route_average_per_task is slow and results in timeouts with No Tasks Available

2

hi @Arnault,

Thanks for your post.

There was a task router fix in v1.13.2 that fixed this issue:

The fixes include:

  • Task router issues occurred due to treating the average variable as non-constant, leading to overwrites.
  • The db.get_hash_count call inaccurately counted hashes by querying both the task and session tables.
  • Users annotating examples from previous sessions could still receive tasks, but on restart, their presence in the annot pool hindered item distribution to other users.

I suspect that fix in get_hash_count may be the problem. Can you update to the latest version of Prodigy and see if the problem remains?