完善 MCP/会话上下文治理并补齐搜索与诊断能力。

补充历史压缩与会话诊断链路，强化 tool pairing 与空响应兜底观测；完善 Admin MCP 导入/绑定与相关前端展示；新增 web_search_fast/web_fetch_clean 及多项测试与文档更新，并同步技能安装与角色绑定策略改进。

Made-with: Cursor

runtime/tools/public/web_fetch_clean_tool.py

@@ -0,0 +1,158 @@
+from __future__ import annotations
+
+import re
+import urllib.parse
+import urllib.request
+from html import unescape
+from typing import Any
+
+from oclaw.runtime.tools.base import ToolSpec
+
+_UA = (
+    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) "
+    "AppleWebKit/537.36 (KHTML, like Gecko) Chrome/124.0 Safari/537.36"
+)
+
+
+def _http_get(url: str, *, timeout_s: int) -> tuple[int, dict[str, str], bytes]:
+    req = urllib.request.Request(
+        str(url),
+        headers={
+            "User-Agent": _UA,
+            "Accept": "text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8",
+            "Accept-Language": "zh-CN,zh;q=0.9,en;q=0.8",
+        },
+    )
+    try:
+        with urllib.request.urlopen(req, timeout=max(3, min(int(timeout_s), 45))) as resp:
+            code = int(getattr(resp, "status", 200) or 200)
+            headers = {str(k).lower(): str(v) for k, v in (getattr(resp, "headers", {}) or {}).items()}
+            body = resp.read()
+            return code, headers, body
+    except urllib.error.HTTPError as e:  # type: ignore[attr-defined]
+        try:
+            body = e.read() or b""
+        except Exception:
+            body = b""
+        headers = {str(k).lower(): str(v) for k, v in (getattr(e, "headers", {}) or {}).items()}
+        return int(getattr(e, "code", 0) or 0), headers, body
+
+
+def _strip_html_keep_lines(html: str) -> str:
+    s = str(html or "")
+    # Drop scripts/styles/noscript
+    s = re.sub(r"(?is)<(script|style|noscript)[^>]*>.*?</\1>", " ", s)
+    # Add line breaks for common block tags before stripping
+    s = re.sub(r"(?i)</(p|div|br|li|h1|h2|h3|h4|h5|h6|tr|section|article)>", "\n", s)
+    s = re.sub(r"(?i)<br\\s*/?>", "\n", s)
+    # Strip tags
+    s = re.sub(r"(?is)<[^>]+>", " ", s)
+    # Decode entities and normalize whitespace
+    s = unescape(s)
+    s = re.sub(r"[\t\r\f\v]+", " ", s)
+    s = re.sub(r" *\\n *", "\n", s)
+    s = re.sub(r"\\n{3,}", "\n\n", s)
+    s = re.sub(r" {2,}", " ", s)
+    return s.strip()
+
+
+def _extract_title(html: str) -> str:
+    m = re.search(r"(?is)<title[^>]*>(.*?)</title>", str(html or ""))
+    if not m:
+        return ""
+    t = _strip_html_keep_lines(m.group(1))
+    return t.splitlines()[0].strip() if t else ""
+
+
+def web_fetch_clean_tool() -> ToolSpec:
+    def _handler(args: dict[str, Any]) -> dict[str, Any]:
+        payload = args if isinstance(args, dict) else {}
+        url = str(payload.get("url") or "").strip()
+        if not url:
+            return {"ok": False, "error_code": "url_required", "error": "url_required"}
+        timeout_s = max(3, min(int(payload.get("timeout") or 20), 45))
+        max_chars = max(500, min(int(payload.get("max_chars") or 18000), 200000))
+
+        try:
+            parsed = urllib.parse.urlparse(url)
+        except Exception:
+            return {"ok": False, "error_code": "invalid_url", "error": "invalid_url"}
+        if parsed.scheme not in {"http", "https"}:
+            return {"ok": False, "error_code": "unsupported_scheme", "error": "unsupported_scheme"}
+
+        try:
+            code, headers, body = _http_get(url, timeout_s=timeout_s)
+        except Exception as exc:
+            return {"ok": False, "error_code": "fetch_failed", "error": f"fetch_failed:{type(exc).__name__}"}
+
+        if int(code) < 200 or int(code) >= 300:
+            return {
+                "ok": False,
+                "error_code": "http_error",
+                "error": "http_error",
+                "status_code": int(code),
+                "content_type": str(headers.get("content-type") or ""),
+            }
+
+        ctype = str(headers.get("content-type") or "").lower()
+        if ctype and ("text/html" not in ctype and "application/xhtml" not in ctype):
+            # Still try to decode as text, but label as non_html for caller decisions.
+            try:
+                text = body.decode("utf-8", errors="replace")
+            except Exception:
+                text = ""
+            text = text.strip()
+            if len(text) > max_chars:
+                text = text[:max_chars]
+            return {
+                "ok": True,
+                "url": url,
+                "status_code": int(code),
+                "content_type": ctype,
+                "title": "",
+                "text": text,
+                "truncated": len(body) > max_chars,
+                "mode": "raw_non_html",
+            }
+
+        html = body.decode("utf-8", errors="replace")
+        title = _extract_title(html)
+        text = _strip_html_keep_lines(html)
+        truncated = False
+        if len(text) > max_chars:
+            text = text[:max_chars]
+            truncated = True
+        return {
+            "ok": True,
+            "url": url,
+            "status_code": int(code),
+            "content_type": ctype or "text/html",
+            "title": title,
+            "text": text,
+            "truncated": truncated,
+            "mode": "html_strip",
+        }
+
+    return ToolSpec(
+        name="web_fetch_clean",
+        description="Fetch a web page and return cleaned main text (simple HTML stripping).",
+        parameters={
+            "type": "object",
+            "properties": {
+                "url": {"type": "string", "description": "HTTP/HTTPS URL to fetch."},
+                "timeout": {"type": "integer", "default": 20, "description": "Timeout seconds [3..45]."},
+                "max_chars": {"type": "integer", "default": 18000, "description": "Max returned text chars [500..200000]."},
+            },
+            "required": ["url"],
+            "additionalProperties": False,
+        },
+        handler=_handler,
+        tags=frozenset({"public", "web", "fetch", "read"}),
+        risk_level="low",
+        read_only=True,
+        timeout_s=55.0,
+    )
+
+
+__all__ = ["web_fetch_clean_tool"]
+