You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
In Python SDK we have send_video since early days of video support. But this methods utilize uploadBlob method. Which performs all communication with video service under the hood. I really appreciate that you introduced such an easy method to start! But this method has lack of access to processing errors. And it started the common issue for our Python community to ask why "video not found" happens. That's why I started writing the complete example of how to use Video Service directly with job status handling and etc. During the development, I faced with a few really strange things that I want to raise.
I have split this issue to 3 sections. Each represents own problem.
1. Server response is out of the lexicon
To Reproduce
Upload video using uploadVideo endpoint of Video Service
You will see the response like this:
The response misses jobStatus root object described in the lexicon.
It is getting really hard to work with. As it requires try-catch for 409 and additional error content deserialization to the proper model. Which is out of the default flow when we parse models in succeed response. It would be really helpful to see the status code from 2xx group.
3. Input of the uploadVideo method is described in the lexicon partially only
To Reproduce
Try to use uploadVideo only with the info described in the lexicon
You will be rejected by the server validation that you are missing DID or name input data.
Expected behavior
Probably this could be unique situation when the procedure has both input and parameters, but at least it should be described somehow in the lexicon (IMO). My understanding of missing parameters comes from the traffic sniffing of the official app...
Not an issue, just a question
Any participial reason to make name parameter required? As I can see it is randomly generated with the same extension of the uploaded file. Could be easily done on server side? And the DID could be obtained from the service token? Am I wrong?
P.S. I am sorry for being so direct ❣️ I am understand that Video Service is like 3d party service and could do not follow the best practices of atproto, but since is it developed by the same developers I expected more smooth work with it. I just shared my experience. Happy to hear your thoughts. Thank you!
The text was updated successfully, but these errors were encountered:
Describe the bug
In Python SDK we have
send_video
since early days of video support. But this methods utilizeuploadBlob
method. Which performs all communication with video service under the hood. I really appreciate that you introduced such an easy method to start! But this method has lack of access to processing errors. And it started the common issue for our Python community to ask why "video not found" happens. That's why I started writing the complete example of how to use Video Service directly with job status handling and etc. During the development, I faced with a few really strange things that I want to raise.I have split this issue to 3 sections. Each represents own problem.
1. Server response is out of the lexicon
To Reproduce
Upload video using
uploadVideo
endpoint of Video ServiceYou will see the response like this:
The response misses
jobStatus
root object described in the lexicon.Expected behavior
atproto/lexicons/app/bsky/video/uploadVideo.json
Lines 11 to 23 in c34426f
jobStatus
is required field which should be this definition:atproto/lexicons/app/bsky/video/defs.json
Lines 2 to 5 in c34426f
2. Server status code responses
To Reproduce
Use
uploadVideo
again with the same video file.HTTP 409 with
Video processed successfully
is a bit wild and IMO does not fit XRPC ideology of status codes (https://atproto.com/specs/xrpc#summary-of-http-status-codes)Expected behavior
It is getting really hard to work with. As it requires try-catch for 409 and additional error content deserialization to the proper model. Which is out of the default flow when we parse models in succeed response. It would be really helpful to see the status code from 2xx group.
3. Input of the
uploadVideo
method is described in the lexicon partially onlyTo Reproduce
Try to use
uploadVideo
only with the info described in the lexiconatproto/lexicons/app/bsky/video/uploadVideo.json
Lines 8 to 10 in c34426f
You will be rejected by the server validation that you are missing
DID
orname
input data.Expected behavior
Probably this could be unique situation when the procedure has both
input
andparameters
, but at least it should be described somehow in the lexicon (IMO). My understanding of missing parameters comes from the traffic sniffing of the official app...Not an issue, just a question
Any participial reason to make
name
parameter required? As I can see it is randomly generated with the same extension of the uploaded file. Could be easily done on server side? And the DID could be obtained from the service token? Am I wrong?P.S. I am sorry for being so direct ❣️ I am understand that Video Service is like 3d party service and could do not follow the best practices of atproto, but since is it developed by the same developers I expected more smooth work with it. I just shared my experience. Happy to hear your thoughts. Thank you!
The text was updated successfully, but these errors were encountered: