* pds: place-holder robots.txt and / (as txt)
robots.txt: want to be explicit about allowing crawling (generically)
/: a bit of dev-experience polish. in prod we already redirect this
route, but may be helpful for self-hosting folks and people just
exploring
* pds: move basic routes (health, robots.txt, /) to separate file
Improve BskyAgent performance
By using z.discriminatedUnion instead of z.union
we can hint to zod at which field it should look when
validating/parsing schemas
This improves BskyAgent instantiation time around 4 times,
when creating 10 agents using stuff from #1015,
the time went down from ~1200ms to 330ms
* Normalize record datetimes used for sort to simplified iso
* Update tests and snapshots for alternate valid datetime formats
* Fix notifications race during record update
* enable disabling account invites
* xtra test
* couple fixes
* allow re-enabling user accoutn invites
* add in getter route for if invites are disabled or not
* fix admin search test
* gneerated type on user account
* lex: 'escalate' mod action type
* api: codegen for escalate mod action type
* pds: codegen+impl esclate mod action type
* bsky: codegen+impl esclate mod action type
* pds: prettier:fix
* Tidy
---------
Co-authored-by: Devin Ivy <devinivy@gmail.com>
* Perform takedown then background hard-deletion process for deleteAccount
* Remove transactions from account deletion process, perform steps sequentially
* Add hostname and protocol validation for pds in getBlob()
Hostname and protocol validation for the pds' url in the getBlob() to prevent potential SSRF (Server-Side Request Forgery) attacks by ensuring that only valid protocols and hostnames are used in the URL. Function also will now throw a 400 error if detects any invalid.
* Migrate changes
* delete the line with console.log...
* Return `XRPCError` instead of 404 if a handler errors
Currently, if there is an uncaught error inside a handler, a 404 response is returned instead of the error itself. This fixes that.
* Catch it in the higher-level catch block
* Add test
* merge
* Improve test
Somehow the output of this log line has been ending up in datadog under
the wrong 'service' (with the 'service' being an "wss://"-prefixed URL,
not "pds" or "bav" or whatever expected).
Tweaks the 'service' key, and some others that might have conflict.
Tune "porn", "sexual", and "nudity" label categorization, including
re-structuring of code paths.
This comes out of some image testing as well as review of Hive API docs.
* block lexicons
* table schema
* block processing
* Actor-block migration
* Add blocked post lexicon to getPostThread
* getBlocks routes
* Add missing lexicons to bsky
* Add block state to actor views
* throw on blocked accounts in getAuthorFeed
* blcok state in profile views
* Update getPostThread to support block views, update existing test
* strip out blocks out oof timeline & popular
* more routes
* Test blocked posts in threads
* Support blocks on record embeds
* notifs
* Update pds test snapshots w/ block state
* fix search
* Update pds actor search snapshots w/ blocks
* test
* tests
* update lex
* fix up from returning blocking as record
* ci for branch
* fix build issue in bsky
* more blocks tests
* fix actor search test
* hook getBlocks up to router
* bugfix & actually test lol
---------
Co-authored-by: Devin Ivy <devinivy@gmail.com>
* Background notification creation/deletion
* Add todo
* Update tests for backgrounded notifs
* Add index to notifications supporting record deletion
* Use profile_agg to power getSuggestions
* Fetch record blocks separately from getting notification resultset
* Update background queue to drain pending tasks
* Lengthen pass reset and invite codes, invalidation on reset
* Tidy
* Adjust token format for reset and deletion
* Update invite codes to use same token gen as pass reset and acct deletion
Patryk
devin ivy
dholms
bnewbold
Jaga Santagostino
Wes Todd
Mustafa Yafes Akbaş
Roj
Ryan Winchester
Max Graey
Wes Bos
Paul Frazee