@remixmate/cli 0.9.72 → 0.9.73

package/skills/web-screenshot/scripts/_media_screenshot/reader.py

@@ -1,14 +1,14 @@
-"""Read operation: load a page and extract its main content as text.
 
-The DOM work lives in ``js/extract_article.js`` (one injected pass returning
-typed blocks); this module drives the browser, renders those blocks into the
-requested format, and enforces the character budget.
 
-Truncation is the part that matters for the caller: a 200k-character page
-pasted into a model's context is worse than useless. ``max_chars`` cuts on a
-block boundary and says so in-band, and ``--output`` keeps the full text on
-disk so nothing is actually lost.
-"""
+
+
+
+
+
+
+
+
+
 from __future__ import annotations
 
 import json
@@ -20,20 +20,20 @@ from . import js_loader
 from .bootstrap import ensure_runtime
 from .browser import apply_pre_action_waits, build_context_options, launch_with_browser_install
 
-# 读正文只需要 HTML 和（SPA 场景下）脚本 / 接口；图片、字体、媒体、样式表一律不拉。
-# 抽取逻辑只读 DOM 文本与属性（img 的 src 照样在），不依赖布局。
+
+
 _SKIPPED_RESOURCE_TYPES = {"image", "media", "font", "stylesheet"}
-# 单个子资源的拉取上限。浏览器要等 defer 脚本 / 阻塞样式表下载完才触发
-# DOMContentLoaded，一个从本机出口挂死的 CDN 就能让整页永远「打不开」
-# （2026-10-06：新加坡出口读腾讯云社区文章，_next chunk 与广州 COS 资源全部挂起）。
-# 超过上限的请求被 abort，浏览器跳过它继续解析。
+
+
+
+
 _SUBRESOURCE_TIMEOUT_MS = 8000
-# 拿到主文档响应头之后，最多再等这么久的 DOMContentLoaded；到点不报错，按当前 DOM 抽取。
+
 _DOM_READY_TIMEOUT_MS = 15000
 
 
 class _SubresourceGate:
-    """page.route 处理器：丢弃不需要的资源，给其余子资源加单请求超时。"""
+
 
     def __init__(self) -> None:
         self.stalled: list[str] = []
@@ -53,7 +53,7 @@ class _SubresourceGate:
             try:
                 route.abort()
             except Exception:
-                pass  # 页面已关闭 / 已导航走
+                pass
             return
         try:
             route.fulfill(response=resp)
@@ -67,9 +67,9 @@ class _SubresourceGate:
 
 
 def _navigation_error(url: str, e: Exception) -> SystemExit:
-    # 一条模型能据以行动的错误，胜过 30 行 Playwright traceback。
-    # 走到这里说明连主文档的响应头都没拿到（子资源挂死已由 _SubresourceGate 兜住），
-    # 问题在网络层或对方服务器；换参数重试几乎不可能有不同结果，要明确告诉调用方别再试。
+
+
+
     first = str(e).splitlines()[0] if str(e) else ""
     kind = "超时未响应" if type(e).__name__ == "TimeoutError" else "连接失败"
     return SystemExit(
@@ -82,7 +82,7 @@ def _navigation_error(url: str, e: Exception) -> SystemExit:
 
 
 def extract(cfg: dict) -> dict:
-    """Load cfg['url'] and return the raw extraction dict from the page."""
+
     ensure_runtime()
     from playwright.sync_api import sync_playwright
 
@@ -104,7 +104,7 @@ def extract(cfg: dict) -> dict:
             page.route("**/*", gate)
             try:
                 try:
-                    # commit = 拿到主文档响应头即返回，不让任何子资源决定「页面打不打得开」。
+
                     response = page.goto(cfg["url"], wait_until="commit")
                 except Exception as e:
                     raise _navigation_error(cfg["url"], e)
@@ -120,13 +120,13 @@ def extract(cfg: dict) -> dict:
                     print(f"[web_read] {note}", file=sys.stderr)
                 status = response.status if response else None
                 if status is not None and status >= 400:
-                    # Keep going: many sites serve real content under a 403/404
-                    # (paywalls, soft 404s). The status rides along in the output
-                    # so the caller can tell "empty page" from "blocked".
+
+
+
                     pass
                 apply_pre_action_waits(page, cfg)
-                # SPA 正文往往在 domcontentloaded 之后才注水。networkidle 等不到就
-                # 算了——静态页本来就不会再有请求，硬等只是白白花掉 20 秒。
+
+
                 try:
                     page.wait_for_load_state("networkidle", timeout=cfg.get("networkIdleMs", 8000))
                 except Exception:
@@ -145,10 +145,10 @@ def extract(cfg: dict) -> dict:
             browser.close()
 
 
-# ── rendering ────────────────────────────────────────────────────────────────
+
 
 def _render_blocks(blocks: list, fmt: str) -> list[str]:
-    """One string per block, ready to be joined with a blank line."""
+
     md = fmt == "markdown"
     out: list[str] = []
     for block in blocks:
@@ -227,14 +227,14 @@ def _header(data: dict, fmt: str) -> str:
 
 
 def render(data: dict, cfg: dict) -> tuple[str, str, bool]:
-    """Return (stdout_text, full_text, truncated).
 
-    ``full_text`` is always the complete document; ``stdout_text`` is what the
-    caller should print, i.e. the same thing cut to ``maxChars``.
-    """
+
+
+
+
     fmt = cfg.get("format", "markdown")
     blocks = list(data.get("blocks") or [])
-    # 标题已经在 header 里印过一次；正文第一个块又是同一句时，去掉重复的那个。
+
     title = ((data.get("metadata") or {}).get("title") or "").strip()
     if blocks and title and blocks[0].get("type") == "heading":
         first = (blocks[0].get("text") or "").strip().rstrip("¶").strip()
@@ -254,8 +254,8 @@ def render(data: dict, cfg: dict) -> tuple[str, str, bool]:
             "blocks": data.get("blocks"),
         }
         full = json.dumps(payload, ensure_ascii=False, indent=2)
-        # JSON 不做块级截断：切一半的 JSON 不是 JSON。超预算时只报告，
-        # 让调用方自己决定是改格式还是配 --output。
+
+
         return full, full, False
 
     full = f"{_header(data, fmt)}\n\n{body}".strip()
@@ -264,7 +264,7 @@ def render(data: dict, cfg: dict) -> tuple[str, str, bool]:
     if max_chars <= 0 or len(full) <= max_chars:
         return full, full, False
 
-    # Cut on a block boundary so the tail is a whole paragraph, not half a word.
+
     head = f"{_header(data, fmt)}\n\n"
     kept: list[str] = []
     used = len(head)
@@ -274,7 +274,7 @@ def render(data: dict, cfg: dict) -> tuple[str, str, bool]:
         kept.append(part)
         used += len(part) + 2
     if not kept:
-        # Single oversized block (one giant <pre>, say) — fall back to a hard cut.
+
         kept = [body[: max(0, max_chars - len(head))]]
     shown = (head + "\n\n".join(kept)).strip()
     notice = (
@@ -285,7 +285,7 @@ def render(data: dict, cfg: dict) -> tuple[str, str, bool]:
 
 
 def do_read(cfg: dict) -> dict:
-    """Extract, render, optionally persist. Returns a small result summary."""
+
     data = extract(cfg)
 
     if data.get("error") == "selector-not-found":

package/skills/web-screenshot/scripts/_media_screenshot/recording.py

@@ -1,4 +1,4 @@
-"""Plain recording: fixed duration / stop-when / scroll-through."""
+
 from __future__ import annotations
 
 import sys
@@ -59,11 +59,11 @@ def _wait_for_stop_condition(page, cfg: dict) -> str:
 
 
 def do_record(cfg: dict) -> tuple[Path, int, int]:
-    """Plain recording — returns (output_path, start_ms, end_ms).
 
-    start_ms/end_ms 是相对视频起点（context 创建）的内容窗口：start_ms=页面加载就绪时刻，
-    end_ms=内容动作结束时刻。调用方据此端锚裁掉开头加载段（skill 侧或委托 ab-render）。
-    """
+
+
+
+
     ensure_runtime()
     from playwright.sync_api import sync_playwright
 
@@ -74,8 +74,8 @@ def do_record(cfg: dict) -> tuple[Path, int, int]:
     with sync_playwright() as p:
         browser = launch_with_browser_install(p, cfg.get("browser"))
         try:
-            # t0 ≈ 视频开始录制的时刻（context 创建即开录）；用于测量加载头/内容窗口，
-            # 供 trim_video 端锚裁剪。
+
+
             t0 = time.time()
             context = browser.new_context(**build_context_options(p, cfg, out_dir))
             if cfg.get("timeout"):
@@ -87,8 +87,8 @@ def do_record(cfg: dict) -> tuple[Path, int, int]:
             try:
                 page.goto(cfg["url"], wait_until="domcontentloaded")
                 apply_pre_action_waits(page, cfg)
-                # 无条件等页面真正加载完成（networkidle + settled 检测 + settleMs）再开录内容——
-                # 对所有 URL 生效。加载/白屏段仍会被录进原始视频，随后由 trim_video 端锚裁掉。
+
+
                 wait_until_settled(page, cfg)
                 start_ms = int((time.time() - t0) * 1000)
                 if cfg.get("scrollThrough"):
@@ -112,6 +112,6 @@ def do_record(cfg: dict) -> tuple[Path, int, int]:
     src = Path(video_path)
     if src.resolve() != output:
         src.replace(output)
-    # 不在此处裁剪：加载头由调用方在收尾/转码那一步端锚裁掉（见 record.py），
-    # 避免 skill 侧依赖 ffmpeg 并省去一次重编码。
+
+
     return output, start_ms, end_ms

package/skills/web-screenshot/scripts/_media_screenshot/scenes/__init__.py

@@ -1,9 +1,9 @@
-"""Scene registry.
 
-Each scene type registers a handler ``(page, scene_cfg) -> None`` via the
-``@register("type")`` decorator. Import the per-scene modules at the bottom so
-their decorators fire at package import time.
-"""
+
+
+
+
+
 from __future__ import annotations
 
 from typing import Callable
@@ -28,9 +28,9 @@ def known_types() -> list[str]:
     return sorted(_REGISTRY.keys())
 
 
-# Auto-discover every non-underscore module in this package so its
-# @register decorators fire at import time. Adding a new scene only
-# requires dropping a `scenes/<type>.py` file — no edits here.
+
+
+
 import importlib as _importlib
 import pkgutil as _pkgutil
 

package/skills/web-screenshot/scripts/_media_screenshot/scenes/arrow.py

@@ -1,4 +1,4 @@
-"""Arrow scene: curved arrow from one anchor (selector/coords) to another."""
+
 from __future__ import annotations
 
 from .. import js_loader
@@ -7,7 +7,7 @@ from . import register
 
 
 def _resolve_anchor(page, anchor: dict) -> tuple[float, float]:
-    """Anchor is either {x, y} or {selector, side}. Returns page-fixed (x, y)."""
+
     if anchor is None:
         raise RuntimeError("arrow: missing anchor (need from/to)")
     if "x" in anchor and "y" in anchor:

package/skills/web-screenshot/scripts/_media_screenshot/scenes/caption.py

@@ -1,4 +1,4 @@
-"""Caption scene: subtitle strip sliding in from top or bottom edge."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/click.py

@@ -1,4 +1,4 @@
-"""Click scene: animated cursor flies to target, ripple, then optional real click."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/code_lines.py

@@ -1,9 +1,9 @@
-"""Code-lines scene: highlight a contiguous line range in a code view.
 
-Default pattern targets GitHub's React blob view (``[data-line-number="{n}"]``),
-which also covers diff views. Override ``linePattern`` to target the old blob
-table (``#LC{n}``) or any other line-keyed element.
-"""
+
+
+
+
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/focus.py

@@ -1,4 +1,4 @@
-"""Focus scene: dim the page with a spotlight cut-out around the target."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/highlight.py

@@ -1,4 +1,4 @@
-"""Highlight scene: rectangle around a target element (or union of several)."""
+
 from __future__ import annotations
 
 from .. import js_loader
@@ -7,8 +7,8 @@ from . import register
 
 
 def _union_box(page, selectors: list[str]) -> dict | None:
-    """Compute the union bounding box of every matching selector. Skips misses."""
-    # Make sure the first one is visible / on-screen before measuring.
+
+
     resolve_locator(page, selectors[0])
     boxes: list[dict] = []
     for sel in selectors:

package/skills/web-screenshot/scripts/_media_screenshot/scenes/hover.py

@@ -1,4 +1,4 @@
-"""Hover scene: move cursor onto target and hold so tooltips/cards reveal."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/redact.py

@@ -1,4 +1,4 @@
-"""Redact scene: blur / pixelate / solid-cover sensitive regions."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/scroll.py

@@ -1,4 +1,4 @@
-"""Scroll scene: step-scroll for ``duration`` or until ``toSelector`` is hit."""
+
 from __future__ import annotations
 
 import time

package/skills/web-screenshot/scripts/_media_screenshot/scenes/sequence.py

@@ -1,4 +1,4 @@
-"""Sequence scene: multiple targets get numbered outlines, revealed in order."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/title_card.py

@@ -1,4 +1,4 @@
-"""Title card scene: full-screen opener/outro with title + subtitle + accent."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/type.py

@@ -1,4 +1,4 @@
-"""Type scene: move cursor to an input, focus, then typewriter the text."""
+
 from __future__ import annotations
 
 from .. import js_loader

package/skills/web-screenshot/scripts/_media_screenshot/scenes/wait.py

@@ -1,4 +1,4 @@
-"""Wait scene: a simple pause between other scenes."""
+
 from __future__ import annotations
 
 from . import register

package/skills/web-screenshot/scripts/_media_screenshot/scenes/zoom.py

@@ -1,4 +1,4 @@
-"""Zoom scene: element-level OR page-level (browser-like) zoom animation."""
+
 from __future__ import annotations
 
 from .. import js_loader
@@ -60,7 +60,7 @@ def scene_zoom(page, scene: dict) -> None:
     target = scene.get("target", "element")
     locator = resolve_locator(page, scene["selector"])
     locator.scroll_into_view_if_needed()
-    # Center in viewport so the element has room to grow.
+
     locator.evaluate(
         "el => el.scrollIntoView({ block: 'center', inline: 'center', behavior: 'instant' })"
     )

package/skills/web-screenshot/scripts/_media_screenshot/screenshot.py

@@ -1,6 +1,6 @@
-"""Screenshot operation: full-page / element / region, with optional
-hide-before / native mask / drawn annotations.
-"""
+
+
+
 from __future__ import annotations
 
 import json
@@ -18,7 +18,7 @@ from .overlay import ensure_overlay
 
 
 def _apply_hide(page, selectors: list[str]) -> None:
-    """display:none every match of every selector (idempotent, scoped to this page)."""
+
     if not selectors:
         return
     page.evaluate(
@@ -36,7 +36,7 @@ def _apply_hide(page, selectors: list[str]) -> None:
 
 
 def _apply_annotations(page, annotations: list[dict], settle_ms: int) -> None:
-    """Reuse scene handlers to draw static overlays before snapshotting."""
+
     if not annotations:
         return
     ensure_overlay(page)
@@ -49,8 +49,8 @@ def _apply_annotations(page, annotations: list[dict], settle_ms: int) -> None:
                 file=sys.stderr,
             )
             continue
-        # Scene handlers wait on `duration`; for static annotations we want
-        # 0-wait so they all stack up before the snapshot fires.
+
+
         ann_copy = dict(ann)
         ann_copy.setdefault("duration", 0)
         ann_copy.setdefault("holdMs", 0)
@@ -63,7 +63,7 @@ def _apply_annotations(page, annotations: list[dict], settle_ms: int) -> None:
 
 
 def _load_annotations(spec) -> tuple[list[dict], int]:
-    """Accept either a list, a dict with 'annotations' key, or a path to such."""
+
     if spec is None:
         return [], 0
     settle_ms = 900
@@ -93,7 +93,7 @@ def _build_mask_locators(page, selectors: list[str]):
 
 
 def do_screenshot(cfg: dict) -> Path:
-    """Take a screenshot per cfg. Returns the absolute output path."""
+
     ensure_runtime()
     from playwright.sync_api import sync_playwright
     from playwright.sync_api import TimeoutError as PlaywrightTimeoutError
@@ -126,14 +126,14 @@ def do_screenshot(cfg: dict) -> Path:
 
                 if cfg.get("selector"):
                     locator = page.locator(cfg["selector"]).first
-                    # 先用较短超时确认元素确实存在。否则选择器不匹配时，
-                    # scroll_into_view_if_needed() 会干等全局超时（默认 30s）才抛出
-                    # 一大段难懂的 Playwright traceback、以 exit 1 崩溃。这里快速失败，
-                    # 返回一条模型可读、可据以换选择器的错误信息。
+
+
+
+
                     probe_timeout = cfg.get("selectorTimeout")
                     if probe_timeout is None:
-                        # 页面 domcontentloaded 已触发，元素在则几乎立即命中；
-                        # 不在则无需等满 30s。取全局 timeout 与 8s 的较小值。
+
+
                         probe_timeout = min(cfg.get("timeout") or 8000, 8000)
                     try:
                         locator.wait_for(state="attached", timeout=probe_timeout)

package/skills/web-screenshot/scripts/_media_screenshot/storyboard.py

@@ -1,4 +1,4 @@
-"""Storyboard runner: orchestrate a sequence of scenes with smooth transitions."""
+
 from __future__ import annotations
 
 import sys
@@ -45,7 +45,7 @@ def _run_storyboard(page, cfg: dict) -> None:
 
 
 def do_storyboard(cfg: dict) -> tuple[Path, int, int]:
-    """Run a storyboard recording. Returns (output_path, start_ms, end_ms)."""
+
     ensure_runtime()
     from playwright.sync_api import sync_playwright
 

package/skills/web-screenshot/scripts/_media_screenshot/template.py

@@ -1,12 +1,12 @@
-"""Storyboard JSON template engine.
-
-Templates live in ``skills/media-screenshot/templates/<name>.json``.
-Each template is a regular storyboard JSON optionally prefixed with a ``_meta``
-block describing the variables it accepts. ``{{var_name}}`` placeholders inside
-string values are replaced; when a string consists of a single placeholder its
-substituted value is auto-coerced to int / float / bool / null so e.g.
-``"scale": "{{zoom_scale}}"`` becomes a number after rendering.
-"""
+
+
+
+
+
+
+
+
+
 from __future__ import annotations
 
 import json
@@ -71,7 +71,7 @@ def list_templates() -> list[Path]:
 
 
 def load_template(name: str) -> dict:
-    """Resolve a template name (or path) to its parsed JSON."""
+
     p = Path(name)
     if not p.exists():
         p = _TEMPLATES_DIR / f"{name}.json"
@@ -85,7 +85,7 @@ def load_template(name: str) -> dict:
 
 
 def render_template(name: str, params: dict[str, str]) -> dict:
-    """Load, validate against ``_meta.params``, render, return the storyboard."""
+
     tpl = load_template(name)
     meta = tpl.pop("_meta", {})
     declared = meta.get("params", {}) or {}

package/skills/web-screenshot/scripts/_media_screenshot/trim.py

@@ -1,11 +1,11 @@
-"""End-anchored ffmpeg trim for storyboard videos.
-
-Playwright's recorded video timeline does NOT align 1:1 with Python wall-clock —
-recording starts before the first navigation finishes and the file is typically
-a few seconds longer than the wall-clock window we measured. The storyboard
-ends right before ``context.close()``, which IS the video's tail. So we anchor
-on the END.
-"""
+
+
+
+
+
+
+
+
 from __future__ import annotations
 
 import os
@@ -48,11 +48,11 @@ def trim_video(out: Path, start_ms: int, end_ms: int) -> None:
 
     video_dur = _probe_duration(out)
     if video_dur is None:
-        # Fallback: start-anchored (legacy, may be off if recording drifts).
+
         start_s = max(0.0, (start_ms - lead_ms) / 1000.0)
         duration_s = keep_ms / 1000.0
     else:
-        # End-anchored: video tail == storyboard end == context.close.
+
         start_s = max(0.0, video_dur - keep_ms / 1000.0)
         duration_s = video_dur - start_s
 

package/skills/web-screenshot/scripts/_media_screenshot/urlguard.py

@@ -1,20 +1,20 @@
-"""Refuse to fetch private / link-local addresses (SSRF guard).
-
-Only ``read_page.py`` uses this today, and that asymmetry is deliberate:
-screenshot and record hand back a *file path*, while web_read hands the page's
-**text** straight to a model that is driven by user input. A prompt that says
-"read http://169.254.169.254/latest/meta-data/iam/..." is an exfiltration
-attempt, not a reading request.
-
-``WEB_CAPTURE_ALLOW_PRIVATE_HOSTS=1`` opts out — needed when pointing the skill
-at a dev server or an intranet page on purpose.
-
-Caveat: the check resolves DNS once and then hands the URL to the browser,
-which resolves it again. A name that flips between answers (DNS rebinding) can
-slip through that gap. Closing it properly needs request-level interception in
-Playwright; this guard is aimed at the ordinary case of a hostile or careless
-URL, not at an attacker who controls a nameserver.
-"""
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
+
 from __future__ import annotations
 
 import ipaddress
@@ -41,7 +41,7 @@ def _is_blocked(ip: ipaddress._BaseAddress) -> bool:
 
 
 def assert_public_url(raw_url: str) -> None:
-    """Raise SystemExit when *raw_url* is not an ordinary public http(s) page."""
+
     parsed = urlparse(raw_url)
     if parsed.scheme not in ("http", "https"):
         raise SystemExit(
@@ -54,7 +54,7 @@ def assert_public_url(raw_url: str) -> None:
     if allow_private_hosts():
         return
 
-    # A literal IP needs no DNS round-trip.
+
     try:
         literal = ipaddress.ip_address(host)
     except ValueError:
@@ -75,9 +75,9 @@ def assert_public_url(raw_url: str) -> None:
             ip = ipaddress.ip_address(addr)
         except ValueError:
             continue
-        # Any resolved address being internal is enough to refuse — a name that
-        # answers with one public and one private A record is the exact shape of
-        # an SSRF attempt.
+
+
+
         if _is_blocked(ip):
             raise SystemExit(_refusal(host, addr))