registry
+from .config import default_config, deep_merge, load_config, save_config, deep_get
+from .util import (
+    APP_DIR, PROGRESS_FILE, STATE_FILE, STOP_FILE,
+    load_json, now_text, read_text, restore_working_dir, save_json, write_text,
+)
+
+_RUNS: dict[str, subprocess.Popen] = {}
+_LOCK = threading.Lock()
+
+
+# --------------------------------------------------------------------------- #
+# HTTP helpers
+# --------------------------------------------------------------------------- #
+def _json_response(handler: BaseHTTPRequestHandler, data: Any, status: int = 200) -> None:
+    body = json.dumps(data, ensure_ascii=False).encode("utf-8")
+    handler.send_response(status)
+    handler.send_header("Content-Type", "application/json; charset=utf-8")
+    handler.send_header("Content-Length", str(len(body)))
+    handler.end_headers()
+    handler.wfile.write(body)
+
+
+def _text_response(handler: BaseHTTPRequestHandler, text: str,
+                   content_type: str = "text/plain; charset=utf-8", status: int = 200) -> None:
+    body = text.encode("utf-8")
+    handler.send_response(status)
+    handler.send_header("Content-Type", content_type)
+    handler.send_header("Content-Length", str(len(body)))
+    handler.end_headers()
+    handler.wfile.write(body)
+
+
+def _safe_within(root: Path, candidate: Path) -> bool:
+    try:
+        candidate.resolve().relative_to(root.resolve())
+        return True
+    except ValueError:
+        return False
+
+
+# --------------------------------------------------------------------------- #
+# Run process tracking
+# --------------------------------------------------------------------------- #
+def _is_running(dir_str: str) -> bool:
+    with _LOCK:
+        proc = _RUNS.get(str(Path(dir_str).resolve()))
+    return proc is not None and proc.poll() is None
+
+
+def _kill(proc: subprocess.Popen | None) -> None:
+    if proc is None:
+        return
+    try:
+        if os.name == "nt":
+            subprocess.run(["taskkill", "/F", "/T", "/PID", str(proc.pid)], capture_output=True)
+        else:
+            proc.terminate()
+    except Exception:  # noqa: BLE001
+        pass
+    try:
+        proc.kill()
+    except Exception:  # noqa: BLE001
+        pass
+
+
+# --------------------------------------------------------------------------- #
+# Project summaries / config
+# --------------------------------------------------------------------------- #
+def _project_summary(entry: dict[str, Any]) -> dict[str, Any]:
+    root = Path(entry["dir"])
+    state = load_json(root / APP_DIR / STATE_FILE, {})
+    config = load_config(root)
+    running = _is_running(str(root))
+    status = "running" if running else (state.get("status") or "initialized")
+    return {
+        "dir": str(root),
+        "name": entry.get("name") or deep_get(config, "project.name", "") or root.name,
+        "status": status,
+        "phase": state.get("phase", "build"),
+        "current_version": state.get("current_version", 0),
+        "max_versions": state.get("max_versions") or deep_get(config, "project.max_versions", 0),
+        "goal_progress": state.get("goal_progress", 0),
+        "running": running,
+    }
+
+
+def _create_project(payload: dict[str, Any]) -> dict[str, Any]:
+    raw_dir = str(payload.get("dir") or "").strip()
+    goal = str(payload.get("goal") or "").strip()
+    if not raw_dir:
+        return {"ok": False, "error": "project directory is required"}
+    if not goal:
+        return {"ok": False, "error": "goal is required"}
+    root = Path(raw_dir).expanduser().resolve()
+    root.mkdir(parents=True, exist_ok=True)
+
+    config = load_config(root)
+    overrides: dict[str, Any] = {"project": {"goal": goal}, "provider": {}, "pipeline": {}}
+    if payload.get("name"):
+        overrides["project"]["name"] = payload["name"]
+    if payload.get("max_versions"):
+        overrides["project"]["max_versions"] = int(payload["max_versions"])
+    if payload.get("arch_hint"):
+        overrides["project"]["arch_hint"] = payload["arch_hint"]
+    if payload.get("mode"):
+        overrides["pipeline"]["mode"] = payload["mode"]
+    if payload.get("provider"):
+        overrides["provider"]["name"] = payload["provider"]
+    if payload.get("provider_command"):
+        overrides["provider"]["command"] = payload["provider_command"]
+    if payload.get("model"):
+        overrides["provider"]["model"] = payload["model"]
+    save_config(root, deep_merge(config, {k: v for k, v in overrides.items() if v}))
+    (root / APP_DIR).mkdir(parents=True, exist_ok=True)
+    prompts.ensure_templates(root / APP_DIR)
+    registry.register(root, str(payload.get("name") or root.name))
+    return {"ok": True, "dir": str(root)}
+
+
+def _start_run(payload: dict[str, Any]) -> dict[str, Any]:
+    raw_dir = str(payload.get("dir") or "").strip()
+    if not raw_dir:
+        return {"ok": False, "error": "project directory is required"}
+    root = Path(raw_dir).expanduser().resolve()
+    if _is_running(str(root)):
+        return {"ok": False, "error": "a run is already active for this project"}
+
+    config = load_config(root)
+    state = load_json(root / APP_DIR / STATE_FILE, {})
+    goal = deep_get(config, "project.goal", "") or state.get("goal", "")
+    if not goal:
+        return {"ok": False, "error": "no goal configured; create the project first"}
+
+    cmd = [sys.executable, "-u", "-m", "autodevloop", "run", "--project-dir", str(root), "--non-interactive"]
+    if payload.get("reset"):
+        cmd += ["--reset"]
+
+    (root / APP_DIR).mkdir(parents=True, exist_ok=True)
+    log_file = (root / APP_DIR / "web_run.log").open("a", encoding="utf-8")
+    log_file.write(f"\n=== run started {now_text()} ===\n")
+    log_file.flush()
+    pkg_root = str(Path(__file__).resolve().parents[1])
+    env = dict(os.environ)
+    env["PYTHONPATH"] = pkg_root + os.pathsep + env.get("PYTHONPATH", "")
+    env["PYTHONIOENCODING"] = "utf-8"
+    proc = subprocess.Popen(cmd, stdout=log_file, stderr=subprocess.STDOUT, cwd=pkg_root, env=env)
+    with _LOCK:
+        _RUNS[str(root)] = proc
+    return {"ok": True, "dir": str(root), "pid": proc.pid}
+
+
+def _rollback_current(root: Path) -> tuple[bool, int]:
+    """Restore current/ to the last completed version. Returns (rolled, version)."""
+    state = load_json(root / APP_DIR / STATE_FILE, {})
+    cur = int(state.get("current_version", 0) or 0)
+    target = root / "current"
+    before = root / APP_DIR / "work" / f"v{cur + 1}" / "_before"
+    if before.exists():
+        restore_working_dir(before, target)
+        return True, cur
+    snap = root / "versions" / f"v{cur}"
+    if cur > 0 and snap.exists():
+        restore_working_dir(snap, target)
+        return True, cur
+    return False, cur
+
+
+def _stop_run(payload: dict[str, Any]) -> dict[str, Any]:
+    raw_dir = str(payload.get("dir") or "").strip()
+    mode = str(payload.get("mode") or "graceful")
+    if not raw_dir:
+        return {"ok": False, "error": "dir required"}
+    root = Path(raw_dir).expanduser().resolve()
+    app_dir = root / APP_DIR
+    app_dir.mkdir(parents=True, exist_ok=True)
+
+    if mode == "immediate":
+        with _LOCK:
+            proc = _RUNS.pop(str(root), None)
+        if proc is None or proc.poll() is not None:
+            # Not tracked (e.g. server restarted). Fall back to a graceful stop.
+            write_text(app_dir / STOP_FILE, f"stop requested at {now_text()}\n")
+            return {"ok": True, "mode": "graceful_fallback",
+                    "note": "process not tracked; requested graceful stop instead"}
+        _kill(proc)
+        rolled, rolled_to = _rollback_current(root)
+        # Patch state + progress so the UI reflects the abort.
+        state = load_json(app_dir / STATE_FILE, {})
+        if state:
+            state["status"] = "stopped"
+            state["stop_reason"] = "Immediate stop: current version discarded and rolled back"
+            save_json(app_dir / STATE_FILE, state)
+        prog = load_json(app_dir / PROGRESS_FILE, {})
+        if prog:
+            prog["status"] = "stopped"
+            prog["active"] = []
+            prog["run_ended_at"] = now_text()
+            prog.setdefault("events", []).append(
+                {"time": now_text(), "step": "STOP", "agent": "", "message": "immediate stop; version discarded"})
+            save_json(app_dir / PROGRESS_FILE, prog, stamp=False)
+        return {"ok": True, "mode": "immediate", "rolled_back": rolled, "rolled_to": rolled_to}
+
+    # graceful
+    write_text(app_dir / STOP_FILE, f"stop requested at {now_text()}\n")
+    return {"ok": True, "mode": "graceful"}
+
+
+def _get_config(dir_str: str) -> dict[str, Any]:
+    root = Path(dir_str).expanduser().resolve()
+    config = load_config(root)
+    app_dir = root / APP_DIR
+    prompts.ensure_templates(app_dir)
+    templates = {name: prompts.load_template(app_dir, name) for name in prompts.TEMPLATE_NAMES}
+    return {"config": config, "templates": templates,
+            "template_names": prompts.TEMPLATE_NAMES,
+            "required_tokens": prompts.REQUIRED_TOKENS,
+            "running": _is_running(str(root))}
+
+
+def _save_config(dir_str: str, payload: dict[str, Any]) -> dict[str, Any]:
+    root = Path(dir_str).expanduser().resolve()
+    if _is_running(str(root)):
+        return {"ok": False, "error": "cannot edit settings while a run is active"}
+    templates = payload.get("templates") or {}
+    # Refuse to save prompts that dropped tokens the engine depends on, so an
+    # edited prompt can't silently break the pipeline.
+    invalid: dict[str, list[str]] = {}
+    for name, body in templates.items():
+        if name in prompts.TEMPLATE_NAMES and isinstance(body, str):
+            missing = prompts.validate_template(name, body)
+            if missing:
+                invalid[name] = missing
+    if invalid:
+        return {"ok": False, "error": "invalid_templates", "invalid": invalid}
+
+    config = payload.get("config")
+    if isinstance(config, dict):
+        save_config(root, deep_merge(default_config(), config))
+    base = prompts.templates_dir(root / APP_DIR)
+    base.mkdir(parents=True, exist_ok=True)
+    for name, body in templates.items():
+        if name in prompts.TEMPLATE_NAMES and isinstance(body, str):
+            write_text(base / f"{name}.md", body.rstrip() + "\n")
+    return {"ok": True}
+
+
+# --------------------------------------------------------------------------- #
+# HTTP handler
+# --------------------------------------------------------------------------- #
+class Handler(BaseHTTPRequestHandler):
+    def log_message(self, *_args: Any) -> None:
+        pass
+
+    def _query(self) -> dict[str, str]:
+        return {k: v[0] for k, v in parse_qs(urlparse(self.path).query).items()}
+
+    def _read_body(self) -> dict[str, Any]:
+        length = int(self.headers.get("Content-Length", 0) or 0)
+        if not length:
+            return {}
+        try:
+            return json.loads(self.rfile.read(length).decode("utf-8"))
+        except (json.JSONDecodeError, UnicodeDecodeError):
+            return {}
+
+    def do_GET(self) -> None:  # noqa: N802
+        path = urlparse(self.path).path
+        if path in ("/", "/index.html"):
+            return _text_response(self, INDEX_HTML, "text/html; charset=utf-8")
+        if path == "/api/projects":
+            return _json_response(self, {"projects": [_project_summary(e) for e in registry.load()]})
+        if path == "/api/state":
+            root = Path(self._query().get("dir", ""))
+            return _json_response(self, load_json(root / APP_DIR / STATE_FILE, {}))
+        if path == "/api/progress":
+            root = Path(self._query().get("dir", ""))
+            prog = load_json(root / APP_DIR / PROGRESS_FILE, {})
+            if isinstance(prog, dict):
+                prog["running"] = _is_running(str(root))
+            return _json_response(self, prog)
+        if path == "/api/config":
+            return _json_response(self, _get_config(self._query().get("dir", "")))
+        if path == "/api/log":
+            return self._serve_log()
+        if path == "/api/doc":
+            return self._serve_doc()
+        return _text_response(self, "not found", status=404)
+
+    def do_POST(self) -> None:  # noqa: N802
+        path = urlparse(self.path).path
+        body = self._read_body()
+        if path == "/api/create":
+            return _json_response(self, _create_project(body))
+        if path == "/api/start":
+            return _json_response(self, _start_run(body))
+        if path == "/api/stop":
+            return _json_response(self, _stop_run(body))
+        if path == "/api/config":
+            return _json_response(self, _save_config(self._query().get("dir", ""), body))
+        return _text_response(self, "not found", status=404)
+
+    def _serve_log(self) -> None:
+        q = self._query()
+        root = Path(q.get("dir", "")).resolve()
+        name = q.get("name", "")
+        target = root / APP_DIR / "logs" / name
+        if not name or not _safe_within(root / APP_DIR / "logs", target) or not target.exists():
+            return _text_response(self, "(log not found)")
+        return _text_response(self, read_text(target))
+
+    def _serve_doc(self) -> None:
+        q = self._query()
+        root = Path(q.get("dir", "")).resolve()
+        which = q.get("name", "changelog")
+        mapping = {"changelog": root / "CHANGELOG.md", "features": root / "FEATURES.md",
+                   "report": root / APP_DIR / "final_report.md", "weblog": root / APP_DIR / "web_run.log"}
+        target = mapping.get(which)
+        if target is None or not target.exists():
+            return _text_response(self, "(not available yet)")
+        return _text_response(self, read_text(target))
+
+
+def serve(host: str = "127.0.0.1", port: int = 8787) -> None:
+    server = ThreadingHTTPServer((host, port), Handler)
+    print(f"[AutoDevLoop] Web dashboard: http://{host}:{port}")
+    print("[AutoDevLoop] Press Ctrl+C to stop the server (running projects keep going).")
+    try:
+        server.serve_forever()
+    except KeyboardInterrupt:
+        print("\n[AutoDevLoop] Web server stopped.")
+    finally:
+        server.server_close()
+
+
+INDEX_HTML = r"""<!doctype html>
+<html lang="en">
+<head>
+<meta charset="utf-8">
+<meta name="viewport" content="width=device-width, initial-scale=1">
+<title>AutoDevLoop</title>
+<style>
+:root{
+  --bg:#f4f6fb; --panel:#ffffff; --panel2:#f0f3f9; --line:#e2e7f0; --line2:#d4dbe8;
+  --text:#1f2733; --muted:#6b7686; --brand:#3b6fe0; --brand-soft:#e8f0ff;
+  --ok:#1f9d57; --ok-soft:#e3f6ec; --bad:#d8472b; --bad-soft:#fdeae6;
+  --warn:#c07a13; --warn-soft:#fdf2dd; --run:#3b6fe0; --run-soft:#e8f0ff;
+  --shadow:0 1px 3px rgba(20,30,60,.07),0 1px 2px rgba(20,30,60,.04);
+}
+*{box-sizing:border-box}
+html,body{height:100%}
+body{margin:0;font-family:-apple-system,"Segoe UI",Roboto,"Helvetica Neue",Arial,"PingFang SC","Microsoft YaHei",sans-serif;background:var(--bg);color:var(--text);font-size:14px;display:flex;flex-direction:column;height:100vh;overflow:hidden}
+header{display:flex;align-items:center;gap:10px;padding:10px 18px;background:var(--panel);border-bottom:1px solid var(--line);box-shadow:var(--shadow);flex-wrap:nowrap;flex-shrink:0}
+header .brand{font-size:16px;font-weight:700;letter-spacing:.2px;white-space:nowrap}
+header .brand b{color:var(--brand)}
+header .sub{color:var(--muted);font-size:12px;white-space:nowrap;overflow:hidden;text-overflow:ellipsis}
+header .spacer{flex:1}
+button{background:var(--brand);color:#fff;border:0;border-radius:8px;padding:8px 14px;cursor:pointer;font-size:13px;font-weight:600;transition:.15s}
+button:hover{filter:brightness(.96)}
+button.sec{background:var(--panel2);border:1px solid var(--line2);color:var(--text)}
+button.ok{background:var(--ok)}
+button.warn{background:var(--warn)}
+button.danger{background:var(--bad)}
+button:disabled{opacity:.45;cursor:not-allowed;filter:none}
+#newBtn{white-space:nowrap}
+.iconbtn{background:var(--panel2);border:1px solid var(--line2);color:var(--text);border-radius:8px;padding:7px 10px;font-weight:600;font-size:13px;white-space:nowrap;display:inline-flex;align-items:center;gap:6px}
+.iconbtn:hover{background:var(--brand-soft);border-color:var(--brand)}
+.langwrap{position:relative}
+.menu{position:absolute;right:0;top:calc(100% + 6px);background:var(--panel);border:1px solid var(--line2);border-radius:10px;box-shadow:0 12px 30px rgba(20,30,60,.18);min-width:140px;z-index:30;display:none;overflow:hidden}
+.menu.show{display:block}
+.menu div{padding:9px 14px;cursor:pointer;font-size:13px}
+.menu div:hover{background:var(--brand-soft)}
+.menu div.sel{color:var(--brand);font-weight:700}
+.layout{display:flex;flex:1;min-height:0}
+.sidebar{width:264px;background:var(--panel);border-right:1px solid var(--line);overflow:auto;padding:14px}
+.sidebar .h{font-size:12px;text-transform:uppercase;letter-spacing:.6px;color:var(--muted);margin-bottom:10px}
+.main{flex:1;overflow:auto;padding:20px 24px}
+.proj{padding:10px 12px;border:1px solid var(--line);border-radius:10px;margin-bottom:9px;cursor:pointer;background:var(--panel)}
+.proj:hover{border-color:var(--line2)}
+.proj.active{border-color:var(--brand);background:var(--brand-soft)}
+.proj .nm{font-weight:600}
+.proj .meta{color:var(--muted);font-size:12px;margin-top:4px;display:flex;gap:6px;align-items:center;flex-wrap:wrap}
+.tabs{display:flex;gap:6px;margin-bottom:18px}
+.tabs .t{padding:8px 14px;border-radius:9px;cursor:pointer;color:var(--muted);font-weight:600}
+.tabs .t:hover{background:var(--panel2)}
+.tabs .t.active{background:var(--brand);color:#fff}
+.toolbar{display:flex;align-items:center;gap:10px;margin-bottom:16px;flex-wrap:wrap}
+.toolbar .rt{margin-left:auto;color:var(--muted);font-size:13px}
+.toolbar .rt b{color:var(--text);font-variant-numeric:tabular-nums}
+.cards{display:grid;grid-template-columns:repeat(auto-fit,minmax(150px,1fr));gap:14px;margin-bottom:18px}
+.card{background:var(--panel);border:1px solid var(--line);border-radius:14px;padding:14px;box-shadow:var(--shadow)}
+.card .lbl{color:var(--muted);font-size:12px;font-weight:600;text-transform:uppercase;letter-spacing:.4px}
+.card .val{font-size:22px;font-weight:700;margin-top:6px;font-variant-numeric:tabular-nums}
+.card .sub{color:var(--muted);font-size:12px;margin-top:4px}
+.bar{height:7px;background:var(--panel2);border-radius:6px;overflow:hidden;margin-top:9px}
+.bar>i{display:block;height:100%;background:var(--brand);transition:width .4s}
+.panel{background:var(--panel);border:1px solid var(--line);border-radius:14px;padding:16px;margin-bottom:16px;box-shadow:var(--shadow)}
+.panel h3{margin:0 0 12px;font-size:14px;display:flex;align-items:center;gap:8px}
+table{width:100%;border-collapse:collapse;font-size:13px}
+th,td{text-align:left;padding:8px 9px;border-bottom:1px solid var(--line);vertical-align:top}
+th{color:var(--muted);font-weight:600}
+.vtable{border:1px solid var(--line);border-radius:10px;overflow:hidden}
+.vtable th{white-space:nowrap;background:var(--panel2);border-bottom:1px solid var(--line2)}
+.vtable td.nowrap,.vtable th{white-space:nowrap}
+.vtable tbody tr:nth-child(even){background:var(--panel2)}
+.vtable tbody tr:hover{background:var(--brand-soft)}
+.badge{display:inline-flex;align-items:center;gap:4px;padding:2px 9px;border-radius:7px;font-size:12px;font-weight:700;background:var(--panel2);color:var(--muted);border:1px solid var(--line2)}
+.badge.ok{color:var(--ok);background:var(--ok-soft);border-color:transparent}
+.badge.bad{color:var(--bad);background:var(--bad-soft);border-color:transparent}
+.badge.warn{color:var(--warn);background:var(--warn-soft);border-color:transparent}
+.pill{display:inline-flex;align-items:center;gap:5px;padding:3px 10px;border-radius:20px;font-size:12px;font-weight:600;border:1px solid var(--line2);background:var(--panel2);color:var(--muted)}
+.pill.ok{color:var(--ok);background:var(--ok-soft);border-color:transparent}
+.pill.bad{color:var(--bad);background:var(--bad-soft);border-color:transparent}
+.pill.run{color:var(--run);background:var(--run-soft);border-color:transparent}
+.pill.warn{color:var(--warn);background:var(--warn-soft);border-color:transparent}
+.pill .dot{width:7px;height:7px;border-radius:50%;background:currentColor}
+.pill.run .dot{animation:blink 1s infinite}
+@keyframes blink{50%{opacity:.3}}
+.agents{display:flex;flex-direction:column;gap:8px}
+.agent-row{display:flex;align-items:center;gap:10px;padding:9px 12px;border:1px solid var(--line);border-radius:10px;background:var(--panel2)}
+.agent-row .nm{font-weight:600}
+.agent-row .st{color:var(--muted);font-size:12px;flex:1;overflow:hidden;text-overflow:ellipsis;white-space:nowrap}
+.agent-row .tm{font-variant-numeric:tabular-nums;color:var(--brand);font-weight:700}
+.events{max-height:360px;overflow:auto;display:flex;flex-direction:column;gap:2px}
+.ev{padding:6px 8px;border-radius:8px;font-size:12.5px;display:flex;gap:8px;align-items:baseline;flex-wrap:wrap}
+.ev:hover{background:var(--panel2)}
+.ev .tm{color:var(--muted);font-variant-numeric:tabular-nums;font-size:11px}
+.ev .stp{font-weight:700;color:var(--brand)}
+.ev .ag{color:var(--warn);font-weight:600}
+.ev .msg{color:var(--text)}
+.ev .snip{flex-basis:100%;color:var(--muted);margin:2px 0 0 0;padding-left:8px;border-left:2px solid var(--line2)}
+.ev-div{display:flex;align-items:center;gap:10px;margin:8px 2px;color:var(--brand);font-weight:700;font-size:12px}
+.ev-div:before,.ev-div:after{content:"";flex:1;height:2px;background:linear-gradient(90deg,transparent,var(--brand-soft),var(--brand))}
+.ev-div:after{background:linear-gradient(90deg,var(--brand),var(--brand-soft),transparent)}
+.logbtn{background:var(--brand-soft);color:var(--brand);border:1px solid var(--brand);border-radius:7px;padding:2px 9px;font-size:11.5px;font-weight:700;cursor:pointer}
+.logbtn:hover{background:var(--brand);color:#fff}
+pre{white-space:pre-wrap;background:var(--panel2);padding:12px;border-radius:10px;max-height:380px;overflow:auto;font-size:12.5px;font-family:ui-monospace,Consolas,monospace;margin:0}
+.viewer-head{display:flex;align-items:center;justify-content:space-between;margin-bottom:8px}
+input,select,textarea{width:100%;background:var(--panel);border:1px solid var(--line2);color:var(--text);border-radius:9px;padding:9px;font-size:13px;font-family:inherit}
+input:disabled,select:disabled,textarea:disabled{background:var(--panel2);color:var(--muted)}
+textarea{min-height:150px;font-family:ui-monospace,Consolas,monospace}
+label{display:block;margin:11px 0 5px;color:var(--muted);font-size:12px;font-weight:600}
+.row{display:grid;grid-template-columns:1fr 1fr;gap:14px}
+.banner{padding:10px 12px;border-radius:10px;background:var(--warn-soft);color:var(--warn);font-weight:600;margin-bottom:12px}
+.modal{position:fixed;inset:0;background:rgba(20,30,60,.45);display:none;align-items:center;justify-content:center;z-index:40}
+.modal.show{display:flex}
+.modal .box{background:var(--panel);border-radius:16px;padding:22px;width:580px;max-height:92vh;overflow:auto;box-shadow:0 20px 60px rgba(20,30,60,.25)}
+.modal .box.wide{width:760px}
+.modal h3{margin:0 0 6px}
+.muted{color:var(--muted)}
+.hint{font-size:12px;color:var(--muted);margin-top:7px}
+.help{display:inline-flex;align-items:center;justify-content:center;width:15px;height:15px;border-radius:50%;background:var(--panel2);border:1px solid var(--line2);color:var(--muted);font-size:10px;font-weight:700;cursor:help;margin-left:5px;vertical-align:middle}
+.help:hover{background:var(--brand);color:#fff;border-color:var(--brand)}
+.tip{position:fixed;z-index:90;max-width:340px;background:#1f2733;color:#fff;padding:8px 11px;border-radius:8px;font-size:12px;line-height:1.5;box-shadow:0 8px 24px rgba(0,0,0,.25);display:none;pointer-events:none}
+.tmpl-tabs{display:flex;flex-wrap:wrap;gap:6px;margin-bottom:10px}
+.tmpl-tabs span{padding:5px 10px;border:1px solid var(--line2);border-radius:8px;cursor:pointer;font-size:12px;font-weight:600;color:var(--muted)}
+.tmpl-tabs span.active{border-color:var(--brand);color:var(--brand);background:var(--brand-soft)}
+.tmpl-tabs span.bad{border-color:var(--bad);color:var(--bad)}
+.tmpl-tabs span.inactive{opacity:.4}
+.linklike{color:var(--brand);cursor:pointer;text-decoration:underline;font-size:12px}
+.tmplnote{font-size:12px;color:var(--warn);margin-top:6px;display:none}
+.chips{display:flex;flex-wrap:wrap;gap:6px;margin:6px 0}
+.chip{font-family:ui-monospace,Consolas,monospace;font-size:11px;padding:2px 7px;border-radius:6px;background:var(--panel2);border:1px solid var(--line2);color:var(--text)}
+.chip.miss{background:var(--bad-soft);border-color:transparent;color:var(--bad)}
+.steptable{border:1px solid var(--line);border-radius:10px;overflow:hidden;margin-top:6px}
+.grouphdr{background:var(--panel2);padding:7px 12px;font-size:11px;font-weight:700;text-transform:uppercase;letter-spacing:.5px;color:var(--muted);border-bottom:1px solid var(--line)}
+.step-row{display:flex;align-items:center;gap:10px;padding:9px 12px;border-bottom:1px solid var(--line)}
+.step-row:last-child{border-bottom:0}
+.step-row input{width:auto;margin:0}
+.step-row .sa{font-weight:700;min-width:148px}
+.step-row .sd{color:var(--muted);font-size:12px;flex:1}
+.step-row .stmpl{font-family:ui-monospace,Consolas,monospace;font-size:11px;color:var(--brand);background:var(--brand-soft);padding:2px 7px;border-radius:6px;white-space:nowrap}
+.step-row.req{background:var(--panel2)}
+.step-row.req .sa{color:var(--muted)}
+.step-row.off{opacity:.45}
+.lockicon{font-size:11px;color:var(--muted)}
+.helpsec{margin:14px 0}
+.helpsec h4{margin:0 0 6px;font-size:14px;color:var(--brand)}
+.helpsec p{margin:4px 0;color:var(--text);font-size:13px;line-height:1.6}
+.helpsec table{margin-top:8px}
+.helpsec td,.helpsec th{font-size:12.5px}
+</style>
+</head>
+<body>
+<header>
+  <div class="brand">Auto<b>Dev</b>Loop</div>
+  <div class="sub" id="brandSub"></div>
+  <div class="spacer"></div>
+  <button class="iconbtn" id="helpBtn" data-tip="" onclick="showHelp()">? <span id="helpLbl"></span></button>
+  <div class="langwrap">
+    <button class="iconbtn" id="langBtn" onclick="toggleLang(event)">🌐 <span id="langCur"></span> ▾</button>
+    <div class="menu" id="langMenu">
+      <div data-l="en" onclick="setLang('en')">English</div>
+      <div data-l="zh" onclick="setLang('zh')">简体中文</div>
+      <div data-l="ja" onclick="setLang('ja')">日本語</div>
+    </div>
+  </div>
+  <button onclick="openNew()" id="newBtn">+ New</button>
+</header>
+<div class="layout">
+  <div class="sidebar">
+    <div class="h" id="sideProjects">Projects</div>
+    <div id="projects"></div>
+  </div>
+  <div class="main" id="main"></div>
+</div>
+
+<div class="modal" id="newModal">
+  <div class="box">
+    <h3 id="mTitle"></h3>
+    <div class="muted" id="mDesc" style="font-size:12px;margin-bottom:8px"></div>
+    <label id="lDir"></label><input id="f_dir" placeholder="E:\path\to\my-app">
+    <label id="lName"></label><input id="f_name">
+    <label id="lGoal"></label><textarea id="f_goal"></textarea>
+    <div class="row">
+      <div><label id="lVer"></label><input id="f_versions" type="number" value="6" min="1"></div>
+      <div><label id="lMode"></label><select id="f_mode"><option value="advanced">advanced</option><option value="simple">simple</option></select></div>
+    </div>
+    <div class="row">
+      <div><label id="lProv"></label><select id="f_provider"><option>claude</option><option>codex</option><option>gemini</option></select></div>
+      <div><label id="lPcmd"></label><input id="f_pcmd" placeholder="claude"></div>
+    </div>
+    <label id="lArch"></label><input id="f_hint">
+    <div class="hint" id="mTip"></div>
+    <div style="margin-top:18px;display:flex;gap:8px;justify-content:flex-end">
+      <button class="sec" onclick="closeNew()" id="bCancel"></button>
+      <button onclick="createProject()" id="bCreate"></button>
+    </div>
+  </div>
+</div>
+
+<div class="modal" id="infoModal">
+  <div class="box" id="infoBox">
+    <h3 id="infoTitle"></h3>
+    <div id="infoBody"></div>
+    <div style="margin-top:18px;display:flex;justify-content:flex-end">
+      <button onclick="closeInfo()" id="infoClose"></button>
+    </div>
+  </div>
+</div>
+
+<div class="tip" id="tip"></div>
+
+<script>
+const I18N = {
+  en:{brandSub:"autonomous AI iteration · local dashboard",projects:"Projects",noProjects:"No projects yet.",
+    selectHint:"Select a project, or create a new one.",newBtn:"+ New project",helpLbl:"Help",close:"Close",
+    tabDash:"Dashboard",tabVer:"Versions",tabDocs:"Docs",tabSet:"Settings",
+    run:"▶ Run",stopG:"Stop (graceful)",stopI:"Stop (discard)",
+    status:"Status",version:"Version",phase:"Phase",goalProg:"Goal progress",calls:"Agent calls",tokens:"Tokens",runtime:"Run time",
+    activeAgents:"Running now",noActive:"No agent running.",liveProgress:"Activity log",noEvents:"No activity yet.",
+    outputViewer:"Agent output",outputHint:"Click an output button in the log to view an agent's full output here.",
+    cVer:"Version",cPhase:"Phase",cScore:"Score",cTests:"Tests",cGoal:"Goal",cSummary:"Summary",cNew:"What's new",pass:"pass",fail:"fail",
+    backlog:"Accepted feature backlog",backlogEmpty:"Empty (fills after the goal is met).",
+    featuresTitle:"FEATURES.md (overview table)",changelogTitle:"CHANGELOG.md",
+    setTitle:"Run configuration",mode:"Mode",maxVer:"Max versions (default)",provider:"Provider",providerCmd:"Provider command",
+    model:"Model (optional)",reviewTh:"Review threshold",valueTh:"Value threshold (feature gate)",fixRetries:"Fix retries",
+    maxPar:"Max parallel agents",retries:"Provider call retries",testCmd:"Test command override",gitVer:"Git versioning",
+    steps:"Pipeline steps",stepsHint:"Each step is an agent with an editable prompt template. Required agents are always on.",
+    grpReq:"Required agents (always on)",grpOpt:"Optional steps",reqd:"required",
+    promptTpl:"Prompt templates",promptHint:"Edit the wording in any language. Keep the {{placeholders}} and JSON field names listed below — they are how the engine injects context and reads the reply.",
+    reqTokens:"Must keep these tokens:",reqOk:"format OK ✓",reqBad:"missing required tokens",
+    save:"Save settings",saved:"saved ✓",locked:"A run is active. Settings are read-only until it stops. Saved settings take effect on the next run.",
+    mTitle:"Create a project",mDesc:"This only creates the project. Edit settings first, then press Run on the dashboard.",
+    lDir:"Project directory (absolute path)",lName:"Project name",lGoal:"Goal / requirement",lVer:"Max versions",lMode:"Mode",
+    lProv:"Provider",lPcmd:"Provider command (optional)",lArch:"Architecture hint (optional)",
+    mTip:"The provider CLI must already be installed and authenticated locally. No API key is entered here.",
+    cancel:"Cancel",create:"Create",on:"on",off:"off",confirmImmediate:"Discard the current unfinished version and roll back to the last completed version?",
+    st_running:"running",st_completed:"completed",st_failed:"failed",st_stopped:"stopped",st_initialized:"not started",st_unknown:"unknown",
+    stopGT:"Graceful stop requested",stopGB:"The current version will finish, then the run stops — nothing is discarded. Any settings you saved apply to the next run.",
+    stopIT:"Stopped immediately",stopIB1:"The unfinished version was discarded and current/ was rolled back to the last completed version (v{n}).",
+    stopIB0:"The run was stopped and the unfinished version discarded. There was no earlier completed version to roll back to, so current/ was reset.",
+    stopFB:"This server wasn't tracking the process (it may have been restarted), so a graceful stop was requested instead.",
+    reqMissTitle:"Prompt format incomplete",reqMissBody:"These templates are missing tokens the engine needs. Add them back, then save again:",
+    helpTitle:"AutoDevLoop — Help",
+    tip_help:"Open the help guide: what every setting, agent and button does.",tip_lang:"Change the interface language.",
+    tip_mode:"simple = a lean loop (plan → develop → test → review) that saves tokens. advanced = adds an architect, goal check, an AI test planner, docs, feature scouting and the value gate.",
+    tip_maxVer:"How many versions to build before stopping. The loop keeps going to this number even after the goal is met — it then adds extra useful features.",
+    tip_provider:"Which coding-agent CLI to drive: claude, codex or gemini. It must already be installed and logged in on this machine.",
+    tip_providerCmd:"Override the command that gets run (a full path or a wrapper). Leave blank to use the provider's default command.",
+    tip_model:"Optional model name/alias passed to the CLI. Leave blank to use the CLI's own default model.",
+    tip_reviewTh:"If a version's review score (0-100) is below this, a fix pass runs before the version is accepted.",
+    tip_valueTh:"In the expand phase, a proposed feature is only accepted when its value score (0-100) reaches this number.",
+    tip_fixRetries:"Maximum fix → re-test rounds when tests fail or the score is too low.",
+    tip_maxPar:"How many development agents may run at the same time when the planner splits work across files.",
+    tip_retries:"Automatic retries when a provider call fails transiently (network / timeout).",
+    tip_testCmd:"Force one specific test/build command for every version (e.g. npm test). Leave blank to auto-detect or let the AI choose.",
+    tip_gitVer:"When on, each version is committed and tagged inside current/. Falls back to folder snapshots if git is unavailable.",
+    tip_run:"Start the loop with the saved settings. Disabled while a run is active.",
+    tip_stopG:"Graceful: let the current version finish, then stop. Nothing is discarded.",
+    tip_stopI:"Discard: kill the agent now, throw away the unfinished version, and roll current/ back to the last completed version.",
+    tip_plan:"Decides what this version delivers and how many dev agents to spawn.",
+    tip_dev:"Writes the code for the version (one or more agents working in parallel).",
+    tip_test:"Runs the tests / build for the version and reports pass or fail.",
+    tip_review:"Scores quality, flags blockers, judges goal completeness and writes the 'what's new' summary.",
+    tip_fix:"Repairs the version when tests fail or the score is below the review threshold.",
+    tip_arch:"One-time architect: picks the stack, layout and test strategy at the very start of the project.",
+    tip_goal_check:"A second, independent agent that confirms whether the original goal is genuinely met.",
+    tip_test_agent:"Let an AI choose the test commands (advanced). Off = built-in test detection only, no extra call.",
+    tip_doc:"Keeps README / design docs accurate on every version.",
+    tip_scout:"After the goal is met, proposes genuinely valuable new features to add next.",
+    tip_evaluate:"Independently scores the scouted features; only high-value ones enter the backlog.",
+    tip_features_doc:"Writes the FEATURES.md overview table (no AI call).",
+    h_modes:"Modes",hb_modes:"simple keeps a lean, cheap loop (plan → develop → test → review). advanced turns on the architect, a separate goal check, an AI test planner, a docs agent, plus feature scouting and the value gate after your goal is reached. You can also toggle individual optional steps below the mode selector.",
+    h_pipe:"Pipeline & agents",hb_pipe:"Every version runs a fixed sequence of agents. Required agents always run; optional ones are toggled in Settings. Each agent is driven by an editable prompt template of the same name.",
+    h_set:"Settings glossary",hb_set:"Hover the ? next to any field for a one-line explanation. Settings are locked while a run is active and take effect on the next run.",
+    h_stop:"Stopping a run",hb_stop:"Graceful stop lets the current version finish and keeps it. Discard stop kills the agent immediately, throws away the half-built version, and rolls the working copy back to the last completed version.",
+    h_score:"Scores & goal progress",hb_score:"Score (0-100) is the reviewer's quality rating of a version. Goal progress (0-100%) is how much of your ORIGINAL request is done. Once goal progress reaches 100% / goal is met, the run switches from the build phase to the expand phase and starts proposing extra features.",
+    h_sec:"Safety",hb_sec:"AutoDevLoop runs AI-generated code and shell test commands on your machine, unattended. Use a dedicated folder or VM, and keep this dashboard on localhost only — it has no login.",
+    h_ph:"Prompt placeholders",hb_ph:"A prompt template is the instruction sent to an agent. Before sending, the engine replaces each {{placeholder}} with real data, and afterwards reads specific JSON fields back. Keep every placeholder its template requires; put it on its own line, introduced by a short label so the model knows what follows. Below: what each one is, which templates use it, and a suggested way to introduce it.",
+    colAgent:"Agent",colKind:"Kind",colTmpl:"Prompt",colWhat:"What it does",colPh:"Placeholder",colUsed:"Used in",colEx:"Suggested phrasing",
+    tmplInactive:"This template's step is turned off in the current mode/config, so it is not used right now and is read-only. Enable its step (or switch to advanced) to edit it.",
+    phHelpLink:"Not sure what a placeholder means? See the Help guide →"},
+
+  zh:{brandSub:"自治 AI 迭代 · 本地面板",projects:"项目",noProjects:"还没有项目。",
+    selectHint:"选择一个项目，或新建一个。",newBtn:"+ 新建项目",helpLbl:"帮助",close:"关闭",
+    tabDash:"总览",tabVer:"版本",tabDocs:"文档",tabSet:"设置",
+    run:"▶ 运行",stopG:"停止（优雅）",stopI:"停止（废弃本版）",
+    status:"状态",version:"版本",phase:"阶段",goalProg:"目标进度",calls:"Agent 调用",tokens:"Tokens",runtime:"运行时长",
+    activeAgents:"正在运行",noActive:"当前没有 agent 在运行。",liveProgress:"活动日志",noEvents:"暂无活动。",
+    outputViewer:"Agent 输出",outputHint:"点击日志中的「输出」按钮，可在此查看该 agent 的完整输出。",
+    cVer:"版本",cPhase:"阶段",cScore:"评分",cTests:"测试",cGoal:"目标",cSummary:"摘要",cNew:"新增/变化",pass:"通过",fail:"失败",
+    backlog:"已接受的功能待办",backlogEmpty:"空（目标达成后才会填充）。",
+    featuresTitle:"FEATURES.md（一览表）",changelogTitle:"CHANGELOG.md",
+    setTitle:"运行配置",mode:"模式",maxVer:"最大版本数（默认）",provider:"Provider",providerCmd:"Provider 命令",
+    model:"模型（可选）",reviewTh:"评审阈值",valueTh:"价值阈值（功能闸门）",fixRetries:"修复重试次数",
+    maxPar:"最大并行 agent 数",retries:"Provider 调用重试",testCmd:"测试命令覆盖",gitVer:"Git 版本管理",
+    steps:"流程步骤",stepsHint:"每个步骤都是一个 agent，对应一个可编辑的 prompt 模板。必需的 agent 始终开启。",
+    grpReq:"必需 agent（始终开启）",grpOpt:"可选步骤",reqd:"必需",
+    promptTpl:"Prompt 模板",promptHint:"可用任意语言修改文字，但请保留 {{占位符}} 和下方列出的 JSON 字段名——引擎靠它们注入上下文并解析回复。",
+    reqTokens:"必须保留这些标记：",reqOk:"格式正确 ✓",reqBad:"缺少必需标记",
+    save:"保存设置",saved:"已保存 ✓",locked:"项目运行中，设置为只读，停止后方可修改。保存的设置在下次运行时生效。",
+    mTitle:"新建项目",mDesc:"这里只会创建项目，不会立即运行。先去设置里调整参数，再到总览页点击「运行」。",
+    lDir:"项目目录（绝对路径）",lName:"项目名称",lGoal:"目标 / 需求",lVer:"最大版本数",lMode:"模式",
+    lProv:"Provider",lPcmd:"Provider 命令（可选）",lArch:"架构提示（可选）",
+    mTip:"所选 provider 的 CLI 必须已在本地安装并登录。此处不需要填写任何 API key。",
+    cancel:"取消",create:"创建",on:"开",off:"关",confirmImmediate:"废弃当前未完成的版本，并回退到上一个已完成的版本？",
+    st_running:"运行中",st_completed:"已完成",st_failed:"失败",st_stopped:"已停止",st_initialized:"未开始",st_unknown:"未知",
+    stopGT:"已请求优雅停止",stopGB:"当前版本会先跑完，然后停止——不会废弃任何内容。你保存的设置将在下次运行时生效。",
+    stopIT:"已立即停止",stopIB1:"未完成的版本已被废弃，current/ 已回退到上一个已完成版本（v{n}）。",
+    stopIB0:"运行已停止，未完成的版本已废弃。没有更早的已完成版本可回退，current/ 已被重置。",
+    stopFB:"本服务没有在跟踪该进程（可能服务重启过），因此改为请求优雅停止。",
+    reqMissTitle:"Prompt 格式不完整",reqMissBody:"以下模板缺少引擎需要的标记。请补回这些标记后再保存：",
+    helpTitle:"AutoDevLoop — 帮助",
+    tip_help:"打开帮助：解释每个设置、agent 和按钮的作用。",tip_lang:"切换界面语言。",
+    tip_mode:"simple = 精简循环（计划 → 开发 → 测试 → 评审），省 token。advanced = 额外加入架构师、目标检查、AI 测试规划、文档、功能发掘和价值闸门。",
+    tip_maxVer:"停止前要构建多少个版本。即使目标已达成，循环也会一直跑到这个数——之后会继续添加有用的周边功能。",
+    tip_provider:"驱动哪个编码 agent 的 CLI：claude、codex 或 gemini。它必须已在本机安装并登录。",
+    tip_providerCmd:"覆盖实际运行的命令（完整路径或包装命令）。留空则使用该 provider 的默认命令。",
+    tip_model:"可选，传给 CLI 的模型名/别名。留空则使用 CLI 自己的默认模型。",
+    tip_reviewTh:"某版本的评审分（0-100）低于此值时，会先跑一轮修复再接受该版本。",
+    tip_valueTh:"扩展阶段，提议的功能只有价值分（0-100）达到此值才会被接受。",
+    tip_fixRetries:"测试失败或评分过低时，最多进行多少轮「修复 → 重测」。",
+    tip_maxPar:"当计划把工作拆给多个开发 agent 时，最多允许多少个同时运行。",
+    tip_retries:"provider 调用因网络/超时等瞬时失败时，自动重试的次数。",
+    tip_testCmd:"为每个版本强制使用某条测试/构建命令（如 npm test）。留空则自动检测或由 AI 决定。",
+    tip_gitVer:"开启后，每个版本都会在 current/ 内 commit 并打 tag。没有 git 时回退为文件夹快照。",
+    tip_run:"用已保存的设置启动循环。运行中时禁用。",
+    tip_stopG:"优雅：让当前版本跑完再停，不废弃任何内容。",
+    tip_stopI:"废弃：立即终止 agent，丢弃未完成的版本，并把 current/ 回退到上一个已完成版本。",
+    tip_plan:"决定本版本要交付什么，以及开几个开发 agent。",
+    tip_dev:"编写本版本的代码（一个或多个 agent 并行）。",
+    tip_test:"运行本版本的测试/构建，并报告通过或失败。",
+    tip_review:"打分、标记阻断问题、判断目标完成度，并写出「本版新增」摘要。",
+    tip_fix:"当测试失败或评分低于评审阈值时，修复该版本。",
+    tip_arch:"一次性架构师：在项目最开始选定技术栈、目录结构和测试策略。",
+    tip_goal_check:"第二个独立 agent，确认原始目标是否真正达成。",
+    tip_test_agent:"由 AI 选择测试命令（高级）。关闭则只用内置测试检测，不额外调用。",
+    tip_doc:"每个版本维护 README / 设计文档的准确性。",
+    tip_scout:"目标达成后，发掘真正有价值的、可继续添加的新功能。",
+    tip_evaluate:"独立给发掘到的功能打分；只有高价值的才进入待办清单。",
+    tip_features_doc:"生成 FEATURES.md 一览表（不调用 AI）。",
+    h_modes:"模式",hb_modes:"simple 保持精简省钱的循环（计划 → 开发 → 测试 → 评审）。advanced 会开启架构师、独立的目标检查、AI 测试规划、文档 agent，以及目标达成后的功能发掘和价值闸门。你也可以在模式选择下方单独开关可选步骤。",
+    h_pipe:"流程与 agent",hb_pipe:"每个版本都会按固定顺序运行一串 agent。必需 agent 总会运行；可选的在设置里开关。每个 agent 由同名的、可编辑的 prompt 模板驱动。",
+    h_set:"设置词条",hb_set:"把鼠标悬停在字段旁的 ? 上即可看到一行解释。运行中设置为只读，保存的设置在下次运行时生效。",
+    h_stop:"停止运行",hb_stop:"优雅停止会让当前版本跑完并保留它。废弃停止会立即终止 agent，丢弃这个还没建完的版本，并把工作目录回退到上一个已完成版本。",
+    h_score:"评分与目标进度",hb_score:"评分（0-100）是评审对某版本质量的打分。目标进度（0-100%）是你最初需求完成了多少。一旦目标进度达到 100% / 目标达成，运行会从 build 阶段切到 expand 阶段，并开始提议额外功能。",
+    h_sec:"安全",hb_sec:"AutoDevLoop 会在你机器上无人值守地运行 AI 生成的代码和 shell 测试命令。请使用专门的文件夹或虚拟机，并让本面板只绑定 localhost——它没有登录鉴权。",
+    h_ph:"Prompt 占位符",hb_ph:"prompt 模板就是发给某个 agent 的指令。发送前，引擎会把每个 {{占位符}} 替换成真实数据；返回后，再从中读取特定的 JSON 字段。请保留该模板要求的所有占位符；把占位符单独放一行，前面加一句简短说明，让模型知道下面是什么内容。下表列出每个占位符的含义、用在哪些模板、以及建议的引导写法。",
+    colAgent:"Agent",colKind:"类型",colTmpl:"Prompt",colWhat:"作用",colPh:"占位符",colUsed:"用于模板",colEx:"建议写法",
+    tmplInactive:"该模板对应的步骤在当前模式/配置下是关闭的，所以现在用不到，处于只读状态。开启它对应的步骤（或切到 advanced 模式）即可编辑。",
+    phHelpLink:"不清楚占位符含义？查看帮助文档 →"},
+
+  ja:{brandSub:"自律型 AI 反復 · ローカルダッシュボード",projects:"プロジェクト",noProjects:"プロジェクトがありません。",
+    selectHint:"プロジェクトを選ぶか、新規作成してください。",newBtn:"+ 新規作成",helpLbl:"ヘルプ",close:"閉じる",
+    tabDash:"ダッシュボード",tabVer:"バージョン",tabDocs:"ドキュメント",tabSet:"設定",
+    run:"▶ 実行",stopG:"停止（安全）",stopI:"停止（破棄）",
+    status:"状態",version:"バージョン",phase:"フェーズ",goalProg:"目標達成度",calls:"エージェント呼出",tokens:"トークン",runtime:"実行時間",
+    activeAgents:"実行中",noActive:"実行中のエージェントはありません。",liveProgress:"アクティビティログ",noEvents:"まだアクティビティはありません。",
+    outputViewer:"エージェント出力",outputHint:"ログ内の「出力」ボタンをクリックすると、ここで全文を表示します。",
+    cVer:"バージョン",cPhase:"フェーズ",cScore:"スコア",cTests:"テスト",cGoal:"目標",cSummary:"概要",cNew:"変更点",pass:"成功",fail:"失敗",
+    backlog:"承認済み機能バックログ",backlogEmpty:"空（目標達成後に追加されます）。",
+    featuresTitle:"FEATURES.md（一覧表）",changelogTitle:"CHANGELOG.md",
+    setTitle:"実行設定",mode:"モード",maxVer:"最大バージョン数（既定）",provider:"プロバイダ",providerCmd:"プロバイダコマンド",
+    model:"モデル（任意）",reviewTh:"レビュー閾値",valueTh:"価値閾値（機能ゲート）",fixRetries:"修正リトライ回数",
+    maxPar:"最大並列エージェント数",retries:"プロバイダ呼出リトライ",testCmd:"テストコマンド上書き",gitVer:"Git バージョン管理",
+    steps:"パイプライン手順",stepsHint:"各手順は編集可能なプロンプトテンプレートを持つエージェントです。必須エージェントは常に有効です。",
+    grpReq:"必須エージェント（常時オン）",grpOpt:"任意の手順",reqd:"必須",
+    promptTpl:"プロンプトテンプレート",promptHint:"文言は任意の言語で編集できますが、{{プレースホルダ}} と下記の JSON フィールド名は残してください。エンジンがそれらでコンテキストを注入し、応答を解析します。",
+    reqTokens:"次のトークンは必須です：",reqOk:"形式OK ✓",reqBad:"必須トークンが不足",
+    save:"設定を保存",saved:"保存しました ✓",locked:"実行中のため設定は読み取り専用です。停止後に編集できます。保存した設定は次回実行時に反映されます。",
+    mTitle:"プロジェクト作成",mDesc:"ここでは作成のみ行い、すぐには実行しません。先に設定を調整し、ダッシュボードで「実行」を押してください。",
+    lDir:"プロジェクトディレクトリ（絶対パス）",lName:"プロジェクト名",lGoal:"目標 / 要件",lVer:"最大バージョン数",lMode:"モード",
+    lProv:"プロバイダ",lPcmd:"プロバイダコマンド（任意）",lArch:"アーキテクチャのヒント（任意）",
+    mTip:"選択したプロバイダの CLI は事前にインストール・認証済みである必要があります。API キーの入力は不要です。",
+    cancel:"キャンセル",create:"作成",on:"オン",off:"オフ",confirmImmediate:"未完成の現在のバージョンを破棄し、最後に完了したバージョンに戻しますか？",
+    st_running:"実行中",st_completed:"完了",st_failed:"失敗",st_stopped:"停止",st_initialized:"未開始",st_unknown:"不明",
+    stopGT:"安全な停止を要求しました",stopGB:"現在のバージョンが完了してから停止します。破棄はされません。保存した設定は次回実行時に反映されます。",
+    stopIT:"即座に停止しました",stopIB1:"未完成のバージョンは破棄され、current/ は最後に完了したバージョン（v{n}）に戻されました。",
+    stopIB0:"実行を停止し、未完成のバージョンを破棄しました。戻せる完了済みバージョンが無かったため current/ はリセットされました。",
+    stopFB:"このサーバーはプロセスを追跡していなかった（再起動された可能性）ため、代わりに安全な停止を要求しました。",
+    reqMissTitle:"プロンプト形式が不完全です",reqMissBody:"次のテンプレートにエンジンが必要とするトークンがありません。追加してから保存してください：",
+    helpTitle:"AutoDevLoop — ヘルプ",
+    tip_help:"ヘルプを開く：各設定・エージェント・ボタンの役割を説明します。",tip_lang:"表示言語を変更します。",
+    tip_mode:"simple = 軽量ループ（計画 → 開発 → テスト → レビュー）でトークン節約。advanced = アーキテクト、目標チェック、AIテスト計画、ドキュメント、機能探索、価値ゲートを追加。",
+    tip_maxVer:"停止までに作るバージョン数。目標達成後もこの数まで続き、その後は有用な追加機能を作ります。",
+    tip_provider:"駆動する CLI：claude / codex / gemini。本機にインストール・ログイン済みである必要があります。",
+    tip_providerCmd:"実際に実行されるコマンドを上書き（フルパスやラッパー）。空ならプロバイダ既定のコマンド。",
+    tip_model:"任意。CLI に渡すモデル名/別名。空なら CLI 既定のモデル。",
+    tip_reviewTh:"バージョンのレビュースコア（0-100）がこれ未満なら、採用前に修正を行います。",
+    tip_valueTh:"拡張フェーズでは、提案機能の価値スコア（0-100）がこの値に達した場合のみ採用されます。",
+    tip_fixRetries:"テスト失敗やスコア不足時の「修正 → 再テスト」の最大回数。",
+    tip_maxPar:"計画が作業を分割した際に同時実行できる開発エージェントの数。",
+    tip_retries:"ネットワーク/タイムアウト等の一時的失敗時に自動リトライする回数。",
+    tip_testCmd:"全バージョンで使うテスト/ビルドコマンドを固定（例: npm test）。空なら自動検出かAIが選択。",
+    tip_gitVer:"オンにすると各バージョンを current/ 内で commit してタグ付け。git が無ければフォルダスナップショットに退避。",
+    tip_run:"保存済み設定でループを開始。実行中は無効。",
+    tip_stopG:"安全：現在のバージョンを完了させてから停止。破棄しません。",
+    tip_stopI:"破棄：エージェントを即停止し、未完成バージョンを捨て、current/ を最後の完了版に戻します。",
+    tip_plan:"このバージョンで何を提供するか、開発エージェントを何体にするかを決定。",
+    tip_dev:"バージョンのコードを記述（複数体が並列で動作）。",
+    tip_test:"バージョンのテスト/ビルドを実行し、成功か失敗かを報告。",
+    tip_review:"品質を採点し、阻害要因を指摘、目標達成度を判定、変更点の要約を作成。",
+    tip_fix:"テスト失敗やスコアが閾値未満のとき、バージョンを修復。",
+    tip_arch:"一度きりのアーキテクト：開始時にスタック・構成・テスト戦略を決定。",
+    tip_goal_check:"元の目標が本当に達成されたかを確認する独立エージェント。",
+    tip_test_agent:"AI にテストコマンドを選ばせる（advanced）。オフなら組み込み検出のみ。",
+    tip_doc:"各バージョンで README / 設計文書を正確に保つ。",
+    tip_scout:"目標達成後、本当に価値ある新機能を提案。",
+    tip_evaluate:"探索した機能を独立して採点。高価値のものだけバックログ入り。",
+    tip_features_doc:"FEATURES.md 一覧表を作成（AI呼出なし）。",
+    h_modes:"モード",hb_modes:"simple は軽量・低コストなループ（計画 → 開発 → テスト → レビュー）。advanced はアーキテクト、独立した目標チェック、AIテスト計画、ドキュメント、目標達成後の機能探索と価値ゲートを有効化します。モード選択の下で各任意手順を個別に切替もできます。",
+    h_pipe:"パイプラインとエージェント",hb_pipe:"各バージョンは固定順のエージェント列を実行します。必須は常に実行、任意は設定で切替。各エージェントは同名の編集可能なプロンプトで駆動します。",
+    h_set:"設定用語",hb_set:"各項目の ? にカーソルを合わせると一行説明が出ます。実行中は読み取り専用、保存内容は次回実行時に反映。",
+    h_stop:"実行の停止",hb_stop:"安全停止は現在のバージョンを完了させて保持します。破棄停止はエージェントを即停止し、作りかけのバージョンを捨て、作業コピーを最後の完了版に戻します。",
+    h_score:"スコアと目標達成度",hb_score:"スコア（0-100）はレビューによる品質評価。目標達成度（0-100%）は元の要件の達成割合。100%/目標達成になると build フェーズから expand フェーズへ移り、追加機能の提案を始めます。",
+    h_sec:"安全性",hb_sec:"AutoDevLoop は AI 生成コードと shell テストを無人で実行します。専用フォルダや VM を使い、本ダッシュボードは localhost のみに——ログイン認証はありません。",
+    h_ph:"プロンプトのプレースホルダ",hb_ph:"プロンプトテンプレートはエージェントへ送る指示です。送信前にエンジンが各 {{プレースホルダ}} を実データに置換し、返答から特定の JSON フィールドを読み取ります。テンプレートが要求するプレースホルダは必ず残し、独立した行に置き、短いラベルで何が続くか示してください。下表に各プレースホルダの意味・使用テンプレート・推奨の導入文を示します。",
+    colAgent:"エージェント",colKind:"種別",colTmpl:"プロンプト",colWhat:"役割",colPh:"プレースホルダ",colUsed:"使用テンプレート",colEx:"推奨の書き方",
+    tmplInactive:"このテンプレートの手順は現在のモード/設定でオフのため、今は使われず読み取り専用です。対応手順を有効化（または advanced に切替）すると編集できます。",
+    phHelpLink:"プレースホルダの意味が不明ですか？ヘルプを参照 →"}
+};
+const LANG_LABEL={en:"EN",zh:"中文",ja:"日本語"};
+const REQUIRED_STEPS=[{key:"plan",agent:"AgentPLAN",tmpl:"plan"},{key:"dev",agent:"AgentDEV",tmpl:"dev"},
+  {key:"test",agent:"AgentTEST",tmpl:"test"},{key:"review",agent:"AgentREVIEW",tmpl:"review"},{key:"fix",agent:"AgentFIX",tmpl:"fix"}];
+const OPTIONAL_STEPS=[{key:"arch",agent:"AgentARCH",tmpl:"arch"},{key:"goal_check",agent:"AgentGOALCHECK",tmpl:"goal_check"},
+  {key:"test_agent",agent:"AgentTEST",tmpl:"test"},{key:"doc",agent:"AgentDOC",tmpl:"doc"},{key:"scout",agent:"AgentSCOUT",tmpl:"scout"},
+  {key:"evaluate",agent:"AgentEVALUATE",tmpl:"evaluate"},{key:"features_doc",agent:"—",tmpl:"—"}];
+// Mirror config.py: which optional steps each mode runs by default.
+const SIMPLE_STEPS={arch:true,goal_check:false,test_agent:false,doc:false,scout:false,evaluate:false,features_doc:true};
+const ADVANCED_STEPS={arch:true,goal_check:true,test_agent:true,doc:true,scout:true,evaluate:true,features_doc:true};
+// Which optional step gates each prompt template (null = always used / required).
+const TMPL_STEP={arch:"arch",plan:null,dev:null,doc:"doc",test:"test_agent",review:null,fix:null,scout:"scout",evaluate:"evaluate",goal_check:"goal_check"};
+// Per-placeholder docs for the Help guide: where used + meaning + a suggested
+// lead-in phrase, in each language. [desc, example phrasing].
+const PH={
+  goal:{used:"all",en:["The user's overall goal/requirement for the whole project; present in almost every template.","User goal:"],zh:["用户对整个项目的总目标/需求，几乎每个模板都会用到。","用户目标："],ja:["プロジェクト全体の目標/要件。ほぼ全テンプレートで使用。","ユーザー目標："]},
+  arch_hint:{used:"arch",en:["Optional architecture hint the user typed when creating the project (may be empty). Only AgentARCH sees it.","Extra architecture hints from the user (may be empty):"],zh:["创建项目时用户填写的可选架构提示（可能为空）。仅 AgentARCH 用到。","用户提供的额外架构提示（可能为空）："],ja:["作成時に入力した任意のアーキテクチャヒント（空の場合あり）。AgentARCH のみ使用。","ユーザーからの追加アーキテクチャヒント（空の場合あり）："]},
+  version:{used:"plan, dev, doc, test, review, fix, scout",en:["The version number currently being built (1, 2, 3…).","for version v{{version}}"],zh:["当前正在构建的版本号（1、2、3……）。","针对版本 v{{version}}"],ja:["現在構築中のバージョン番号（1, 2, 3…）。","バージョン v{{version}} 向け"]},
+  phase:{used:"plan, review",en:["Which phase the loop is in: build (driving at the goal) or expand (adding features after it is met).","Phase: {{phase}}"],zh:["循环所处阶段：build（冲目标）或 expand（达成后加功能）。","当前阶段：{{phase}}"],ja:["ループの段階：build（目標へ）か expand（達成後の機能追加）。","フェーズ：{{phase}}"]},
+  architecture:{used:"plan, dev",en:["The architecture report AgentARCH produced (stack, layout, test strategy). Stay consistent with it.","Architecture contract (stay consistent with this):"],zh:["AgentARCH 产出的架构报告（技术栈、目录、测试策略）。请与之保持一致。","架构约定（请与此保持一致）："],ja:["AgentARCH 作成のアーキテクチャ報告（スタック・構成・テスト戦略）。これに従う。","アーキテクチャ規約（これに従う）："]},
+  phase_guidance:{used:"plan",en:["Auto-generated guidance telling the planner whether to complete the goal (build) or extend it (expand).","Guidance for this version:"],zh:["自动生成的指引，告诉计划 agent 是冲目标（build）还是扩展（expand）。","本版本指引："],ja:["計画に build か expand かを伝える自動生成の指針。","本バージョンの指針："]},
+  backlog:{used:"plan, scout",en:["The list of accepted future features (filled during the expand phase). Pick from here when extending.","Accepted feature backlog:"],zh:["已接受的待开发功能清单（扩展阶段填充）。扩展时从这里挑。","已接受的功能待办："],ja:["承認済みの今後の機能リスト（拡張フェーズで蓄積）。拡張時はここから選ぶ。","承認済み機能バックログ："]},
+  previous:{used:"plan",en:["A summary of the previous iteration: last review, last test result, recent version summaries.","Previous iteration context:"],zh:["上一轮迭代的摘要：上次评审、上次测试结果、最近版本摘要。","上一轮迭代上下文："],ja:["前回反復の要約：前回レビュー・テスト結果・直近バージョン概要。","前回反復のコンテキスト："]},
+  context:{used:"plan, test, review, scout, goal_check",en:["A snapshot of the current project's files/contents so the agent can see the real code.","Current project context:"],zh:["当前项目文件/内容的快照，让 agent 能看到真实代码。","当前项目上下文："],ja:["現在のプロジェクトのファイル/内容のスナップショット。実コードを参照可能に。","現在のプロジェクトコンテキスト："]},
+  agent_name:{used:"dev",en:["The name of this specific dev agent (the planner may spawn several).","You are {{agent_name}}"],zh:["当前这个开发 agent 的名字（计划可能会开多个）。","你是 {{agent_name}}"],ja:["この開発エージェントの名前（計画が複数生成する場合あり）。","あなたは {{agent_name}} です"]},
+  plan:{used:"dev, doc, test, review, fix",en:["The JSON plan for this version (version goal, acceptance criteria, dev agents, test focus).","Version plan:"],zh:["本版本的 JSON 计划（版本目标、验收标准、开发 agent、测试重点）。","版本计划："],ja:["本バージョンの JSON 計画（目標・受入基準・開発エージェント・テスト重点）。","バージョン計画："]},
+  task:{used:"dev",en:["This dev agent's specific assignment for the version.","Your specific task:"],zh:["该开发 agent 在本版本中的具体任务。","你的具体任务："],ja:["この開発エージェントの具体的な担当。","あなたの具体的タスク："]},
+  owns:{used:"dev",en:["The files/paths this agent should edit, to avoid clobbering peers.","Files you own (edit only these):"],zh:["该 agent 应编辑的文件/路径，避免覆盖其他 agent。","你负责的文件（只改这些）："],ja:["このエージェントが編集すべきファイル/パス（他者の上書き回避）。","担当ファイル（これのみ編集）："]},
+  candidates:{used:"test, evaluate",en:["In the test template: detected test/build commands. In evaluate: the scouted feature ideas to score.","Candidates:"],zh:["在 test 模板里：检测到的测试/构建命令；在 evaluate 里：待打分的功能点子。","候选项："],ja:["test では検出したテスト/ビルドコマンド。evaluate では採点対象の機能案。","候補："]},
+  test_result:{used:"review, fix",en:["The JSON result of the test run (success flag, commands, output).","Test result:"],zh:["测试运行的 JSON 结果（是否成功、命令、输出）。","测试结果："],ja:["テスト実行の JSON 結果（成否・コマンド・出力）。","テスト結果："]},
+  dev_summaries:{used:"review",en:["What each dev agent reported doing this version.","Development agent summaries:"],zh:["本版本各开发 agent 报告自己做了什么。","开发 agent 摘要："],ja:["本バージョンで各開発エージェントが報告した作業内容。","開発エージェント概要："]},
+  review:{used:"fix, scout, goal_check",en:["The reviewer's JSON verdict (score, issues, goal progress).","Review:"],zh:["评审的 JSON 结论（评分、问题、目标进度）。","评审结果："],ja:["レビューの JSON 判定（スコア・問題・目標達成度）。","レビュー："]},
+  attempt:{used:"fix",en:["Which fix attempt this is (1, 2, …).","fix attempt {{attempt}}"],zh:["这是第几次修复尝试（1、2……）。","第 {{attempt}} 次修复"],ja:["何回目の修正試行か（1, 2…）。","{{attempt}} 回目の修正"]},
+  threshold:{used:"evaluate",en:["The minimum value score a feature needs to be accepted by the gate.","A feature is accepted only when value >= {{threshold}}."],zh:["功能被闸门接受所需的最低价值分。","只有价值 >= {{threshold}} 的功能才被接受。"],ja:["機能がゲートに承認されるための最低価値スコア。","価値 >= {{threshold}} の機能のみ承認。"]}
+};
+
+let lang = localStorage.getItem('adl_lang') || 'en';
+let current=null, tab='dashboard', projects=[], tmplActive=null, cfgCache=null;
+let curEvents=[], renderedCount=0, selectedLog=null, lastProgress=null, dashBuilt=false;
+
+function t(k){return (I18N[lang] && I18N[lang][k]) || (I18N.en[k]) || k;}
+function h(s){return (s===null||s===undefined)?'':String(s).replace(/[&<>"]/g,c=>({'&':'&amp;','<':'&lt;','>':'&gt;','"':'&quot;'}[c]));}
+async function api(p,o){const r=await fetch(p,o);const ct=r.headers.get('content-type')||'';return ct.includes('json')?r.json():r.text();}
+function val(id){const e=document.getElementById(id);return e?e.value.trim():'';}
+function statusLabel(s){return t('st_'+(s||'unknown'))||s;}
+function fmtDur(sec){sec=Math.max(0,Math.floor(sec));const h=Math.floor(sec/3600),m=Math.floor((sec%3600)/60),s=sec%60;return (h?h+'h ':'')+(h||m?m+'m ':'')+s+'s';}
+function helpDot(key){return '<span class="help" data-tip="'+h(t(key))+'">?</span>';}
+
+// ---- floating tooltip ----
+const tipEl=document.getElementById('tip');
+function posTip(e){const pad=14;let x=e.clientX+pad,y=e.clientY+pad;const w=tipEl.offsetWidth,hh=tipEl.offsetHeight;
+  if(x+w>window.innerWidth-8)x=window.innerWidth-w-8;if(y+hh>window.innerHeight-8)y=e.clientY-hh-pad;
+  tipEl.style.left=Math.max(8,x)+'px';tipEl.style.top=Math.max(8,y)+'px';}
+document.addEventListener('mouseover',e=>{const el=e.target.closest('[data-tip]');if(el&&el.getAttribute('data-tip')){tipEl.textContent=el.getAttribute('data-tip');tipEl.style.display='block';posTip(e);}});
+document.addEventListener('mousemove',e=>{if(tipEl.style.display==='block'){const el=e.target.closest('[data-tip]');if(el)posTip(e);else tipEl.style.display='none';}});
+document.addEventListener('mouseout',e=>{const el=e.target.closest('[data-tip]');if(el)tipEl.style.display='none';});
+
+// ---- language menu ----
+function toggleLang(e){e.stopPropagation();document.getElementById('langMenu').classList.toggle('show');}
+document.addEventListener('click',()=>document.getElementById('langMenu').classList.remove('show'));
+function setLang(l){lang=l;localStorage.setItem('adl_lang',l);document.getElementById('langMenu').classList.remove('show');applyStatic();dashBuilt=false;render();loadProjects();}
+function applyStatic(){
+  document.getElementById('brandSub').textContent=t('brandSub');
+  document.getElementById('sideProjects').textContent=t('projects');
+  document.getElementById('newBtn').textContent=t('newBtn');
+  document.getElementById('helpLbl').textContent=t('helpLbl');
+  document.getElementById('helpBtn').setAttribute('data-tip',t('tip_help'));
+  document.getElementById('langBtn').setAttribute('data-tip',t('tip_lang'));
+  document.getElementById('langCur').textContent=LANG_LABEL[lang];
+  document.querySelectorAll('#langMenu div').forEach(d=>d.classList.toggle('sel',d.getAttribute('data-l')===lang));
+}
+
+// ---- generic info / help modal ----
+function showInfo(title,bodyHtml,wide){
+  document.getElementById('infoTitle').textContent=title;
+  document.getElementById('infoBody').innerHTML=bodyHtml;
+  document.getElementById('infoClose').textContent=t('close');
+  document.getElementById('infoBox').classList.toggle('wide',!!wide);
+  document.getElementById('infoModal').classList.add('show');
+}
+function closeInfo(){document.getElementById('infoModal').classList.remove('show');}
+function showHelp(){
+  const secs=[['h_modes','hb_modes'],['h_pipe','hb_pipe'],['h_set','hb_set'],['h_stop','hb_stop'],['h_score','hb_score'],['h_sec','hb_sec']];
+  let html='';
+  html+='<div class="helpsec"><h4>'+h(t('h_modes'))+'</h4><p>'+h(t('hb_modes'))+'</p></div>';
+  // pipeline section + auto table
+  let rows='';
+  REQUIRED_STEPS.forEach(s=>{rows+=pipeRow(s,true);});
+  OPTIONAL_STEPS.forEach(s=>{rows+=pipeRow(s,false);});
+  html+='<div class="helpsec"><h4>'+h(t('h_pipe'))+'</h4><p>'+h(t('hb_pipe'))+'</p>'
+    +'<table><thead><tr><th>'+h(t('colAgent'))+'</th><th>'+h(t('colKind'))+'</th><th>'+h(t('colTmpl'))+'</th><th>'+h(t('colWhat'))+'</th></tr></thead><tbody>'+rows+'</tbody></table></div>';
+  // prompt placeholders section
+  html+='<div class="helpsec" id="help_ph"><h4>'+h(t('h_ph'))+'</h4><p>'+h(t('hb_ph'))+'</p>'
+    +'<table><thead><tr><th>'+h(t('colPh'))+'</th><th>'+h(t('colUsed'))+'</th><th>'+h(t('colWhat'))+'</th><th>'+h(t('colEx'))+'</th></tr></thead><tbody>'+phRows()+'</tbody></table></div>';
+  secs.slice(2).forEach(s=>{html+='<div class="helpsec"><h4>'+h(t(s[0]))+'</h4><p>'+h(t(s[1]))+'</p></div>';});
+  showInfo(t('helpTitle'),html,true);
+}
+function phRows(){
+  return Object.keys(PH).map(k=>{const e=PH[k][lang]||PH[k].en;
+    return '<tr><td><code>{{'+h(k)+'}}</code></td><td class="muted">'+h(PH[k].used)+'</td><td>'+h(e[0])+'</td><td><code>'+h(e[1])+'</code></td></tr>';
+  }).join('');
+}
+function pipeRow(s,req){
+  const kind=req?('<span class="badge">'+h(t('reqd'))+'</span>'):('<span class="badge warn">'+h(t('grpOpt'))+'</span>');
+  return '<tr><td><b>'+h(s.agent)+'</b></td><td>'+kind+'</td><td><code>'+h(s.tmpl)+'</code></td><td>'+h(t('tip_'+s.key))+'</td></tr>';
+}
+
+function openNew(){
+  const m=document.getElementById('newModal');m.classList.add('show');
+  document.getElementById('mTitle').textContent=t('mTitle');
+  document.getElementById('mDesc').textContent=t('mDesc');
+  document.getElementById('lDir').textContent=t('lDir');document.getElementById('lName').textContent=t('lName');
+  document.getElementById('lGoal').textContent=t('lGoal');document.getElementById('lVer').textContent=t('lVer');
+  document.getElementById('lMode').textContent=t('lMode');document.getElementById('lProv').textContent=t('lProv');
+  document.getElementById('lPcmd').textContent=t('lPcmd');document.getElementById('lArch').textContent=t('lArch');
+  document.getElementById('mTip').textContent=t('mTip');
+  document.getElementById('bCancel').textContent=t('cancel');document.getElementById('bCreate').textContent=t('create');
+}
+function closeNew(){document.getElementById('newModal').classList.remove('show');}
+async function createProject(){
+  const payload={dir:val('f_dir'),name:val('f_name'),goal:val('f_goal'),max_versions:val('f_versions'),
+    mode:val('f_mode'),provider:val('f_provider'),provider_command:val('f_pcmd'),arch_hint:val('f_hint')};
+  const res=await api('/api/create',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify(payload)});
+  if(res.ok){closeNew();await loadProjects();selectDir(res.dir);tab='settings';dashBuilt=false;render();}
+  else alert(res.error||'failed');
+}
+
+async function loadProjects(){
+  const d=await api('/api/projects');projects=d.projects||[];
+  const box=document.getElementById('projects');
+  if(!projects.length){box.innerHTML='<div class="muted">'+t('noProjects')+'</div>';return;}
+  box.innerHTML=projects.map((p,i)=>{
+    const cls=p.running?'run':(p.status&&p.status.startsWith('completed')?'ok':(p.status==='failed'?'bad':(p.status==='stopped'?'warn':'')));
+    return '<div class="proj '+(p.dir===current?'active':'')+'" onclick="selectIdx('+i+')">'
+      +'<div class="nm">'+h(p.name)+'</div><div class="meta">'
+      +'<span class="pill '+cls+'">'+(p.running?'<span class="dot"></span>':'')+statusLabel(p.status)+'</span>'
+      +'<span>v'+p.current_version+'/'+p.max_versions+'</span><span>'+(p.goal_progress||0)+'%</span></div></div>';
+  }).join('');
+}
+function selectIdx(i){selectDir(projects[i].dir);}
+function selectDir(dir){current=dir;tab='dashboard';dashBuilt=false;curEvents=[];renderedCount=0;selectedLog=null;loadProjects();render();}
+function setTab(tt){tab=tt;dashBuilt=false;render();}
+function curProj(){return projects.find(p=>p.dir===current);}
+
+async function render(){
+  const main=document.getElementById('main');
+  if(!current){main.innerHTML='<div class="muted">'+t('selectHint')+'</div>';return;}
+  const tabs='<div class="tabs">'
+    +['dashboard','versions','docs','settings'].map(x=>'<div class="t '+(tab===x?'active':'')+'" onclick="setTab(\''+x+'\')">'+t(x==='dashboard'?'tabDash':x==='versions'?'tabVer':x==='docs'?'tabDocs':'tabSet')+'</div>').join('')
+    +'</div>';
+  if(tab==='dashboard'){ if(!dashBuilt){ main.innerHTML=tabs+dashSkeleton(); dashBuilt=true; curEvents=[];renderedCount=0; } await pollDashboard(true); }
+  else if(tab==='versions'){ main.innerHTML=tabs+await renderVersions(); }
+  else if(tab==='docs'){ main.innerHTML=tabs+await renderDocs(); }
+  else if(tab==='settings'){ main.innerHTML=tabs+await renderSettings(); bindTemplate(); }
+}
+
+function dashSkeleton(){
+  return '<div class="toolbar">'
+    +'<button class="ok" id="btnRun" data-tip="'+h(t('tip_run'))+'" onclick="doRun()">'+t('run')+'</button>'
+    +'<button class="warn" id="btnStopG" data-tip="'+h(t('tip_stopG'))+'" onclick="doStop(\'graceful\')">'+t('stopG')+'</button>'
+    +'<button class="danger" id="btnStopI" data-tip="'+h(t('tip_stopI'))+'" onclick="doStop(\'immediate\')">'+t('stopI')+'</button>'
+    +'<span class="rt">'+t('runtime')+': <b id="rtVal">-</b></span></div>'
+    +'<div class="cards" id="cards"></div>'
+    +'<div class="panel"><h3>'+t('activeAgents')+'</h3><div class="agents" id="agents"></div></div>'
+    +'<div class="panel"><h3>'+t('liveProgress')+'</h3><div class="events" id="events"></div></div>'
+    +'<div class="panel"><div class="viewer-head"><h3 style="margin:0">'+t('outputViewer')+'</h3><span class="muted" id="viewerName"></span></div>'
+    +'<pre id="logview" class="muted">'+t('outputHint')+'</pre></div>';
+}
+
+async function pollDashboard(force){
+  if(tab!=='dashboard'||!current) return;
+  const p=await api('/api/progress?dir='+encodeURIComponent(current));
+  lastProgress=p; const running=!!p.running;
+  const br=document.getElementById('btnRun'),bg=document.getElementById('btnStopG'),bi=document.getElementById('btnStopI');
+  if(br){br.disabled=running;bg.disabled=!running;bi.disabled=!running;}
+  const cls=running?'run':((p.status||'').startsWith('completed')?'ok':(p.status==='failed'?'bad':(p.status==='stopped'?'warn':'')));
+  const gp=p.goal_progress||0,tk=p.tokens||{};
+  const cards=document.getElementById('cards');
+  if(cards) cards.innerHTML=
+    card(t('status'),'<span class="pill '+cls+'">'+(running?'<span class="dot"></span>':'')+statusLabel(p.status)+'</span>','')
+    +card(t('version'),'v'+(p.current_version||0)+'/'+(p.max_versions||0),'')
+    +card(t('phase'),p.phase==='expand'?'expand':'build',(p.goal_completed_version?('🎯 v'+p.goal_completed_version):''))
+    +card(t('goalProg'),gp+'%','<div class="bar"><i style="width:'+gp+'%"></i></div>')
+    +card(t('calls'),(p.calls||0),'')
+    +card(t('tokens'),(tk.input||0),'↓in · '+(tk.output||0)+' ↑out');
+  const agbox=document.getElementById('agents');
+  const active=p.active||[];
+  if(agbox){
+    agbox.innerHTML=active.length?active.map(a=>
+      '<div class="agent-row"><span class="nm">'+h(a.agent)+'</span><span class="st">'+h(a.message||'')+'</span>'
+      +'<span class="tm" data-start="'+(a.started_ts||0)+'">0s</span></div>').join('')
+      :'<div class="muted">'+t('noActive')+'</div>';
+  }
+  const evs=p.events||[]; const ebox=document.getElementById('events');
+  if(ebox){
+    if(evs.length<curEvents.length){curEvents=[];renderedCount=0;ebox.innerHTML='';}
+    curEvents=evs;
+    const atBottom=ebox.scrollHeight-ebox.scrollTop-ebox.clientHeight<40;
+    let html='';
+    for(let i=renderedCount;i<evs.length;i++){ html+=evHtml(evs[i],i); }
+    if(html){ ebox.insertAdjacentHTML('beforeend',html); renderedCount=evs.length; if(atBottom||force) ebox.scrollTop=ebox.scrollHeight; }
+    if(!evs.length) ebox.innerHTML='<div class="muted">'+t('noEvents')+'</div>';
+  }
+  tickTimers();
+}
+function card(l,v,s){return '<div class="card"><div class="lbl">'+l+'</div><div class="val">'+v+'</div><div class="sub">'+(s||'')+'</div></div>';}
+function evHtml(e,i){
+  if(e.step==='VERSION_START'){
+    const vno=e.vno||e.version||'';
+    return '<div class="ev-div">v'+h(vno)+' · '+h(e.phase||'')+'</div>';
+  }
+  const log = e.log? '<button class="logbtn" onclick="viewLog('+i+')">📄 '+h(e.agent||'output')+'</button>' : '';
+  const snip = e.snippet? '<div class="snip">'+h(e.snippet)+'</div>' : '';
+  return '<div class="ev"><span class="tm">'+h((e.time||'').slice(11))+'</span>'
+    +'<span class="stp">'+h(e.step)+'</span>'+(e.agent?'<span class="ag">'+h(e.agent)+'</span>':'')
+    +'<span class="msg">'+h(e.message)+'</span>'+log+snip+'</div>';
+}
+async function viewLog(i){
+  const e=curEvents[i]; if(!e||!e.log) return; selectedLog=e.log;
+  const txt=await api('/api/log?dir='+encodeURIComponent(current)+'&name='+encodeURIComponent(e.log));
+  const v=document.getElementById('logview'); if(v){v.classList.remove('muted');v.textContent=txt;}
+  const vn=document.getElementById('viewerName'); if(vn) vn.textContent=(e.agent||'')+' · '+e.log;
+}
+function tickTimers(){
+  const now=Date.now()/1000;
+  document.querySelectorAll('.tm[data-start]').forEach(el=>{const s=parseFloat(el.getAttribute('data-start'))||0;if(s>0)el.textContent=fmtDur(now-s);});
+  const rt=document.getElementById('rtVal');
+  if(rt&&lastProgress){const st=lastProgress.run_started_ts,en=lastProgress.run_ended_ts;
+    if(st){const end=(lastProgress.running||!en)?now:en;rt.textContent=fmtDur(end-st);}else rt.textContent='-';}
+}
+
+async function doRun(){const r=await api('/api/start',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({dir:current})});if(!r.ok)alert(r.error||'failed');setTimeout(()=>pollDashboard(true),600);}
+async function doStop(mode){
+  if(mode==='immediate'&&!confirm(t('confirmImmediate')))return;
+  const r=await api('/api/stop',{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({dir:current,mode:mode})});
+  setTimeout(()=>{pollDashboard(true);loadProjects();},600);
+  if(r&&r.ok){
+    if(r.mode==='graceful'){ showInfo(t('stopGT'),'<p>'+h(t('stopGB'))+'</p>'); }
+    else if(r.mode==='graceful_fallback'){ showInfo(t('stopGT'),'<p>'+h(t('stopFB'))+'</p>'); }
+    else if(r.mode==='immediate'){ const body=r.rolled_back?t('stopIB1').replace('{n}',r.rolled_to):t('stopIB0'); showInfo(t('stopIT'),'<p>'+h(body)+'</p>'); }
+  } else if(r&&r.error){ alert(r.error); }
+}
+
+async function renderVersions(){
+  const s=await api('/api/state?dir='+encodeURIComponent(current));
+  const gv=s.goal_completed_version;
+  const rows=(s.versions||[]).map(v=>{
+    const tests=(v.test_result&&v.test_result.success)?'<span class="badge ok">✅ '+t('pass')+'</span>':'<span class="badge bad">❌ '+t('fail')+'</span>';
+    const gp=v.goal_progress||0;
+    const nw=(v.whats_new||[]).slice(0,3).map(x=>'• '+h(x)).join('<br>');
+    return '<tr><td class="nowrap"><b>v'+v.version+'</b>'+(v.version===gv?' 🎯':'')+'</td>'
+      +'<td class="nowrap"><span class="badge '+(v.phase==='expand'?'warn':'')+'">'+h(v.phase)+'</span></td>'
+      +'<td class="nowrap">'+scoreBadge(v.review_score)+'</td><td class="nowrap">'+tests+'</td>'
+      +'<td class="nowrap">'+gp+'%</td><td>'+h(v.feature_summary)+'</td><td class="muted">'+nw+'</td></tr>';
+  }).join('');
+  const bl=(s.backlog||[]).filter(b=>b.status==='accepted').map(b=>'<tr><td class="nowrap">'+scoreBadge100(b.value)+'</td><td><b>'+h(b.title)+'</b><div class="muted">'+h(b.description)+'</div></td></tr>').join('');
+  return '<div class="panel"><h3>'+t('tabVer')+'</h3><table class="vtable"><thead><tr><th>'+t('cVer')+'</th><th>'+t('cPhase')+'</th><th>'+t('cScore')+'</th><th>'+t('cTests')+'</th><th>'+t('cGoal')+'</th><th>'+t('cSummary')+'</th><th>'+t('cNew')+'</th></tr></thead><tbody>'+(rows||'<tr><td colspan=7 class="muted">-</td></tr>')+'</tbody></table></div>'
+    +'<div class="panel"><h3>'+t('backlog')+'</h3><table class="vtable"><thead><tr><th>'+t('cScore')+'</th><th>'+t('cSummary')+'</th></tr></thead><tbody>'+(bl||'<tr><td colspan=2 class="muted">'+t('backlogEmpty')+'</td></tr>')+'</tbody></table></div>';
+}
+function scoreBadge(n){n=parseInt(n)||0;const c=n>=80?'ok':(n>=60?'warn':'bad');return '<span class="badge '+c+'">'+n+'/100</span>';}
+function scoreBadge100(n){n=parseInt(n)||0;const c=n>=80?'ok':(n>=60?'warn':'');return '<span class="badge '+c+'">'+n+'</span>';}
+
+async function renderDocs(){
+  const ft=await api('/api/doc?dir='+encodeURIComponent(current)+'&name=features');
+  const cl=await api('/api/doc?dir='+encodeURIComponent(current)+'&name=changelog');
+  return '<div class="panel"><h3>'+t('featuresTitle')+'</h3><pre>'+h(ft)+'</pre></div><div class="panel"><h3>'+t('changelogTitle')+'</h3><pre>'+h(cl)+'</pre></div>';
+}
+
+async function renderSettings(){
+  cfgCache=await api('/api/config?dir='+encodeURIComponent(current));
+  const c=cfgCache.config,names=cfgCache.template_names,locked=cfgCache.running;
+  if(!tmplActive||names.indexOf(tmplActive)<0)tmplActive=names[0];
+  const dis=locked?'disabled':'';
+  const banner=locked?'<div class="banner">'+t('locked')+'</div>':'';
+  const tabsHtml=names.map(n=>'<span class="'+(n===tmplActive?'active':'')+'" id="tt_'+n+'" onclick="pickTmpl(\''+n+'\')">'+n+'</span>').join('');
+  const lf=(key,inner)=>'<div><label>'+t(key)+helpDot('tip_'+key)+'</label>'+inner+'</div>';
+  return banner+'<div class="panel"><h3>'+t('setTitle')+'</h3>'
+    +'<div class="row">'
+    +lf('mode','<select id="c_mode" onchange="applyMode(this.value)" '+dis+'><option value="advanced" '+(c.pipeline.mode==='advanced'?'selected':'')+'>advanced</option><option value="simple" '+(c.pipeline.mode==='simple'?'selected':'')+'>simple</option></select>')
+    +lf('maxVer','<input id="c_versions" type="number" value="'+c.project.max_versions+'" '+dis+'>')+'</div>'
+    +'<div class="row">'+lf('provider','<input id="c_provider" value="'+h(c.provider.name)+'" '+dis+'>')
+    +lf('providerCmd','<input id="c_pcmd" value="'+h(c.provider.command)+'" placeholder="claude / codex / gemini" '+dis+'>')+'</div>'
+    +'<div class="row">'+lf('model','<input id="c_model" value="'+h(c.provider.model)+'" '+dis+'>')
+    +lf('reviewTh','<input id="c_review" type="number" value="'+c.review.threshold+'" '+dis+'>')+'</div>'
+    +'<div class="row">'+lf('valueTh','<input id="c_value" type="number" value="'+c.value.threshold+'" '+dis+'>')
+    +lf('fixRetries','<input id="c_fix" type="number" value="'+c.fix.retries+'" '+dis+'>')+'</div>'
+    +'<div class="row">'+lf('maxPar','<input id="c_par" type="number" value="'+c.agents.max_parallel+'" '+dis+'>')
+    +lf('retries','<input id="c_retries" type="number" value="'+c.agents.retries+'" '+dis+'>')+'</div>'
+    +'<div class="row">'+lf('testCmd','<input id="c_test" value="'+h(c.tests.command)+'" '+dis+'>')
+    +lf('gitVer','<select id="c_git" '+dis+'><option value="true" '+(c.vcs.git?'selected':'')+'>'+t('on')+'</option><option value="false" '+(!c.vcs.git?'selected':'')+'>'+t('off')+'</option></select>')+'</div>'
+    +'<label>'+t('steps')+helpDot('stepsHint')+'</label><div class="hint">'+t('stepsHint')+'</div>'
+    +stepsHtml(c,locked)+'</div>'
+    +'<div class="panel"><h3>'+t('promptTpl')+'</h3><div class="tmpl-tabs">'+tabsHtml+'</div>'
+    +'<textarea id="tmpl_text" oninput="validateActive()" '+dis+'></textarea>'
+    +'<div class="tmplnote" id="tmplNote">'+t('tmplInactive')+'</div>'
+    +'<div class="hint">'+t('promptHint')+'</div>'
+    +'<div style="margin-top:6px"><span class="muted" style="font-size:12px">'+t('reqTokens')+'</span><div class="chips" id="reqChips"></div><div id="reqStatus" class="hint"></div>'
+    +'<div style="margin-top:6px"><span class="linklike" onclick="showHelp()">'+t('phHelpLink')+'</span></div></div></div>'
+    +(locked?'':'<div style="display:flex;gap:10px;align-items:center"><button id="saveBtn" onclick="saveSettings()">'+t('save')+'</button><span id="savemsg" class="muted"></span></div>');
+}
+function stepsHtml(c,locked){
+  const ov=(c.pipeline&&c.pipeline.steps)||{};
+  const mode=(c.pipeline&&c.pipeline.mode)||'advanced';
+  const base=mode==='simple'?SIMPLE_STEPS:ADVANCED_STEPS;
+  const row=(s,req)=>{
+    let attrs,off='';
+    if(req){ attrs='checked disabled'; }
+    else {
+      const supported=base[s.key]!==false;          // does this mode run it at all?
+      const forceOff=(mode==='simple'&&!supported);  // advanced-only step under simple
+      const resolved=(ov[s.key]===true||ov[s.key]===false)?ov[s.key]:base[s.key];
+      const checked=forceOff?false:resolved;
+      const disabled=locked||forceOff;
+      attrs='id="step_'+s.key+'" onchange="refreshTmplState()"'+(checked?' checked':'')+(disabled?' disabled':'');
+      if(forceOff)off=' off';
+    }
+    const lock=req?'<span class="lockicon" data-tip="'+h(t('reqd'))+'">🔒</span>':'';
+    return '<div class="step-row'+(req?' req':'')+off+'">'
+      +'<input type="checkbox" '+attrs+'>'
+      +'<span class="sa">'+h(s.agent)+' '+lock+'</span>'
+      +'<span class="stmpl">'+h(s.tmpl)+'</span>'
+      +'<span class="sd">'+h(t('tip_'+s.key))+'</span></div>';
+  };
+  let html='<div class="steptable"><div class="grouphdr">'+t('grpReq')+'</div>';
+  REQUIRED_STEPS.forEach(s=>html+=row(s,true));
+  html+='<div class="grouphdr">'+t('grpOpt')+'</div>';
+  OPTIONAL_STEPS.forEach(s=>html+=row(s,false));
+  return html+'</div>';
+}
+// Live: switching mode resets the optional steps to that mode's defaults and
+// disables/greys the advanced-only ones when simple is selected.
+function applyMode(mode){
+  const base=mode==='simple'?SIMPLE_STEPS:ADVANCED_STEPS;
+  const locked=cfgCache&&cfgCache.running;
+  OPTIONAL_STEPS.forEach(s=>{
+    const el=document.getElementById('step_'+s.key);if(!el)return;
+    const forceOff=(mode==='simple'&&base[s.key]===false);
+    el.checked=forceOff?false:!!base[s.key];
+    el.disabled=!!locked||forceOff;
+    const r=el.closest('.step-row');if(r)r.classList.toggle('off',forceOff);
+  });
+  refreshTmplState();
+}
+function pickTmpl(n){
+  // swap in place (do NOT re-render: that would re-fetch config and drop unsaved edits)
+  if(cfgCache){const ta=document.getElementById('tmpl_text');if(ta)cfgCache.templates[tmplActive]=ta.value;}
+  tmplActive=n;
+  document.querySelectorAll('.tmpl-tabs span').forEach(sp=>sp.classList.toggle('active',sp.id==='tt_'+n));
+  bindTemplate();
+}
+function bindTemplate(){
+  if(!cfgCache)return;
+  const ta=document.getElementById('tmpl_text');if(ta)ta.value=cfgCache.templates[tmplActive]||'';
+  validateActive();
+  refreshTmplState();
+}
+function tmplStepOn(name){const k=TMPL_STEP[name];if(!k)return true;const el=document.getElementById('step_'+k);return el?el.checked:true;}
+// Grey the prompt tabs whose step is off, and make the editor read-only for the
+// active template when its step is off — so prompts follow the mode like agents.
+function refreshTmplState(){
+  if(!cfgCache)return;
+  (cfgCache.template_names||[]).forEach(n=>{const sp=document.getElementById('tt_'+n);if(sp)sp.classList.toggle('inactive',!tmplStepOn(n));});
+  const inactive=!tmplStepOn(tmplActive);
+  const ta=document.getElementById('tmpl_text');if(ta)ta.disabled=(!!cfgCache.running)||inactive;
+  const note=document.getElementById('tmplNote');if(note)note.style.display=inactive?'block':'none';
+}
+function missingTokens(name,body){const req=(cfgCache.required_tokens||{})[name]||[];body=body||'';return req.filter(x=>body.indexOf(x)<0);}
+function validateActive(){
+  if(!cfgCache)return;
+  const ta=document.getElementById('tmpl_text');const body=ta?ta.value:(cfgCache.templates[tmplActive]||'');
+  const req=(cfgCache.required_tokens||{})[tmplActive]||[];
+  const miss=missingTokens(tmplActive,body);
+  const chips=document.getElementById('reqChips');
+  if(chips)chips.innerHTML=req.map(x=>'<span class="chip'+(miss.indexOf(x)>=0?' miss':'')+'">'+h(x)+'</span>').join('')||'<span class="muted" style="font-size:12px">—</span>';
+  const st=document.getElementById('reqStatus');
+  if(st){st.textContent=miss.length?(t('reqBad')+': '+miss.join(', ')):t('reqOk');st.style.color=miss.length?'var(--bad)':'var(--ok)';}
+  const tabSpan=document.getElementById('tt_'+tmplActive);if(tabSpan)tabSpan.classList.toggle('bad',miss.length>0);
+}
+async function saveSettings(){
+  if(cfgCache){const ta=document.getElementById('tmpl_text');if(ta)cfgCache.templates[tmplActive]=ta.value;}
+  // client-side format check across all templates
+  const bad={};(cfgCache.template_names||[]).forEach(n=>{const m=missingTokens(n,cfgCache.templates[n]);if(m.length)bad[n]=m;});
+  if(Object.keys(bad).length){
+    showInfo(t('reqMissTitle'),'<p>'+h(t('reqMissBody'))+'</p>'+Object.keys(bad).map(n=>'<p><b>'+h(n)+'</b>: <code>'+bad[n].map(h).join('</code> <code>')+'</code></p>').join(''));
+    return;
+  }
+  const steps={};['arch','goal_check','test_agent','doc','scout','evaluate','features_doc'].forEach(s=>{const el=document.getElementById('step_'+s);if(el)steps[s]=el.checked;});
+  const config={project:{max_versions:parseInt(val('c_versions')||'6')},
+    provider:{name:val('c_provider'),command:val('c_pcmd'),model:val('c_model')},
+    pipeline:{mode:document.getElementById('c_mode').value,steps:steps},
+    agents:{max_parallel:parseInt(val('c_par')||'3'),retries:parseInt(val('c_retries')||'3')},
+    review:{threshold:parseInt(val('c_review')||'80')},value:{threshold:parseInt(val('c_value')||'65')},
+    fix:{retries:parseInt(val('c_fix')||'2')},tests:{command:val('c_test')},vcs:{git:document.getElementById('c_git').value==='true'}};
+  const btn=document.getElementById('saveBtn');if(btn)btn.disabled=true;
+  const r=await api('/api/config?dir='+encodeURIComponent(current),{method:'POST',headers:{'Content-Type':'application/json'},body:JSON.stringify({config:config,templates:cfgCache.templates})});
+  const m=document.getElementById('savemsg');
+  if(r.ok){if(btn)btn.textContent=t('saved');if(m)m.textContent=t('saved');
+    // reload settings from disk so the user sees the page visibly refresh
+    setTimeout(()=>{cfgCache=null;render();},700);}
+  else{ if(btn)btn.disabled=false;
+    if(r.error==='invalid_templates'){showInfo(t('reqMissTitle'),'<p>'+h(t('reqMissBody'))+'</p>'+Object.keys(r.invalid||{}).map(n=>'<p><b>'+h(n)+'</b>: <code>'+(r.invalid[n]||[]).map(h).join('</code> <code>')+'</code></p>').join(''));}
+    else if(m)m.textContent=r.error||'error'; }
+}
+
+// pollers
+setInterval(()=>{ if(current&&tab==='dashboard') pollDashboard(false).catch(()=>{}); loadProjects().catch(()=>{}); },1800);
+setInterval(tickTimers,1000);
+
+applyStatic();loadProjects();render();
+</script>
+</body>
+</html>
+"""