feat: evidence-grade NIM discovery + all-modality cost-quality benchmark - #90

Draft

seonghobae wants to merge 200 commits into

fix/atheris-interpreter-lockfrom

claude/nim-all-models-support-fecb0b

Commits on Aug 4, 2026

feat: evidence-grade NIM discovery + all-modality cost-quality benchmark (#86)

seonghobae
and
claude
committed
064973e
merge: stack NIM benchmark on provider-egress security base
seonghobae
committed
b48f396
docs(changelog): record NIM benchmark evidence slice
seonghobae
committed
0e19b3a
ci: apply reviewed NIM transport hardening
seonghobae
committed
66c8d55
fix(nim): harden egress and equalize policy budgets
seonghobae
committed
9d546a7
fix(nim): install evidence and transport hardening
seonghobae
committed
785fb15
test(nim): prove secure transport and equal budgets
seonghobae
committed
e07a504
docs(nim): document pinned egress and equal total budgets
seonghobae
committed
d0312ea
chore(ci): remove temporary NIM repair workflow
seonghobae
committed
c7bcad2
chore(ci): stage one-time NIM source repair
seonghobae
committed
e1ae745
chore(ci): remove privileged one-shot NIM repair workflow
seonghobae
committed
1015608
chore(ci): stage exact-head NIM security repair
seonghobae
committed
07128ac
chore(ci): harden bounded NIM repair job
seonghobae
committed
a8393d7
chore(ci): stage bounded NIM source integration
seonghobae
committed
4cc0f55
chore(ci): remove overlapping one-shot NIM workflow
seonghobae
committed
9ad742a
fix(ci): replace write-capable PR repair with isolated publisher
seonghobae
committed
4fd96f1
docs(changelog): record reviewed NIM access evidence
seonghobae
committed
4bb7c85
chore(ci): stage equal-budget regression repair
seonghobae
committed
d9bedd7
ci: complete bounded NIM source integration
seonghobae
committed
c4d47d2
chore(ci): remove redundant NIM test-only repair workflow
seonghobae
committed
db6ebb9
ci: run bounded NIM source integration from PR checks
seonghobae
committed
dba1d7e
fix(ci): execute bounded NIM source transformation deterministically
seonghobae
committed
a3741b5
fix(ci): preserve embedded transformation literals
seonghobae
committed
18d3fe7
fix(security): remove PR-head OIDC source mutation
seonghobae
committed
27eb1b8
fix(security): delete privileged branch-trigger repair workflow
seonghobae
committed
ba6b34b
fix(security): delete branch-controlled OIDC repair source
seonghobae
committed
e4578da
test(security): pin PR workflow OIDC boundary
seonghobae
committed
f80e503
ci(review): export read-only exact-head workspace
seonghobae
committed
bc77bfc
ci(review): include inert transformation evidence
seonghobae
committed
a71ebbc

Commits on Aug 5, 2026

test(ci): require secretless NIM dry-run job
seonghobae
committed
6df72ee
fix(ci): isolate NIM dry runs from provider secrets
seonghobae
committed
09fd519
docs(security): record NIM workflow secret isolation
seonghobae
committed
2b622bd
docs(changelog): record secretless NIM dry runs
seonghobae
committed
40b48ba
docs(benchmark): explain workflow credential isolation
seonghobae
committed
348e5e6
chore(review): stage verified NIM repair patch
seonghobae
committed
667b311
ci(nim): stage reviewed patch part 1 of 7
seonghobae
committed
2498648
ci(nim): stage reviewed patch part 2 of 7
seonghobae
committed
4552084
ci(nim): stage reviewed patch part 3 of 7
seonghobae
committed
45c151e
ci(nim): stage reviewed patch part 4 of 7
seonghobae
committed
9dac6b2
ci(nim): stage reviewed patch part 5 of 7
seonghobae
committed
fecd25f
ci(nim): stage reviewed patch part 6 of 7
seonghobae
committed
19be572
ci(nim): stage reviewed patch part 7 of 7
seonghobae
committed
736893c
ci(nim): apply exact reviewed benchmark integration
seonghobae
committed
07f28a9
ci: retrigger reviewed NIM integration with scoped permissions
seonghobae
committed
7d5d6aa
ci: verify and export reviewed NIM integration read-only
seonghobae
committed
fd6b074
ci: bind NIM verification to security ancestor
seonghobae
committed
cb7e9b3
fix(ci): bind NIM verification to pull-request head
seonghobae
committed
3878e5b
fix(ci): reconcile reviewed NIM patch per file
seonghobae
committed
5bbedc2
fix(ci): reconstruct reviewed NIM tree from patch base
seonghobae
committed
d98fe6c
fix(ci): bind reviewed NIM reconstruction to exact head
seonghobae
committed
428736c
fix(ci): run reviewed NIM verification read-only on push
seonghobae
committed
e4e03c4
fix(ci): use runner temp variable at execution time
seonghobae
committed
6cd2012
fix(ci): discover the exact reviewed patch base
seonghobae
committed
7595919
chore(ci): inspect the staged NIM integration payload
seonghobae
committed
a93ec20
fix(ci): replay the reviewed NIM patch stack
seonghobae
committed
f11193d
fix(ci): rehydrate the reviewed patch index
seonghobae
committed
83b7906
fix(ci): execute the patch rehydration helper
seonghobae
committed
e9bfb85
fix(ci): apply the reviewed patch against the rehydrated index
seonghobae
committed
88274c0
fix(ci): verify the reconstructed reviewed tree
seonghobae
committed
03b30e5
fix(ci): locate the exact reviewed patch context
seonghobae
committed
e256a9a
fix(ci): build the wheel in an isolated backend environment
seonghobae
committed
012f7c7
ci(nim): publish the authenticated verified source tree
seonghobae
committed
dde856f
feat(nim): integrate verified provider-neutral benchmark
seonghobae
committed
8049ca8
docs(nim): record immutable integration provenance
seonghobae
committed
d1516a5
test(ci): require stacked pull-request coverage
seonghobae
committed
f090ffc
ci: validate stacked pull requests
seonghobae
committed
d091b69
ci: fuzz stacked pull requests
seonghobae
committed
0d66e0a
ci: secure stacked pull requests
seonghobae
committed
928214c
test(ci): require isolated wheel builds
seonghobae
committed
c759cd1
fix(ci): isolate benchmark wheel builds
seonghobae
committed
8e19d64
ci(review): export read-only PR90 workspace
seonghobae
committed
e7f7238
test(benchmark): stage complete-budget preflight red tests part 1
seonghobae
committed
c72f1cb
test(benchmark): stage complete-budget preflight red tests part 2
seonghobae
committed
cb20657
fix(benchmark): stage complete request planner part 3
seonghobae
committed
636c3e0
docs(benchmark): stage complete-budget evidence and verification part 4
seonghobae
committed
1053f19
ci(benchmark): apply complete-budget preflight test-first
seonghobae
committed
638a126
ci(review): stage exact PR90 request-plan patch
seonghobae
committed
1f6da53
ci(review): refresh exact PR90 request-plan patch
seonghobae
committed
8a31de7
ci(review): verify and publish exact PR90 request-plan fix
seonghobae
committed
87496c8
ci(review): inspect exact PR90 patch hash
seonghobae
committed
b3c0471
ci(review): export decoded PR90 patch evidence
seonghobae
committed
2efcf08
ci(review): verify exact staged PR90 patch
seonghobae
committed
a06f9c5
ci(review): publish verified PR90 request-plan fix
seonghobae
committed
77cc8cc
test(benchmark): align scheduled complete-plan contract
seonghobae
committed
796e6e9
test(benchmark): require request-plan provenance schema
seonghobae
committed
5956214
fix(nim): require a complete catalog benchmark plan
seonghobae
committed
9b72816
merge: integrate provider-egress security base into NIM benchmark
github-actions[bot]
committed
a2af363
docs(adr): record exact NIM security integration evidence
seonghobae
committed
c5b3876
test(benchmark): expose undersized locked evaluation manifest
seonghobae
committed
d43adaa
feat(benchmark): expand locked manifest to evidence floor
seonghobae
committed
16b5016
test(benchmark): bind evidence floor to existing hard ceiling
seonghobae
committed
ac13a56
docs(benchmark): record thirty-task evidence floor
seonghobae
committed
62e0e59
docs(benchmark): align request plan and evidence floor
seonghobae
committed
43c629f
docs(benchmark): doctor thirty-task evidence floor
seonghobae
committed
ff51584
test(benchmark): expose undersized default request caps
seonghobae
committed
bec1129
ci(benchmark): test and repair default request caps
seonghobae
committed
cdfe08e
ci(benchmark): rearm default request-cap repair
seonghobae
committed
37c78b4
ci(benchmark): add exact-head reopen trigger
seonghobae
committed
545cd64
ci(benchmark): use resolvable checkout action
seonghobae
committed
e56d1a3
ci(benchmark): repair complete-plan default finalizer
seonghobae
committed
5216171
test(benchmark): bind evidence floor to canonical request planner
seonghobae
committed
8cae8d4
ci(benchmark): cover all complete-plan regression anchors
seonghobae
committed
e31f901
ci(benchmark): repair invalid one-shot YAML
seonghobae
committed
da8ec5f
ci(benchmark): split workflow-authorized publication
seonghobae
committed
1b97e48
fix(benchmark): align complete-plan source evidence
github-actions[bot]
committed
0ebb1d5
fix(benchmark): align manual request cap with complete plan
seonghobae
committed
9e4b50b
ci(benchmark): remove completed one-shot workflow
seonghobae
committed
821b58e
test(nim): cover buyer-facing complete request plan
seonghobae
committed
7b23bac
ci(nim): include request-plan regression in quality gate
seonghobae
committed
5c1a3f6
test: require NIM parser Atheris instrumentation
seonghobae
committed
f5a494f
fix: instrument NIM parser imports for Atheris
seonghobae
committed
48f007d
docs: distinguish historical NIM integration evidence
seonghobae
committed
2736113
test: cover NIM review regressions
seonghobae
committed
b60ec5d
chore: defer broader NIM regression batch
seonghobae
committed
bdc9673
docs: make NIM receipt head-stable
seonghobae
committed
646c417
docs: record NIM fuzz instrumentation repair
seonghobae
committed
318feb9
test(nim): require completion-dependent README evidence status
seonghobae
committed
63438d0
docs(nim): clarify completion-dependent evidence status
seonghobae
committed
6fc53f1
test(nim): normalize README contract whitespace
seonghobae
committed
4a79964
docs(changelog): record NIM evidence-status clarification
seonghobae
committed
1217fc9
test(security): inspect every pull-request workflow structurally
seonghobae
committed
11353d4
test(nim): enforce ordered non-empty secret-isolation steps
seonghobae
committed
67c5c84
test(security): parse pull-request branch filters structurally
seonghobae
committed
989d7fe
test(nim): lock review regressions before repair
seonghobae
committed
4329081
fix(fuzz): require normalized NIM catalog depth failures
seonghobae
committed
715ee76
test: skip fixture-bound standalone NIM cases
seonghobae
committed
8c4ee86
fix: close NIM evidence review gaps
seonghobae
committed
4ca0448
docs: record NIM evidence hardening
seonghobae
committed
aed3e7c
test: require NIM review regressions in coverage gate
seonghobae
committed
a5d529b
ci: cover current NIM review regressions
seonghobae
committed
8603bcf
docs: record exact NIM regression gate
seonghobae
committed
e9f2072
test(nim): require complete CSV model-assignment evidence
seonghobae
committed
ffd03ba
feat(nim): preserve model assignment evidence in CSV artifacts
seonghobae
committed
a9bf8bc
feat(nim): fail closed until CSV evidence is complete
seonghobae
committed
d031f53
ci(nim): gate CSV evidence adapter at 100% coverage
seonghobae
committed
554a220
docs(nim): record CSV assignment evidence parity
seonghobae
committed
2db0737
docs(nim): doctor CSV assignment evidence boundary
seonghobae
committed
dfc300a
refactor(nim): preserve CSV mode without cleanup branches
seonghobae
committed
5c2c695
test(nim): cover CSV evidence fail-closed edge cases
seonghobae
committed
7dd5245
ci(nim): execute CSV evidence edge regressions
seonghobae
committed
816707f
test(nim): make routing disclaimer contract case-insensitive
seonghobae
committed
00483f3
ci(nim): preserve benchmark docstring contract and gate adapter
seonghobae
committed
88b0e96
test(ci): require pull request workflows to checkout exact heads
seonghobae
committed
c9fa27c
ci: checkout exact pull request heads in test jobs
seonghobae
committed
5e46772
ci: checkout exact pull request heads in fuzz jobs
seonghobae
committed
f7320aa
ci: checkout exact pull request heads in security jobs
seonghobae
committed
16f1409
docs(ci): record exact-head pull request verification boundary
seonghobae
committed
692e49a
docs(ci): doctor exact-head versus merge-test evidence
seonghobae
committed
1ef2c97
merge: synchronize NIM benchmark stack with PR #96
seonghobae
committed
f042953

Commits on Aug 6, 2026

test(nim): define transactional artifact publication contract
seonghobae
committed
2b84e1e
fix(nim): publish complete evidence sets transactionally
seonghobae
committed
ef28784
test(nim): route CSV wrapper fakes through private staging
seonghobae
committed
ed52bc8
ci(nim): include transactional publication regressions
seonghobae
committed
5771957
test(nim): write fake artifacts into precreated staging
seonghobae
committed
e1f19ff
test(nim): reuse wrapper-created staging directory
seonghobae
committed
c446251
test(nim): cover transactional publication failure edges
seonghobae
committed
30c046f
ci(nim): include publication edge coverage
seonghobae
committed
6ed1fbf
test(nim): cover duplicate split output argument
seonghobae
committed
5e46ba5
docs(nim): record transactional evidence-set publication
seonghobae
committed
1cea0c0

Commits on Aug 7, 2026