Track processed candidates in session and use OAuth for API calls

- Record skips, saves, and no-match results in session["skipped"] so
  revisiting an article resumes past already-checked candidates
- Filter self-links (case-insensitive first letter) from hit list
- Use OAuth session for all API reads when logged in for higher rate limits
- Add "for" template to exclusion list to avoid bad edits
- Improve API error handling with HTTP status codes logged to stderr

Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>

add_links/api.py

@@ -1,4 +1,5 @@
 import re
+import sys
 import typing
 
 import requests
@@ -72,18 +73,30 @@ webpage_error = (
 )
 
 
+def _get_active_session() -> requests.sessions.Session:
+    """Return OAuth session if one is available in Flask context, else plain session."""
+    try:
+        from flask import g
+        if hasattr(g, "oauth_session") and g.oauth_session is not None:
+            return g.oauth_session  # type: ignore[return-value]
+    except RuntimeError:
+        pass
+    return get_session()
+
+
 def api_get(params: StrDict) -> StrDict:
     """Make call to Wikipedia API."""
-    s = get_session()
+    s = _get_active_session()
 
     r = s.get(get_query_url(), params=params)
     try:
         ret: StrDict = r.json()
     except JSONDecodeError:
+        print(f"API request failed: HTTP {r.status_code}", file=sys.stderr)
+        print(f"Response body: {r.text!r}", file=sys.stderr)
         if webpage_error in r.text:
             raise MediawikiError(webpage_error)
-        else:
+        raise MediawikiError(f"HTTP {r.status_code}: unexpected response from Wikipedia API")
-            raise MediawikiError("unknown error")
     check_for_error(ret)
     return ret
 
@@ -271,7 +284,7 @@ def call_get_diff(title: str, section_num: int, section_text: str) -> str:
         "rvdifftotext": section_text.strip(),
     }
 
-    s = get_session()
+    s = _get_active_session()
     r = s.post(get_query_url(), data=data)
     try:
         ret = r.json()

add_links/match.py

@@ -78,7 +78,7 @@ re_cite = re.compile(
     re.I | re.S,
 )
 
-re_cite_template_start = re.compile(r"\{\{(?:cite|citation|short description|gli|defn|annotated link|excerpt|main|see)\b", re.I)
+re_cite_template_start = re.compile(r"\{\{(?:cite|citation|short description|gli|defn|annotated link|excerpt|main|see|for)\b", re.I)
 re_no_param_template = re.compile(r"\{\{[^|{}]+\}\}")
 re_external_link = re.compile(r"\[https?://[^\]]+\]")
 # Italic text (work titles in bibliographies). Handles apostrophes in content

add_links/mediawiki_oauth.py

@@ -99,6 +99,27 @@ def userinfo_call() -> typing.Mapping[str, typing.Any]:
     return api_request(params)
 
 
+def get_oauth_session() -> OAuth1Session | None:
+    """Return an OAuth1Session for the current user, or None if not logged in."""
+    if "owner_key" not in session or "owner_secret" not in session:
+        return None
+    app = current_app
+    client_key = app.config["CLIENT_KEY"]
+    client_secret = app.config["CLIENT_SECRET"]
+    oauth = OAuth1Session(
+        client_key,
+        client_secret=client_secret,
+        resource_owner_key=session["owner_key"],
+        resource_owner_secret=session["owner_secret"],
+    )
+    oauth.headers.update({"User-Agent": ua})
+    oauth.params = typing.cast(
+        dict[str, str | int],
+        {"format": "json", "action": "query", "formatversion": 2},
+    )
+    return oauth
+
+
 def get_username() -> None | str:
     """Get the username or None if not logged in."""
     if "owner_key" not in session:

web_view.py

@@ -111,6 +111,7 @@ def search_no_link(q: str) -> tuple[int, list[Hit]]:
 def global_user() -> None:
     """Make username available everywhere."""
     flask.g.user = mediawiki_oauth.get_username()
+    flask.g.oauth_session = mediawiki_oauth.get_oauth_session()
 
 
 @app.route("/")
@@ -294,6 +295,16 @@ def match_type(q: str, snippet: str) -> str | None:
 
 
 
+def _record_skip(from_title: str, hit_title: str) -> None:
+    """Record that a candidate was skipped or saved for this article."""
+    skipped: dict[str, list[str]] = flask.session.get("skipped", {})
+    article_skipped = skipped.get(from_title, [])
+    if hit_title not in article_skipped:
+        skipped[from_title] = article_skipped + [hit_title]
+        flask.session["skipped"] = skipped
+        flask.session.modified = True
+
+
 def handle_post(url_title: str) -> Response:
     """Handle POST request."""
     from_title = url_title.replace("_", " ").strip()
@@ -305,6 +316,7 @@ def handle_post(url_title: str) -> Response:
     except mediawiki_api.APIError as e:
         return flask.make_response(f"Save failed: {e}", 502)
     flask.session["saves"] = flask.session.get("saves", 0) + 1
+    _record_skip(from_title, hit_title)
     return flask.redirect(
         flask.url_for("article_page", url_title=url_title, after=hit_title)
     )
@@ -318,9 +330,19 @@ def article_page(url_title: str) -> str | Response:
 
     from_title = url_title.replace("_", " ").strip()
 
+    try:
         total = search_count(from_title)
         with_link = search_count_with_link(from_title)
         _no_link_count, hits = search_no_link(from_title)
+    except api.MediawikiError as e:
+        return flask.make_response(
+            flask.render_template("error.html", message=str(e)), 502
+        )
+
+    # Filter out candidates already processed this session
+    session_skipped: set[str] = set(
+        flask.session.get("skipped", {}).get(from_title, [])
+    )
 
     # If a specific candidate was requested, move it to the front
     title_param = flask.request.args.get("title")
@@ -328,12 +350,14 @@ def article_page(url_title: str) -> str | Response:
         hits = [h for h in hits if h["title"] == title_param] + \
                [h for h in hits if h["title"] != title_param]
 
-    # Skip past already-processed candidates
+    # Record and apply explicit skip-past
     after = flask.request.args.get("after")
     if after:
-        hits_iter = itertools.dropwhile(lambda h: h["title"] != after, hits)
+        _record_skip(from_title, after)
-        next(hits_iter, None)  # consume the "after" entry itself
+        session_skipped.add(after)
-        hits = list(hits_iter)
+
+    hits = [h for h in hits if h["title"] not in session_skipped
+            and h["title"] != from_title and h["title"] != case_flip_first(from_title)]
 
     if not hits:
         return flask.render_template("all_done.html")
@@ -397,7 +421,10 @@ def api_valid_hit() -> werkzeug.wrappers.response.Response:
     try:
         found = get_diff(link_to, link_from, None)
     except NoMatch:
+        _record_skip(link_to, link_from)
         return flask.jsonify(valid=False)
+    except api.MediawikiError as e:
+        return flask.jsonify(valid=False, error=str(e))
 
     return flask.jsonify(valid=True, diff=found["diff"], replacement=found["replacement"])