APIv3: proxy these URLs to be served from El Proxito /_/api/v3/
#11831
Conversation
humitos
commented
Dec 9, 2024
humitos
commented
- Related: API: consider using APIv3 standard endpoints addons#356
- Related: API: use APIv3 endpoint for resources addons#468
| api_proxied_urls = [ | ||
| path("embed/", ProxiedEmbedAPI.as_view(), name="embed_api_v3"), | ||
| path("search/", ProxiedSearchAPI.as_view(), name="search_api_v3"), | ||
| ] | ||
|
|
||
| urlpatterns = api_proxied_urls | ||
| urlpatterns += router.urls |
There was a problem hiding this comment.
I know Santos had some worried about handling modifications here. This seems like a pretty large change?
Santos mentioned:
We can cache content over docs domains (assuming we only expose read-only resources, and on .org only).
And this doesn't seem to restrict any of those things? In particular, we need to remove auth from these API endpoints (
readthedocs.org/readthedocs/core/mixins.py
Lines 25 to 27 in bb2aca3
There was a problem hiding this comment.
Yes. I haven't jumped into read-only endpoints / auth / cache yet because I wanted to be sure I'm moving in the right direction here before implementing those. If this POC is 👍🏼 -- we can start exploring how to achieve those goals as well.
There was a problem hiding this comment.
By the way, I understand we want auth here because .com will require authentication requests to return the correct data for those private projects.
There was a problem hiding this comment.
By the way, I understand we want auth here because .com will require authentication requests to return the correct data for those private projects.
Yea, looks like we're setting SessionAuthentication explicitly on the Corporate side for the proxied APIs.
| @@ -406,36 +354,6 @@ def _v1(self, project, version, build, filename, url, request): | |||
|
|
|||
| data = { | |||
| "api_version": "1", | |||
| "projects": { | |||
There was a problem hiding this comment.
I worry a little bit about deleting this on deploy, since it will lead to a timing issue between the client & server. I'd probably just add the new fields to the existing API response, so we can test that appraoch without breaking old API clients during deploy?
There was a problem hiding this comment.
Yeah, it makes sense to deploy everything needed in the backend first without removing the old pattern 👍🏼
