Caching error with groupby

[lisadunlap]

Thanks for fixing the previous caching! Found one more bug

Describe the bug
Getting caching errors when group by is used for sem_agg

Input:

summaries = results.sem_agg(
        create_reduce_prompt(num_final_vibes),
        group_by="cluster_id",
        suffix="reduced axes",
    )

Output:

Traceback (most recent call last):
  File "/home/lisabdunlap/VibeCheck/main.py", line 424, in <module>
    main(
  File "/home/lisabdunlap/VibeCheck/main.py", line 257, in main
    vibes = propose_vibes(
  File "/home/lisabdunlap/VibeCheck/main.py", line 197, in propose_vibes
    summaries = results.sem_agg(
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/site-packages/lotus/cache.py", line 78, in wrapper
    result = func(self, *args, **kwargs)
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/site-packages/lotus/sem_ops/sem_agg.py", line 197, in __call__
    return pd.concat(list(executor.map(SemAggDataframe.process_group, group_args)))
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/concurrent/futures/_base.py", line 621, in result_iterator
    yield _result_or_cancel(fs.pop())
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/concurrent/futures/_base.py", line 319, in _result_or_cancel
    return fut.result(timeout)
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/concurrent/futures/_base.py", line 458, in result
    return self.__get_result()
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/concurrent/futures/_base.py", line 403, in __get_result
    raise self._exception
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/concurrent/futures/thread.py", line 58, in run
    result = self.fn(*self.args, **self.kwargs)
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/site-packages/lotus/sem_ops/sem_agg.py", line 150, in process_group
    return group.sem_agg(user_instruction, all_cols, suffix, None, progress_bar_desc=progress_bar_desc)
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/site-packages/lotus/cache.py", line 71, in wrapper
    cached_result = model.cache.get(cache_key)
  File "/home/lisabdunlap/miniconda3/envs/vibecheck/lib/python3.10/site-packages/lotus/cache.py", line 156, in get
    with self.conn:
sqlite3.ProgrammingError: SQLite objects created in a thread can only be used in that same thread. The object was created in thread id 138740594186048 and this is thread id 138728762771136.

Input that works:

    summaries = results.sem_agg(
        create_reduce_prompt(num_final_vibes),
        suffix="reduced axes",
    )

[StanChan03]

@dhruviyer Any thoughts on this threading issue with sqlite?

[dhruviyer]

@StanChan03 I will look into it! Appears to be a problem when using multi-threading with caching

[StanChan03]

Yea, I was discussing this with Sid, either maybe sync the threads or have an alternative caching backend

[dhruviyer]

Tracking a potential fix for this in #92