fix(ci): switch review model to deepseek-v4.1-flash, pause snyk via workflow_dispatch
PR Check / Security scan (Snyk) (pull_request) Skipped
PR Check / ESLint (pull_request) Successful in 36s
PR Check / Typecheck (pull_request) Successful in 44s
PR Check / Secret scan (gitleaks) (pull_request) Successful in 6s
PR Check / Dependency audit (pull_request) Successful in 9s
PR Check / Dockerfile lint (droast) (pull_request) Successful in 5s
PR Check / AI code review (opencode) (pull_request) Successful in 5m12s

This commit is contained in:
Noel Sabia committed 2026-10-20 20:36:54 +02:00
1 parent db11806c03
commit 9b0f89ef38
3 files changed
+24 -10

No files matched your search

+17 -9
View File
@@ -3,6 +3,8 @@ name: PR Check
on:
pull_request:
types: [opened, synchronize, reopened]
# manual trigger: the only way the paused snyk job below can run
workflow_dispatch:
concurrency:
group: pr-check-${{ github.event.pull_request.number }}
@@ -174,15 +176,17 @@ jobs:
-print0 | xargs -0 -r droast --no-roast --min-severity warning
# ------------------------------------------------------- snyk security ----
# DISABLED: Snyk free plan only includes ~100 test credits/month, which the
# PAUSED: Snyk free plan only includes ~100 test credits/month, which the
# pipeline would burn through quickly (2 tests per PR push).
# To re-enable, delete the "if: false" line below. Consider also running
# Paused the actionlint-clean way: the job only runs on manual
# (workflow_dispatch) triggers, so pushes and PRs never burn credits.
# To re-enable fully, delete the if: line below. Consider also running
# this job only on PRs marked "ready for review" or on a weekly schedule
# instead of on every push.
snyk:
name: Security scan (Snyk)
runs-on: ubuntu-latest
if: false
if: github.event_name == 'workflow_dispatch'
# snyk/snyk:node is amd64-only ("no matching manifest for linux/arm64"),
# so install the CLI via npm in the multi-arch node image instead
container: node:22-bookworm
@@ -251,11 +255,15 @@ jobs:
env:
OPENCODE_API_KEY: ${{ secrets.OPENCODE_API_KEY }}
run: |
# opencode-go/deepseek-v4-pro#max: flagship reasoning model at
# maximum thinking budget - included in the opencode-go $10 plan
# (variants: high, max). NOTE the provider prefix must be
# "opencode-go", NOT "opencode" (Zen is a different, pay-per-call
# provider and fails with 'Insufficient account funds').
# opencode-go/deepseek-v4.1-flash#max: switched from
# deepseek-v4-pro#max on 2026-10-20 after two consecutive runs
# failed with "Upstream request failed: Endpoint is unavailable"
# (~10 min apart; the model is still listed active, the endpoint
# just does not serve). This is the default recorded in
# OKF_Nuxt/open-questions.md for exactly this case. NOTE the
# provider prefix must be "opencode-go", NOT "opencode" (Zen is a
# different, pay-per-call provider and fails with 'Insufficient
# account funds').
#
# Auth is env-only: the CLI picks up OPENCODE_API_KEY from the
# environment (verified headless). Deliberately NO opencode.json:
@@ -289,7 +297,7 @@ jobs:
EXIT_CODE=0
opencode run \
--print-logs \
--model opencode-go/deepseek-v4-pro#max \
--model opencode-go/deepseek-v4.1-flash#max \
"You are a principal engineer performing a rigorous code review on a
pull request. You have tool access: read repo files for context when
the diff alone is ambiguous, but review ONLY code changed in this PR.
+6
View File
@@ -1,5 +1,11 @@
# Update Log
## 2026-10-20 (snyk pause made actionlint-clean)
* **Update**: pr-checker.yml - the paused snyk job's `if: false` (flagged by actionlint as a constant condition once this PR touched the workflow file) became `if: github.event_name == 'workflow_dispatch'` plus a `workflow_dispatch:` trigger. Same pause semantics ([snyk paused](decisions/snyk-paused-and-secrets-policy.md)) - pushes/PRs burn no credits; the job is now runnable manually instead of dead code.
## 2026-10-20 (review model switched to flash after endpoint outage)
* **Update**: [open questions](open-questions.md) - the review-model question is decided: `opencode-go/deepseek-v4.1-flash#max` in [pr-checker.yml](../.gitea/workflows/pr-checker.yml). The `deepseek-v4-pro#max` endpoint answered two consecutive runs with `Upstream request failed: Endpoint is unavailable` (runs 67 and 68 on PR #14, ~10 min apart, model still listed active) - the recorded default ("switch to flash") was applied.
## 2026-10-20 (blocking audit catches the tailwind dep, first override)
* **Update**: [vulnerability audit policy](decisions/vulnerability-audit-policy.md) - the first PR run after adding `@nuxtjs/tailwindcss` failed the blocking audit on GHSA-rj75-hqrm-r3gf (moderate: `postcss-selector-parser` < 7.1.6, quadratic flat-selector parsing, pulled in by `tailwindcss` 3.4.19 → `postcss-nested` 6.2.0). Fixed per policy with the first `overrides` entry in `pnpm-workspace.yaml`, pinning 7.1.6 (already in the tree via `postcss-nesting`, build verified, audit green except the two reviewed highs). Same run: AI review job failed transiently (`Upstream request failed: Endpoint is unavailable` on `opencode-go/deepseek-v4-pro#max`; the model is still listed active, so retry on next push - the model-switch question stays open in [open questions](open-questions.md)).
+1 -1
View File
@@ -17,7 +17,7 @@ generated: { by: opencode-go/glm-5.3-flash, at: 2026-10-18T12:41:52Z }
| When and how often Snyk runs once re-enabled: weekly schedule, ready-for-review gate, or per-push | re-enabling the snyk job | Stay disabled. |
| Whether the AI review job should gate merges (required status check on the verdict) or stay advisory | branch protection setup | Advisory. The review comment is information; blocking merges on an AI verdict is a policy choice, not a technical one. |
| How the opencode CLI version in CI gets upgraded: manually bump the pinned 2.0.32, or automate when a new major appears | first review failure after an opencode release | Manual bump. Pinning exists so upgrades are decisions. |
| Whether the Go plan's per-model usage windows ($3 per 5 hours for DeepSeek V4 Pro) survive real review frequency, or the review model must move to `deepseek-v4.1-flash` ($80 window) | first quota-limited review | Switch to flash. |
| ~~Whether the Go plan's per-model usage windows ($3 per 5 hours for DeepSeek V4 Pro) survive real review frequency, or the review model must move to `deepseek-v4.1-flash` ($80 window)~~ - **decided** (2026-10-20): review model is `opencode-go/deepseek-v4.1-flash#max`; the pro endpoint failed two consecutive review runs with `Upstream request failed: Endpoint is unavailable` (~10 min apart), so the recorded default was applied | done | — |
| Removing the two reviewed audit ignores when `node-forge` or `braces` ship patched releases | next dependency update round | Watch for new releases. |
# The mount trade-off, in detail