feat: upload benchmark results to Braintrust - #101

Draft

Rodriguespn wants to merge 15 commits into

mainfrom

feat/braintrust-benchmark-upload

Commits on Aug 2, 2026

feat: upload benchmark results to Braintrust alongside eval-results.json

Rodriguespn
and
claude
committed
b1883ac
fix: address review findings on Braintrust upload

Rodriguespn
and
claude
committed
4cc4259
feat: overall passed verdict as a score span in Braintrust traces

Rodriguespn
and
claude
committed
bc354a4
feat: structured AgentTrace transcript + hierarchical Braintrust spans

Rodriguespn
and
claude
committed
be01119
feat: turn/step trace semantics, single passed scorer, judge evidence spans

Rodriguespn
and
claude
committed
cb85787
feat: attach the agent CLI's verbatim transcript to Braintrust rows

Rodriguespn
and
claude
committed
9c74d79
refactor: review-pass cleanups on the Braintrust upload branch

Rodriguespn
and
claude
committed
6cc2804
feat: add custom suite mode to eval-refresh for ad hoc branch runs

Rodriguespn
and
claude
committed
1d7be86
feat: baseline comparison + auto-updating PR comment for Braintrust uploads

Rodriguespn
and
claude
committed
0a42f07
feat: distinguish scheduled vs PR runs in Braintrust + pre-filtered list links

Rodriguespn
and
claude
committed
6e10e6a
chore: refresh eval results

github-actions[bot]
authored and
Rodriguespn
committed
12665f4
feat: compare PR runs against the latest-main baseline

Rodriguespn
and
claude
committed
cc6d986
feat: batch baseline rotation — demote old holders only after full upload

Rodriguespn
and
claude
committed
9eb8d3a
decouple experiment metadata field in four
Rodriguespn
committed
25e2b4f
feat: align experiment metadata with the design doc's field table

Rodriguespn
and
claude
committed
40ad9e2

Navigation Menu

Uh oh!

feat: upload benchmark results to Braintrust - #101

feat: upload benchmark results to Braintrust#101
Rodriguespn wants to merge 15 commits into
mainfrom
feat/braintrust-benchmark-upload

Commits on Aug 2, 2026

feat: upload benchmark results to Braintrust alongside eval-results.json

fix: address review findings on Braintrust upload

feat: overall passed verdict as a score span in Braintrust traces

feat: structured AgentTrace transcript + hierarchical Braintrust spans

feat: turn/step trace semantics, single passed scorer, judge evidence spans

feat: attach the agent CLI's verbatim transcript to Braintrust rows

refactor: review-pass cleanups on the Braintrust upload branch

feat: add custom suite mode to eval-refresh for ad hoc branch runs

feat: baseline comparison + auto-updating PR comment for Braintrust uploads

feat: distinguish scheduled vs PR runs in Braintrust + pre-filtered list links

chore: refresh eval results

feat: compare PR runs against the latest-main baseline

feat: batch baseline rotation — demote old holders only after full upload

decouple experiment metadata field in four

feat: align experiment metadata with the design doc's field table