feat(foundry): ASSERT to Azure AI Foundry exporter (SDK-backed) - #267

Open

tangym wants to merge 26 commits into

mainfrom

tangym/foundry-exporter-v2

Commits on Jul 16, 2026

chore(foundry): add extra + empty subpackage scaffold
tangym
committed
22baf9b
feat(foundry): pure ASSERT run loader
tangym
committed
689c842
feat(foundry): custom evaluator spec builder (code + prompt variants)
tangym
committed
2e3c8c1
feat(foundry): dataset row builder (ASSERT scores + reasons on the row)
tangym
committed
1c2cdf7
feat(foundry): SDK-backed push pipeline (idempotent evaluators, content-hash datasets, fail-loud on eval drift)
tangym
committed
f481837
feat(foundry): CLI push subcommand with --evaluator-mode {code,prompt,both}
tangym
committed
33fe6b2
fix(foundry): flatten code eval schema + delete stale evaluators on drift
tangym
committed
5fc54d0
docs(foundry): schema mapping reference + customer runbook
tangym
committed
eb8828c
refactor(foundry-cli): render push output with rich tables + add --json / --no-color
tangym
committed
89d416a
feat(foundry): fingerprint-based evaluator drift detection
tangym
committed
b93987b
feat(foundry): expose evaluator fingerprints in dry-run output
tangym
committed
b11e151
refactor(foundry): bump evaluator versions on drift instead of delete+recreate
tangym
committed
09f6939
fix(foundry): canonicalize numeric bounds + absent text bodies in fingerprint
tangym
committed
41580cb
chore(foundry): pin azure-ai-projects upper bound + drop stale scaffold comments
tangym
committed
f5c153b
fix(foundry): tolerate un-scored rows in data-source + evaluator schemas
tangym
committed
e96da56
fix(foundry): narrow _iter_paged catch, guard _dict_get on methods, UTF-8-safe bounded_record
tangym
committed
5ebf97b
feat(foundry): surface estimated prompt-variant LLM calls in dry-run output
tangym
committed
fb2e1dd
docs(foundry): note description-vs-rubric asymmetry + tool events dropped
tangym
committed
a4f77aa
docs(foundry): sync required-field lists with e96da568 schema relaxations
tangym
committed
10ac083
fix(foundry): narrow _list_evaluator_versions fallback to TypeError
tangym
committed
af9ad2e
fix(foundry-cli): make --project optional in --dry-run mode
tangym
committed
f08a4d4
docs(changelog): add foundry exporter preview entry
tangym
committed
b2ba11e
test(foundry): skip SDK-backed test modules on base install
tangym
committed
ff6bd93
fix(foundry-cli): drop non-ASCII chars from dry-run output
tangym
committed
0ee7c99
docs(foundry): align expected-output blocks with current CLI
tangym
committed
25d8b07

Commits on Aug 6, 2026

Merge branch 'main' into tangym/foundry-exporter-v2
Aaron Aspinwall (AaronAspinwall123)
authored
9957148

Navigation Menu

Uh oh!

feat(foundry): ASSERT to Azure AI Foundry exporter (SDK-backed) - #267

feat(foundry): ASSERT to Azure AI Foundry exporter (SDK-backed)#267
tangym wants to merge 26 commits into
mainfrom
tangym/foundry-exporter-v2

Commits on Jul 16, 2026

chore(foundry): add extra + empty subpackage scaffold

feat(foundry): pure ASSERT run loader

feat(foundry): custom evaluator spec builder (code + prompt variants)

feat(foundry): dataset row builder (ASSERT scores + reasons on the row)

feat(foundry): SDK-backed push pipeline (idempotent evaluators, content-hash datasets, fail-loud on eval drift)

feat(foundry): CLI push subcommand with --evaluator-mode {code,prompt,both}

fix(foundry): flatten code eval schema + delete stale evaluators on drift

docs(foundry): schema mapping reference + customer runbook

refactor(foundry-cli): render push output with rich tables + add --json / --no-color

feat(foundry): fingerprint-based evaluator drift detection

feat(foundry): expose evaluator fingerprints in dry-run output

refactor(foundry): bump evaluator versions on drift instead of delete+recreate

fix(foundry): canonicalize numeric bounds + absent text bodies in fingerprint

chore(foundry): pin azure-ai-projects upper bound + drop stale scaffold comments

fix(foundry): tolerate un-scored rows in data-source + evaluator schemas

fix(foundry): narrow _iter_paged catch, guard _dict_get on methods, UTF-8-safe bounded_record

feat(foundry): surface estimated prompt-variant LLM calls in dry-run output

docs(foundry): note description-vs-rubric asymmetry + tool events dropped

docs(foundry): sync required-field lists with e96da568 schema relaxations

fix(foundry): narrow _list_evaluator_versions fallback to TypeError

fix(foundry-cli): make --project optional in --dry-run mode

docs(changelog): add foundry exporter preview entry

test(foundry): skip SDK-backed test modules on base install

fix(foundry-cli): drop non-ASCII chars from dry-run output

docs(foundry): align expected-output blocks with current CLI

Commits on Aug 6, 2026

Merge branch 'main' into tangym/foundry-exporter-v2