Support configuring a proxy for fetching YouTube transcripts

Add a `YOUTUBE_PROXY` environment variable which specifies a proxy
service to use when fetching caption tracks/transcripts from YouTube.

The format of this variable should be an argument that can be used with the
`proxies` argument to `requests.post`, eg.
"http://proxy_user:proxy_pass@proxy_host:port".

Part of hypothesis/support#143
Slack thread: https://hypothes-is.slack.com/archives/C4K6M7P5E/p1723188976861449

tests/unit/via/services/youtube_transcript_test.py

@@ -289,12 +289,43 @@ def svc(self, http_service):
 
 
 class TestFactory:
-    def test_factory(self, YouTubeTranscriptService, http_service, pyramid_request):
+    def test_factory_without_proxy(
+        self, YouTubeTranscriptService, pyramid_request, HTTPService
+    ):
         svc = factory(sentinel.context, pyramid_request)
 
-        YouTubeTranscriptService.assert_called_once_with(http_service=http_service)
+        HTTPService.assert_called_once()
+        session = HTTPService.call_args[1]["session"]
+        assert session.proxies == {}
+
+        YouTubeTranscriptService.assert_called_once_with(
+            http_service=HTTPService.return_value
+        )
+
         assert svc == YouTubeTranscriptService.return_value
 
+    def test_factory_with_proxy(
+        self, YouTubeTranscriptService, pyramid_request, HTTPService
+    ):
+        proxy_server = "http://proxy_user@proxy_pass:proxy_host.com:123"
+        pyramid_request.registry.settings["youtube_proxy"] = proxy_server
+
+        svc = factory(sentinel.context, pyramid_request)
+
+        HTTPService.assert_called_once()
+        session = HTTPService.call_args[1]["session"]
+        assert session.proxies == {"https": proxy_server}
+
+        YouTubeTranscriptService.assert_called_once_with(
+            http_service=HTTPService.return_value
+        )
+
+        assert svc == YouTubeTranscriptService.return_value
+
+    @pytest.fixture
+    def HTTPService(self, patch):
+        return patch("via.services.youtube_transcript.HTTPService")
+
     @pytest.fixture
     def YouTubeTranscriptService(self, patch):
         return patch("via.services.youtube_transcript.YouTubeTranscriptService")

tox.ini

@@ -48,6 +48,7 @@ passenv =
     dev: CLIENT_EMBED_URL
     dev: SIGNED_URLS_REQUIRED
     dev: YOUTUBE_API_KEY
+    dev: YOUTUBE_PROXY
 deps =
     pip-tools
     pip-sync-faster

via/app.py

@@ -30,6 +30,7 @@
     "youtube_transcripts": {"formatter": asbool},
     "api_jwt_secret": {"required": True},
     "youtube_api_key": {},
+    "youtube_proxy": {},
 }
 
 

via/services/youtube_transcript.py

@@ -4,6 +4,8 @@
 from typing import Dict, List
 from xml.etree import ElementTree
 
+import requests
+
 from via.services.http import HTTPService
 
 
@@ -145,4 +147,8 @@ def strip_html(xml_string):
 
 
 def factory(_context, request):
-    return YouTubeTranscriptService(http_service=request.find_service(HTTPService))
+    session = requests.Session()
+    if youtube_proxy := request.registry.settings.get("youtube_proxy"):
+        session.proxies["https"] = youtube_proxy
+    http_svc = HTTPService(session=session)
+    return YouTubeTranscriptService(http_service=http_svc)