backend: add shared HTTP session factory, use it for the song intro

Introduce core/http_client.py with HTTP_USER_AGENT and make_session(), a
single home for the outbound User-Agent that's currently copy-pasted into
~20 files. song_intro.py now builds its session via make_session() instead
of hardcoding the UA string. A follow-up PR will sweep the remaining call
sites onto the factory.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

Author: dprodger

backend/core/http_client.py

@@ -0,0 +1,34 @@
+"""Shared HTTP client configuration.
+
+A single home for the outbound User-Agent and a `requests.Session` factory,
+so every crawler/integration identifies us the same way and a version bump
+is a one-line change rather than a sweep across ~20 files.
+
+Most external services (MusicBrainz, Wikipedia/MediaWiki, Cover Art Archive,
+Wikimedia Commons, etc.) expect — and in some cases require — a descriptive
+User-Agent. Use `make_session()` to get a session that already carries it.
+
+Note: this only handles identification/headers, not per-service rate
+limiting. Clients that must throttle (e.g. MusicBrainz) keep their own
+rate-limit logic on top of the session.
+"""
+
+import requests
+
+# Outbound identity sent on every API/crawl request. Bump the version here.
+HTTP_USER_AGENT = "ApproachNote/1.0 (+support@approachnote.com)"
+
+
+def make_session(accept_json: bool = True) -> requests.Session:
+    """Return a requests.Session preconfigured with our User-Agent.
+
+    Args:
+        accept_json: Also set ``Accept: application/json`` (the common case
+            for the JSON APIs we call). Pass False for HTML/binary fetches.
+    """
+    session = requests.Session()
+    headers = {'User-Agent': HTTP_USER_AGENT}
+    if accept_json:
+        headers['Accept'] = 'application/json'
+    session.headers.update(headers)
+    return session

backend/integrations/wikipedia/song_intro.py

@@ -21,10 +21,10 @@
 
 import requests
 
+from core.http_client import make_session
+
 logger = logging.getLogger(__name__)
 
-# MediaWiki asks API clients to send a descriptive User-Agent.
-USER_AGENT = "ApproachNote/1.0 (+support@approachnote.com)"
 DEFAULT_SENTENCES = 4
 REQUEST_TIMEOUT = 15
 
@@ -58,7 +58,7 @@ def fetch_wikipedia_intro(page_title: str, api_url: str,
     request fails. Raises nothing for HTTP-level non-200s (logs + returns
     None); network exceptions propagate to the caller.
     """
-    sess = session or requests.Session()
+    sess = session or make_session()
     params = {
         'action': 'query',
         'format': 'json',
@@ -69,8 +69,7 @@ def fetch_wikipedia_intro(page_title: str, api_url: str,
         'explaintext': 1,
         'exsentences': sentences,
     }
-    headers = {'User-Agent': USER_AGENT, 'Accept': 'application/json'}
-    resp = sess.get(api_url, params=params, headers=headers, timeout=REQUEST_TIMEOUT)
+    resp = sess.get(api_url, params=params, timeout=REQUEST_TIMEOUT)
     if resp.status_code != 200:
         logger.warning("Wikipedia returned status %s for %s", resp.status_code, page_title)
         return None