| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
| |
Repeated requests from the server responds in an ALREADY_EXISTS error and does not transcribe the voicemails. In that case we should ask the server if the request is present already on the server, and if so retrieve it.
Bug: 77236260
Test: N/A
PiperOrigin-RevId: 191016332
Change-Id: I605c546a3bec3599c7b162853463ca7e46f63886
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Since we are now using the AlarmManager to decide when to poll
for voicemail transcriptions, we need to ensure that we don't
start transcribing a voicemail until polling for the previous
one ends.
This cl postpones a voicemail transcription task if there is an alarm
set to poll for a transcription result. After a transcription result
is received (or we give up) the db is scanned for any pending transcriptions
and the next one is started.
Bug: 70242961
Test: manual and updated unit tests
PiperOrigin-RevId: 181899975
Change-Id: I7b8fb696164980cf710aa58a79418c6954e2b4d2
|
|
|
result
This cl switches the way asynchronous transcription tasks poll for their
results. it has been observed that sometimes the server side transcription
takes longer than expected (sometimes many minutes), so instead of blocking
a thread for all that time this cl schedules an alarm for the next time to
poll. it also uses an exponential backoff scheme to determine the poll times
and increases the maximum total polling time from 20 seconds to 20 minutes.
Bug: 66966157
Test: manual and unit tests
PiperOrigin-RevId: 177257577
Change-Id: Ib2998f03cc418d5241ccffec71ba3945c9fe4cbc
|
