dw-mcp 0.1.0__tar.gz

dw_mcp-0.1.0/.github/workflows/publish.yml

@@ -0,0 +1,20 @@
+name: Publish to PyPI
+
+on:
+  push:
+    tags: ["v*"]
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write # required for PyPI trusted publishing (OIDC)
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - run: pip install build
+      - run: python -m build
+      - uses: pypa/gh-action-pypi-publish@release/v1

dw_mcp-0.1.0/.github/workflows/registry.yml

@@ -0,0 +1,61 @@
+name: Publish to MCP Registry
+
+# Official MCP Registry (registry.modelcontextprotocol.io) via GitHub OIDC.
+# No secret, no interactive login: the runner's OIDC token proves ownership
+# of the io.github.mohammad-abbasi-sh/* namespace.
+# Refs: https://github.com/modelcontextprotocol/registry
+
+on:
+  push:
+    tags: ["v*"]
+  workflow_dispatch:
+
+jobs:
+  publish:
+    runs-on: ubuntu-latest
+    permissions:
+      id-token: write # required for OIDC -> MCP Registry
+      contents: read
+    steps:
+      - uses: actions/checkout@v4
+
+      - name: Sync server.json version to tag
+        # Only sync when running on a version tag (v*). On workflow_dispatch
+        # (ref = refs/heads/main) keep the version committed in server.json.
+        run: |
+          VER=""
+          if [[ "$GITHUB_REF" == refs/tags/* ]]; then
+            TAG="${GITHUB_REF#refs/tags/}"
+            VER="${TAG#v}"
+          fi
+          if [ -n "$VER" ]; then
+            python3 - "$VER" << 'EOF'
+          import json, sys
+          ver = sys.argv[1]
+          d = json.load(open('server.json'))
+          d['version'] = ver
+          for pkg in d.get('packages', []):
+              pkg['version'] = ver
+          json.dump(d, open('server.json', 'w'), indent=2, ensure_ascii=False)
+          print('server.json version ->', ver)
+          EOF
+          else
+            echo "No version tag ($GITHUB_REF); keeping server.json version as-is"
+          fi
+
+      - name: Install mcp-publisher
+        run: |
+          curl -sL "https://github.com/modelcontextprotocol/registry/releases/download/v1.8.7/mcp-publisher_linux_amd64.tar.gz" -o /tmp/mcp-publisher.tar.gz
+          tar -xzf /tmp/mcp-publisher.tar.gz -C /tmp
+          sudo mv /tmp/mcp-publisher /usr/local/bin/mcp-publisher
+          mcp-publisher --version
+
+      - name: Validate server.json
+        run: mcp-publisher validate server.json
+
+      - name: Publish to MCP Registry (OIDC)
+        run: |
+          mcp-publisher login github-oidc
+          mcp-publisher publish server.json
+        # A registry outage must never break the release; re-run from Actions if needed.
+        continue-on-error: true

dw_mcp-0.1.0/.github/workflows/smoke.yml

@@ -0,0 +1,21 @@
+name: Daily live smoke
+
+# One fixed probe per day: catches fingerprint rot / bot-wall changes
+# before users hit them. Fails visibly on 403/timeout.
+on:
+  schedule:
+    - cron: "0 6 * * *"
+  workflow_dispatch:
+
+jobs:
+  smoke:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: actions/setup-python@v5
+        with:
+          python-version: "3.12"
+      - name: Install
+        run: pip install -e .
+      - name: Smoke probe
+        run: python tests/smoke_live.py

dw_mcp-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Mohammad Abbasi
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

dw_mcp-0.1.0/PKG-INFO

@@ -0,0 +1,72 @@
+Metadata-Version: 2.5
+Name: dw-mcp
+Version: 0.1.0
+Summary: Unofficial read-only MCP server for Deutsche Welle news (dw.com) — headlines, search, articles & topics in 31 languages
+Author: Mohammad Abbasi
+License: MIT
+License-File: LICENSE
+Requires-Python: >=3.10
+Requires-Dist: httpx>=0.27
+Requires-Dist: mcp<2
+Description-Content-Type: text/markdown
+
+# dw-mcp 📰
+
+mcp-name: io.github.mohammad-abbasi-sh/dw-mcp
+
+Unofficial **read-only** MCP server for **[dw.com](https://www.dw.com)** — Deutsche Welle news in 31 languages from DW's official sitemaps: latest headlines, keyword search over recent headlines, latest articles, full article text and topic pages. No key, no login. No key, no login.
+
+> Built by Mohammad Abbasi as part of the "Germany MCP factory". Not affiliated with dw.com.
+
+## For AI agents
+
+Add to your MCP client config (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dw": {
+      "command": "uvx",
+      "args": ["dw-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+Or run locally: `python -m dw_mcp.server` (needs `mcp<2`, `httpx`).
+
+## Tools
+
+| Tool | Description |
+|------|-------------|
+| `dw_headlines` | Latest DW headlines — `language` (e.g. "en", "de", "fa-ir"), `limit` 1..50 |
+| `dw_search_news` | Keyword search over the latest DW headlines — `query`, `language`, `limit` 1..50 |
+| `dw_latest_articles` | Latest DW articles by publication time — `language`, `limit` 1..50 |
+| `dw_get_article` | Full text of one DW article — `url` (a dw.com article link) |
+| `dw_topics` | DW topic pages matching a word — `query`, `language`, `limit` 1..50 |
+| `dw_health` | Reachability probe (ok / blocked / error) — one polite request |
+
+## Notes
+
+- All data comes from **DW's official sitemaps** (free, no key, no login).
+- DW has **no working RSS feeds any more** — every `/rss/` path returns 404 (verified 2026-11-10) — and its site search is gone, so headlines come from the news sitemaps and `dw_search_news` filters the latest ~50 headlines (not the full archive).
+- Persian is two editions: `fa-ir` (Iran) and `fa-af` (Afghanistan/Dari); plain `fa` does not exist.
+- Article bodies are extracted from DW's server-rendered pages (JSON-LD headline/date + cleaned `<p>` text); live blogs and video pages may return little or no body text.
+
+Tests: 22/22 green (15 offline + 7 live, 2026-11-10).
+
+## Development
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## License
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dw_mcp-0.1.0/README-de.md

@@ -0,0 +1,51 @@
+# dw-mcp 📰
+
+mcp-name: io.github.mohammad-abbasi-sh/dw-mcp
+
+Inoffizieller **nur lesender** MCP-Server für **[dw.com](https://www.dw.com)** — Deutsche-Welle-Nachrichten in 31 Sprachen über die offiziellen DW-Sitemaps: neueste Schlagzeilen, Stichwortsuche, neueste Artikel, Volltexte und Themenseiten. Kein Schlüssel, kein Login. Kein Schlüssel, keine Anmeldung.
+
+> Gebaut von Mohammad Abbasi als Teil der "Germany MCP factory". Nicht mit dw.com verbunden.
+
+## Für KI-Agenten
+
+Zur MCP-Client-Konfiguration hinzufügen (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dw": {
+      "command": "uvx",
+      "args": ["dw-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+Oder lokal starten: `python -m dw_mcp.server` (benötigt `mcp<2`, `httpx`).
+
+## Tools
+
+| Tool | Beschreibung |
+|------|-------------|
+| `dw_headlines` | Neueste DW-Schlagzeilen in einer Sprache |
+| `dw_search_news` | Neueste DW-Schlagzeilen nach Stichwort suchen |
+| `dw_latest_articles` | Neueste DW-Artikel in einer Sprache |
+| `dw_get_article` | Volltext eines DW-Artikels |
+| `dw_topics` | DW-Themenseiten zu einem Stichwort |
+| `dw_health` | Erreichbarkeitsprüfung (ok / blocked / error) — eine höfliche Anfrage |
+
+## Entwicklung
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## Lizenz
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dw_mcp-0.1.0/README-fa.md

@@ -0,0 +1,51 @@
+# dw-mcp 📰
+
+mcp-name: io.github.mohammad-abbasi-sh/dw-mcp
+
+سرور MCP غیررسمی و **فقط‌خواندنی** برای **[dw.com](https://www.dw.com)** — اخبار دویچه‌وله به ۳۱ زبان از سایت‌مپ‌های رسمی DW: تازه‌ترین تیترها، جست‌وجوی کلیدواژه، تازه‌ترین مقاله‌ها، متن کامل مقاله و صفحه‌های موضوعی. بدون کلید، بدون ورود. بدون کلید، بدون ورود.
+
+> ساختهٔ محمد عباسی، بخشی از «کارخانهٔ MCP آلمان». وابسته به dw.com نیست.
+
+## برای ایجنت‌های هوش مصنوعی
+
+به پیکربندی MCP کلاینت اضافه کنید (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dw": {
+      "command": "uvx",
+      "args": ["dw-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+یا به‌صورت محلی اجرا کنید: `python -m dw_mcp.server` (نیازمند `mcp<2` و `httpx`).
+
+## ابزارها
+
+| ابزار | توضیح |
+|------|-------|
+| `dw_headlines` | تازه‌ترین تیترهای DW به یک زبان |
+| `dw_search_news` | جست‌وجوی تیترهای اخیر DW با کلیدواژه |
+| `dw_latest_articles` | تازه‌ترین مقاله‌های DW به یک زبان |
+| `dw_get_article` | متن کامل یک مقالهٔ DW |
+| `dw_topics` | صفحه‌های موضوعی DW برای یک واژه |
+| `dw_health` | بررسی دسترس‌پذیری (ok / blocked / error) — یک درخواست مؤدبانه |
+
+## توسعه
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## لایسنس
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dw_mcp-0.1.0/README.md

@@ -0,0 +1,60 @@
+# dw-mcp 📰
+
+mcp-name: io.github.mohammad-abbasi-sh/dw-mcp
+
+Unofficial **read-only** MCP server for **[dw.com](https://www.dw.com)** — Deutsche Welle news in 31 languages from DW's official sitemaps: latest headlines, keyword search over recent headlines, latest articles, full article text and topic pages. No key, no login. No key, no login.
+
+> Built by Mohammad Abbasi as part of the "Germany MCP factory". Not affiliated with dw.com.
+
+## For AI agents
+
+Add to your MCP client config (stdio):
+
+```json
+{
+  "mcpServers": {
+    "dw": {
+      "command": "uvx",
+      "args": ["dw-mcp==0.1.0"]
+    }
+  }
+}
+```
+
+Or run locally: `python -m dw_mcp.server` (needs `mcp<2`, `httpx`).
+
+## Tools
+
+| Tool | Description |
+|------|-------------|
+| `dw_headlines` | Latest DW headlines — `language` (e.g. "en", "de", "fa-ir"), `limit` 1..50 |
+| `dw_search_news` | Keyword search over the latest DW headlines — `query`, `language`, `limit` 1..50 |
+| `dw_latest_articles` | Latest DW articles by publication time — `language`, `limit` 1..50 |
+| `dw_get_article` | Full text of one DW article — `url` (a dw.com article link) |
+| `dw_topics` | DW topic pages matching a word — `query`, `language`, `limit` 1..50 |
+| `dw_health` | Reachability probe (ok / blocked / error) — one polite request |
+
+## Notes
+
+- All data comes from **DW's official sitemaps** (free, no key, no login).
+- DW has **no working RSS feeds any more** — every `/rss/` path returns 404 (verified 2026-11-10) — and its site search is gone, so headlines come from the news sitemaps and `dw_search_news` filters the latest ~50 headlines (not the full archive).
+- Persian is two editions: `fa-ir` (Iran) and `fa-af` (Afghanistan/Dari); plain `fa` does not exist.
+- Article bodies are extracted from DW's server-rendered pages (JSON-LD headline/date + cleaned `<p>` text); live blogs and video pages may return little or no body text.
+
+Tests: 22/22 green (15 offline + 7 live, 2026-11-10).
+
+## Development
+
+```bash
+python3 -m venv .venv && .venv/bin/pip install "mcp<2" httpx pytest
+.venv/bin/pip install -e .
+.venv/bin/python -m pytest tests/ -q
+```
+
+## License
+
+MIT
+
+---
+
+Built by [Mohammad Abbasi](https://github.com/mohammad-abbasi-sh) — [Almanibama](https://almanibama.ir) · [mabbasi.ir](https://mabbasi.ir)

dw_mcp-0.1.0/pyproject.toml

@@ -0,0 +1,25 @@
+[build-system]
+requires = ["hatchling"]
+build-backend = "hatchling.build"
+
+[project]
+name = "dw-mcp"
+version = "0.1.0"
+description = "Unofficial read-only MCP server for Deutsche Welle news (dw.com) — headlines, search, articles & topics in 31 languages"
+readme = "README.md"
+license = { text = "MIT" }
+requires-python = ">=3.10"
+dependencies = [
+    "mcp<2",
+    "httpx>=0.27",
+]
+authors = [{ name = "Mohammad Abbasi" }]
+
+[project.scripts]
+dw-mcp = "dw_mcp.server:main"
+
+[tool.hatch.build.targets.wheel]
+packages = ["src/dw_mcp"]
+
+[tool.pytest.ini_options]
+markers = ["live: tests that hit the real https://www.dw.com site"]

dw_mcp-0.1.0/server.json

@@ -0,0 +1 @@
+{"$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json", "description": "Unofficial read-only MCP server for Deutsche Welle news — headlines, search & articles.", "name": "io.github.mohammad-abbasi-sh/dw-mcp", "packages": [{"identifier": "dw-mcp", "registryBaseUrl": "https://pypi.org", "registryType": "pypi", "transport": {"type": "stdio"}, "version": "0.1.0"}], "repository": {"source": "github", "url": "https://github.com/mohammad-abbasi-sh/dw-mcp"}, "title": "Deutsche Welle — DW news in 31 languages: headlines, search, articles & topics", "version": "0.1.0", "websiteUrl": "https://github.com/mohammad-abbasi-sh/dw-mcp"}

dw_mcp-0.1.0/src/dw_mcp/__init__.py

@@ -0,0 +1,7 @@
+"""dw_mcp — unofficial read-only MCP server for dw.com.
+
+Deutsche Welle news in 31 languages from DW's official sitemaps: latest headlines, keyword search over recent headlines, latest articles, full article text and topic pages. No key, no login.
+"""
+from __future__ import annotations
+
+__version__ = "0.1.0"

dw_mcp-0.1.0/src/dw_mcp/domain.py

@@ -0,0 +1,157 @@
+"""Domain logic: shape Deutsche Welle sitemap XML + article HTML into
+compact dicts (stdlib only).
+
+Data source: DW's official sitemaps
+(https://www.dw.com/{lang}/news-sitemap.xml,
+/{lang}/article-sitemap.xml, /{lang}/auto-topic-sitemap.xml) and
+dw.com article pages — free, no key, no login.
+
+DW has NO working RSS feeds any more (all /rss/ paths 404, verified
+2026-11-10) and its site search is gone too, so `dw_search_news`
+filters the latest news-sitemap headlines by keyword — it searches
+the ~last 50 headlines, not the full archive, and says so.
+"""
+from __future__ import annotations
+
+import html as _html
+import json
+import re
+import xml.etree.ElementTree as ET
+
+BASE = "https://www.dw.com"
+
+# Language editions with a news sitemap (from the DW sitemap index,
+# verified 2026-11-10). Note: Persian is "fa-ir" (Iran) and "fa-af"
+# (Afghanistan/Dari) — plain "fa" 400s.
+LANGUAGES = {
+    "am", "ar", "bg", "bn", "bs", "de", "el", "en", "es", "fa-af",
+    "fa-ir", "fr", "ha", "hi", "hr", "id", "mk", "pl", "ps",
+    "pt-002", "pt-br", "ro", "ru", "sq", "sr", "sw", "tr", "uk",
+    "ur", "zh", "zh-hant",
+}
+
+_NS = {
+    "s": "http://www.sitemaps.org/schemas/sitemap/0.9",
+    "n": "http://www.google.com/schemas/sitemap-news/0.9",
+}
+
+_LD_RE = re.compile(r'<script[^>]*application/ld\+json[^>]*>(.*?)</script>', re.S)
+_P_RE = re.compile(r'<p[^>]*>(.*?)</p>', re.S)
+_TAG_RE = re.compile(r'<[^>]+>')
+_SHORT_URL_RE = re.compile(r'^https?://p\.dw\.com/p/\S+?Prefer DW on Google\s*')
+_SHORT_URL_BARE_RE = re.compile(r'^https?://p\.dw\.com/p/\S+\s+')
+
+
+def validate_language(language: str) -> str:
+    key = (language or "").strip().lower()
+    if key not in LANGUAGES:
+        raise ValueError(
+            f"unsupported language: {language!r} "
+            f"(use one of: {', '.join(sorted(LANGUAGES))})"
+        )
+    return key
+
+
+def validate_limit(limit: int) -> int:
+    if not 1 <= limit <= 50:
+        raise ValueError(f"limit out of range: {limit} (expected 1..50)")
+    return limit
+
+
+def news_sitemap_url(language: str) -> str:
+    return f"{BASE}/{language}/news-sitemap.xml"
+
+
+def article_sitemap_url(language: str) -> str:
+    return f"{BASE}/{language}/article-sitemap.xml"
+
+
+def topic_sitemap_url(language: str) -> str:
+    return f"{BASE}/{language}/auto-topic-sitemap.xml"
+
+
+def _parse_xml(text: str, what: str) -> ET.Element:
+    try:
+        return ET.fromstring(text)
+    except ET.ParseError as e:
+        raise ValueError(f"invalid {what}: {e}") from e
+
+
+def parse_news_sitemap(text: str) -> list[dict]:
+    """[{title, url, published}] — newest first (sitemap order)."""
+    root = _parse_xml(text, "news sitemap")
+    out = []
+    for url in root.findall("s:url", _NS):
+        news = url.find("n:news", _NS)
+        if news is None:
+            continue
+        out.append({
+            "title": news.findtext("n:title", namespaces=_NS) or "",
+            "url": url.findtext("s:loc", namespaces=_NS) or "",
+            "published": news.findtext("n:publication_date", namespaces=_NS) or "",
+        })
+    return out
+
+
+def parse_article_sitemap(text: str) -> list[dict]:
+    """[{url, lastmod}] — sorted newest first by lastmod."""
+    root = _parse_xml(text, "article sitemap")
+    out = []
+    for url in root.findall("s:url", _NS):
+        out.append({
+            "url": url.findtext("s:loc", namespaces=_NS) or "",
+            "lastmod": url.findtext("s:lastmod", namespaces=_NS) or "",
+        })
+    out.sort(key=lambda e: e["lastmod"], reverse=True)
+    return out
+
+
+def parse_topic_sitemap(text: str) -> list[dict]:
+    """[{name, url, lastmod}] — name derived from the topic slug."""
+    root = _parse_xml(text, "topic sitemap")
+    out = []
+    for url in root.findall("s:url", _NS):
+        loc = url.findtext("s:loc", namespaces=_NS) or ""
+        slug = loc.rstrip("/").split("/")[-2] if "/t-" in loc else ""
+        name = slug.replace("-", " ").strip().title() if slug else ""
+        out.append({
+            "name": name,
+            "url": loc,
+            "lastmod": url.findtext("s:lastmod", namespaces=_NS) or "",
+        })
+    out.sort(key=lambda e: e["lastmod"], reverse=True)
+    return out
+
+
+def parse_article(html_text: str) -> dict:
+    """Extract headline/date (JSON-LD NewsArticle) + cleaned body <p>s.
+
+    DW article pages are server-rendered: the body is plain <p> tags.
+    The first <p> is navigation junk ("Skip to main menu...") and is
+    dropped; the intro paragraph carries a leading p.dw.com short URL
+    + "Prefer DW on Google" prefix which is stripped.
+    """
+    headline, published = "", ""
+    m = _LD_RE.search(html_text)
+    if m:
+        try:
+            ld = json.loads(m.group(1))
+            if isinstance(ld, dict) and ld.get("@type") == "NewsArticle":
+                headline = ld.get("headline") or ""
+                published = ld.get("datePublished") or ""
+        except json.JSONDecodeError:
+            pass
+    paras = []
+    for raw in _P_RE.findall(html_text):
+        txt = _html.unescape(_TAG_RE.sub("", raw)).replace("\u00a0", " ").strip()
+        if not txt:
+            continue
+        if txt.startswith("Skip to main menu"):
+            continue
+        txt = _SHORT_URL_RE.sub("", txt)
+        txt = _SHORT_URL_BARE_RE.sub("", txt)
+        txt = txt.strip()
+        if txt:
+            paras.append(txt)
+    body = "\n\n".join(paras)
+    return {"headline": headline, "published": published, "body": body}

dw_mcp-0.1.0/src/dw_mcp/http.py

@@ -0,0 +1,99 @@
+"""HTTP layer for dw_mcp — plain httpx GETs against dw.com.
+
+No key, no login. One retry on 429/5xx; error taxonomy (403/429/other).
+"""
+from __future__ import annotations
+
+import os
+import ssl
+import time
+
+import httpx
+
+UA = (
+    "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 "
+    "(KHTML, like Gecko) Chrome/131.0.0.0 Safari/537.36"
+)
+TIMEOUT = 25.0
+PROBE_URL = "https://www.dw.com/en/news-sitemap.xml"
+SOURCE = "dw.com"
+
+
+class DwMcpError(RuntimeError):
+    pass
+
+
+def _verify():
+    """TLS trust store: the OS bundle when present, else httpx/certifi."""
+    for path in (
+        "/etc/ssl/certs/ca-certificates.crt",
+        ssl.get_default_verify_paths().openssl_cafile,
+    ):
+        if path and os.path.exists(path):
+            return ssl.create_default_context(cafile=path)
+    return True
+
+
+def _client() -> httpx.Client:
+    # NOTE: we read the proxy URL from the environment but never log it.
+    proxy = os.environ.get("https_proxy") or os.environ.get("HTTPS_PROXY")
+    return httpx.Client(
+        headers={"User-Agent": UA, "Accept": "application/json, application/xml, text/html, */*"},
+        timeout=TIMEOUT,
+        proxy=proxy or None,
+        trust_env=False,
+        verify=_verify(),
+    )
+
+
+def fetch_text(url: str) -> str:
+    """GET one URL and return its text. One retry on 429/5xx."""
+    last_exc: Exception | None = None
+    r: httpx.Response | None = None
+    for attempt in range(2):
+        try:
+            with _client() as c:
+                r = c.get(url)
+        except httpx.HTTPError as e:
+            last_exc = e
+            if attempt == 0:
+                time.sleep(1.0)
+                continue
+            raise DwMcpError(f"network error fetching {url}: {e}") from e
+        if r.status_code in (429, 500, 502, 503):
+            if attempt == 0:
+                time.sleep(1.0)
+                continue
+        break
+    assert r is not None  # loop always assigns r or raises
+    if r.status_code == 403:
+        raise DwMcpError(
+            f"{SOURCE} returned HTTP 403 for {url} (bot wall) — try again later"
+        )
+    if r.status_code == 429:
+        raise DwMcpError(
+            f"{SOURCE} rate-limited the request (HTTP 429) — try again later"
+        )
+    if r.status_code != 200:
+        raise DwMcpError(f"{SOURCE} returned HTTP {r.status_code} for {url}")
+    return r.text
+
+
+def health_check() -> dict:
+    """One polite probe of the source site.
+
+    Returns {"status": "ok"|"blocked"|"error", "http_status": int|None,
+    "url": ..., "detail": ...}. Used by the `<prefix>_health` tool
+    and the daily smoke CI.
+    """
+    url = PROBE_URL
+    try:
+        fetch_text(url)
+    except DwMcpError as e:
+        msg = str(e)
+        if "403" in msg:
+            return {"status": "blocked", "http_status": 403, "url": url, "detail": msg}
+        if "429" in msg:
+            return {"status": "blocked", "http_status": 429, "url": url, "detail": msg}
+        return {"status": "error", "http_status": None, "url": url, "detail": msg}
+    return {"status": "ok", "http_status": 200, "url": url}