replace 'silence' with 'verbose' option

[jvmncs]

I've found that the logs are pretty hard to work with when requests come with very long messages (in my case, on the order of 7k tokens). this PR expands the log verbosity options into 3 levels:

• verbose=0: suppress tqdm bars and 100% of log output (what the docstring says silent=True should do, although currently it doesn't suppress logs at all)
• verbose=1: display tqdm bars and log requests with metadata only
• verbose=2: display tqdm bars and log requests with metadata and all request params (including messages). this is roughly equivalent to current behavior of silence=False

I believe verbose=1 is a more sensible default, but kept default verbose=2 to align with the current behavior

[ctjlewis]

@jvmncs shit, you gotta rebase bc of the other PR. otherwise lgtm

[jvmncs]

ah I just merged through github UI. can redo with a proper rebase if you prefer

[ctjlewis]

no that's OK. i might update the name of that param to loglevel: "silent" | "warn" | "verbose" or something similar.

[ctjlewis]

Damn, tests are failing - i forgot to add tests as required workflow for PRs or we'd have seen it here pre-merge.

Edit:

• actually, there's a widespread OpenAI outage;
• AND, we had an outstanding invoice;
• AND, we can't access GitHub secrets from the PR workflow because GitHub is concerned our token will leak.

So I will reset the branch and test manually.

[ctjlewis]

Confirmed failing tests had nothing to do with your code so this will ship after I bump version.

Edit: Shipped as v1.1.2.