Repository navigation

perf(onboarding): optimize onboarding pipeline from ~5min to ~2min - #2777

Merged

Marfuen merged 28 commits into

mainfrom

mariano/onboarding-perf

May 7, 2026

Commits on May 7, 2026

perf(onboarding): parallelize vendor/risk extraction, fire-and-forget policies

Marfuen
and
claude
committed
a7ea2be
perf(onboarding): parallelize risk + vendor matching in linkage step

Marfuen
and
claude
committed
c41f213
perf(onboarding): switch linkage reranker from gpt-5-mini to gemini-3-flash

Marfuen
and
claude
committed
a02aeb7
chore: upgrade AI SDK to v6/v3, switch reranker to gemini-3-flash via gateway

Marfuen
and
claude
committed
d29f7b9
chore: remove caBundleExtension from trigger config

Marfuen
and
claude
committed
e8d0580
fix(api): await convertToModelMessages in assistant-chat controller

Marfuen
and
claude
committed
6ac5f37
fix(onboarding): restore tracker progress for parallel + fire-and-forget flow

Marfuen
and
claude
committed
d9e067f
fix(onboarding): handle pipeda and other framework conditionals in policy templates

Marfuen
and
claude
committed
a562830
fix(onboarding): restore expandable sections in tracker during background work

Marfuen
and
claude
committed
bff0b24
fix(onboarding): use tasks.batchTrigger for policies to preserve parent metadata

Marfuen
and
claude
committed
4b4a88d
fix(onboarding): restore concurrencyKey for policy batch trigger

Marfuen
and
claude
committed
df813cd
fix(onboarding): sync all tracker spinner animations

Marfuen
and
claude
committed
dcd783b
perf(onboarding): raise policy queue concurrency from 5 to 15

Marfuen
and
claude
committed
29bcf37
perf(onboarding): switch policy generation from gpt-5-mini to gemini-3-flash

Marfuen
and
claude
committed
4e29a15
perf(onboarding): policies first, gemini for all LLM calls, non-blocking vendor side-effects

Marfuen
and
claude
committed
41b8cc1
perf(onboarding): replace LLM policy generation with programmatic template processor

Marfuen
and
claude
committed
faa8d0e
fix(onboarding): switch tracker from useRealtimeRun to useRun with polling

Marfuen
and
claude
committed
5840046
feat(onboarding): add targeted LLM pass to rewrite policy cue lines

Marfuen
and
claude
committed
a3335b3
perf(onboarding): switch reranker to gemini-3.1-flash-lite-preview

Marfuen
and
claude
committed
4eff65d
fix(onboarding): use tasks.batchTrigger for vendor/risk mitigations

Marfuen
and
claude
committed
0d3a2dc
fix: switch onboarding status hooks from useRealtimeRun to useRun

Marfuen
and
claude
committed
64a2da3
feat(onboarding): split tracker into 6 steps with separate mitigation tracking

Marfuen
and
claude
committed
a57fced
fix(onboarding): keep task alive for mitigations, redirect via metadata flag

Marfuen
and
claude
committed
b16e311
fix(onboarding): show vendor/risk count on creation steps

Marfuen
and
claude
committed
c67cbfe
fix(onboarding): sequential triggerAndWait for mitigations

Marfuen
and
claude
committed
7dd06da
fix(onboarding): link completed assessments to treatment-plan tab

Marfuen
and
claude
committed
27a7b8d
fix: address cubic review findings in template processor and linkage

Marfuen
and
claude
committed
cd50157
fix(onboarding): fix currentStep mismatch and treat assessing as active

Marfuen
and
claude
committed
da3618d

Navigation Menu

Repository navigation

perf(onboarding): optimize onboarding pipeline from ~5min to ~2min - #2777

perf(onboarding): optimize onboarding pipeline from ~5min to ~2min#2777
Marfuen merged 28 commits into
mainfrom
mariano/onboarding-perf

Commits on May 7, 2026

perf(onboarding): parallelize vendor/risk extraction, fire-and-forget policies

perf(onboarding): parallelize risk + vendor matching in linkage step

perf(onboarding): switch linkage reranker from gpt-5-mini to gemini-3-flash

chore: upgrade AI SDK to v6/v3, switch reranker to gemini-3-flash via gateway

chore: remove caBundleExtension from trigger config

fix(api): await convertToModelMessages in assistant-chat controller

fix(onboarding): restore tracker progress for parallel + fire-and-forget flow

fix(onboarding): handle pipeda and other framework conditionals in policy templates

fix(onboarding): restore expandable sections in tracker during background work

fix(onboarding): use tasks.batchTrigger for policies to preserve parent metadata

fix(onboarding): restore concurrencyKey for policy batch trigger

fix(onboarding): sync all tracker spinner animations

perf(onboarding): raise policy queue concurrency from 5 to 15

perf(onboarding): switch policy generation from gpt-5-mini to gemini-3-flash

perf(onboarding): policies first, gemini for all LLM calls, non-blocking vendor side-effects

perf(onboarding): replace LLM policy generation with programmatic template processor

fix(onboarding): switch tracker from useRealtimeRun to useRun with polling

feat(onboarding): add targeted LLM pass to rewrite policy cue lines

perf(onboarding): switch reranker to gemini-3.1-flash-lite-preview

fix(onboarding): use tasks.batchTrigger for vendor/risk mitigations

fix: switch onboarding status hooks from useRealtimeRun to useRun

feat(onboarding): split tracker into 6 steps with separate mitigation tracking

fix(onboarding): keep task alive for mitigations, redirect via metadata flag

fix(onboarding): show vendor/risk count on creation steps

fix(onboarding): sequential triggerAndWait for mitigations

fix(onboarding): link completed assessments to treatment-plan tab

fix: address cubic review findings in template processor and linkage

fix(onboarding): fix currentStep mismatch and treat assessing as active