Conversation
Abhijeet Prasad (AbhiPrasad)
left a comment
There was a problem hiding this comment.
This also needs a bunch more tests, but we can tackle that after this.
| try { | ||
| const value = await rootSpan.traced(fn, { | ||
| name: spanName, | ||
| spanAttributes: { type: spanType, purpose: "scorer" }, |
There was a problem hiding this comment.
should we make this purpose: "classifier" for classifier spans? We might need to introduce a new runInClassiferSpan helper.
There was a problem hiding this comment.
I'm not totally sure of the implications of this, I think it would require a bunch of backend changes as well if we did. Could be a good idea, though, Alex Z (@CLowbrow) do you know?
There was a problem hiding this comment.
More context from Ankur - this is a PITA to change because there is a bunch of places in the backend that rely on it. We can't do it. Relevant slack thread: https://braintrustdata.slack.com/archives/C083GCUTVDZ/p1773705541982479
662efe8 to
30ca11a
Compare
we're not using confidence yet, so leave it out for now, we can always add it in later
8afd421 to
1421ef7
Compare
|
warning AI slop: I ran this PR against my spec in https://github.com/braintrustdata/braintrust-spec/pull/2 (after cleaning up stuff like the confidence changes) and it codex gave me the two following things:
These seem like things we have to tackle. |

No description provided.