ci: use cheaper review model and cache pipeline installs #15

Merged
Noel merged 1 commits from feat/upgrade-ci-cd-pipeline into main 2026-10-20 19:05:35 +00:00
5 changed files with 110 additions and 44 deletions
Showing only changes of commit fa8fafb7ef - Show all commits

No files matched your search

+66 -33
View File
@@ -17,9 +17,13 @@ env:
APP_DIR: .
jobs:
# ---------------------------------------------------------------- lint ----
lint:
name: ESLint
# ------------------------------------------------------ lint + typecheck --
# One job, not two: both checks use the same image, the same lockfile and
# the same install, so splitting them doubled the install work (and, on the
# 2 vCPU runner, the CPU contention). They still report as separate steps,
# so a failure says which check failed.
checks:
name: Lint + typecheck
runs-on: ubuntu-latest
container: node:22-bookworm
steps:
@@ -32,6 +36,32 @@ jobs:
# missing or empty")
version: 11
- name: Resolve pnpm store and lockfile hash
id: pnpm
working-directory: ${{ env.APP_DIR }}
run: |
# hashFiles() evaluates empty in this act runner, so the cache key
# is built from a real sha256 in a shell step instead. The store
# directory's own name is the layout version (store/v11 here), so
# it is in the key: a store from another pnpm major is unusable.
store="$(pnpm store path --silent)"
echo "store=$store" >> "$GITHUB_OUTPUT"
echo "layout=$(basename "$store")" >> "$GITHUB_OUTPUT"
echo "lockfile=$(sha256sum pnpm-lock.yaml | cut -d' ' -f1)" >> "$GITHUB_OUTPUT"
- name: Restore pnpm store
# v4.1.x, not v4.2+: the runner's cache server may only speak the v1
# API (Gitea runner 2.x). v4.1 works on both v1-only and v3 runners.
# A cache outage must not turn into a red PR: install still works
# without the store, just slower.
continue-on-error: true
uses: actions/cache@v4.1.0
with:
path: ${{ steps.pnpm.outputs.store }}
key: pnpm-store-${{ runner.os }}-${{ steps.pnpm.outputs.layout }}-${{ steps.pnpm.outputs.lockfile }}
restore-keys: |
pnpm-store-${{ runner.os }}-${{ steps.pnpm.outputs.layout }}-
- name: Install dependencies
working-directory: ${{ env.APP_DIR }}
run: pnpm install --frozen-lockfile
@@ -40,25 +70,6 @@ jobs:
working-directory: ${{ env.APP_DIR }}
run: pnpm lint
# ------------------------------------------------------------ typecheck ---
typecheck:
name: Typecheck
runs-on: ubuntu-latest
container: node:22-bookworm
steps:
- uses: actions/checkout@v4
- uses: pnpm/action-setup@v4
with:
# match the pnpm major version used locally (v11.x) - older pnpm
# rejects our config-only pnpm-workspace.yaml ("packages field
# missing or empty")
version: 11
- name: Install dependencies
working-directory: ${{ env.APP_DIR }}
run: pnpm install --frozen-lockfile
- name: Typecheck (vue-tsc)
working-directory: ${{ env.APP_DIR }}
run: pnpm typecheck
@@ -234,14 +245,33 @@ jobs:
with:
fetch-depth: 0
- name: Restore opencode binary
id: opencode-cache
# The installer writes to $HOME/.opencode/bin. Caching it saves the
# ~4s download+extract per run - small next to the model call, but
# free to keep. Key includes the pinned version, so bumping the
# version below invalidates the cached binary automatically.
# As above, a cache outage falls back to the installer.
continue-on-error: true
uses: actions/cache@v4.1.0
with:
path: ~/.opencode/bin
key: opencode-v2.0.32-${{ runner.os }}
- name: Install opencode CLI
if: steps.opencode-cache.outputs.cache-hit != 'true'
run: curl -fsSL https://opencode.ai/v2/install | bash -s -- -v 2.0.32
# v2 installer is required: the classic /install URL ships only
# OpenCode 1.x (resolved to v1.19.3 on arm64), which cannot resolve
# model variants like "#max" and failed with ProviderModelNotFoundError.
# model variants like "#medium" and failed with ProviderModelNotFoundError.
# Version pinned (-v 2.0.32) so a breaking v2 release cannot fail
# the job mid-PR; bump deliberately, not automatically.
- name: Add opencode to PATH
# The installer appends the directory to $GITHUB_PATH itself, but that
# step is skipped on a cache hit. Add it unconditionally.
run: echo "$HOME/.opencode/bin" >> "$GITHUB_PATH"
- name: Get PR diff
run: |
# lockfile changes bloat the diff (175KB here) and are pointless to
@@ -258,15 +288,18 @@ jobs:
env:
OPENCODE_API_KEY: ${{ secrets.OPENCODE_API_KEY }}
run: |
# 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').
# opencode-go/claude-haiku-5-5#medium: cheap ($0.10/$0.50 per M
# tokens) and, unlike deepseek-v4.1-flash, it has a #medium
# reasoning variant - flash only exposes low/high/max, so every
# review there ran at maximum effort. #medium keeps the review
# competent without paying for max reasoning on each push.
# History: deepseek-v4-pro#max -> deepseek-v4.1-flash#max ->
# haiku-5-5#medium (2026-10-20). The pro endpoint failed two
# consecutive runs with "Upstream request failed: Endpoint is
# unavailable" before the flash switch; that fallback is recorded
# in OKF_Nuxt/open-questions.md. 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:
@@ -300,7 +333,7 @@ jobs:
EXIT_CODE=0
opencode run \
--print-logs \
--model opencode-go/deepseek-v4.1-flash#max \
--model opencode-go/claude-haiku-5-5#medium \
"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.
+18 -6
View File
@@ -1,7 +1,7 @@
---
type: Decision
title: AI code review setup
description: The review job runs opencode with the opencode-go plan model, env-only auth, a read-only permission set, and posts only reviews that end with a verdict line.
description: The review job runs opencode with a cheap medium-effort opencode-go model, env-only auth, a read-only permission set, and posts only reviews that end with a verdict line.
tags: [decision, ci, ai, opencode, security]
status: draft
generated: { by: opencode-go/glm-5.3-flash, at: 2026-10-18T16:05:00Z }
@@ -19,7 +19,7 @@ sources:
The review job installs opencode v2 from the dedicated v2 installer,
authenticates with `OPENCODE_API_KEY` from the environment only, runs
`opencode-go/deepseek-v4-pro#max` (covered by the Go plan subscription,
`opencode-go/claude-haiku-5-5#medium` (covered by the Go plan subscription,
no per-token cost), denies `edit`/`shell`/`webfetch`/`websearch` via a
global permission config written outside the checkout, and posts the
review as a PR comment only if the output ends with a `VERDICT:` line.
@@ -29,6 +29,7 @@ review as a PR comment only if the output ends with a `VERDICT:` line.
| Piece | Why |
|---|---|
| Provider prefix `opencode-go`, not `opencode` | `opencode` is the pay-per-call Console gateway; the Go plan lives under its own provider id. The wrong prefix failed every run with "Insufficient account funds". |
| Model `claude-haiku-5-5#medium` | Cheap per token and cheap in the usage window, with a real `#medium` reasoning level. `deepseek-v4.1-flash` only exposes low/high/max, so its reviews always paid for the maximum effort. |
| The v2 installer (`/v2/install`) | The classic installer ships OpenCode 1.x (v1.19.3 on arm64), which cannot resolve model variants and failed with `ProviderModelNotFoundError` for `#max`. |
| Env-only auth, no `opencode.json` in the checkout | A config file in the workspace holds the API key in plaintext where the review agent can read it; a malicious PR diff could prompt-inject the agent into quoting the key in the publicly visible review comment. Environment variables are invisible to file reads. |
| Deny write tools via a global config outside the checkout | Headless runs cannot answer permission prompts; denials prevent both hangs and repo modification. The config file itself holds no secret, so its location leaks nothing. |
@@ -54,17 +55,28 @@ derives the verdict from the findings, so the gate and the prompt agree.
* Fresh-container replica test (node:22-bookworm, key-only auth)
reproduced the CI failure exactly, then confirmed the v2 + `#max` fix
before it was pushed.
* Model history: `deepseek-v4-pro#max` was replaced by
`deepseek-v4.1-flash#max` after two consecutive endpoint outages, then
by `claude-haiku-5-5#medium` (2026-10-20) to stop paying maximum
reasoning effort on every review. The flash model had no `#medium`
variant to fall back to.
# Not verified
* Whether `#max` still resolves after a future CLI upgrade — the variant
list comes from the model catalog, not a contract.
* Long-run behaviour of Go plan usage windows ($3 per 5 hours for
DeepSeek V4 Pro) under real review frequency.
* Whether `#medium` still resolves after a future CLI upgrade — the
variant list comes from the model catalog, not a contract.
* Whether the smaller model still catches the same class of findings as
the first flash review. The PR #14 findings are the baseline; re-run
the eval on the next few PRs.
* Long-run behaviour of Go plan usage windows under real review
frequency, now that the model is cheaper.
# Revisit when
* A PR review fails again with a new error class — check the provider
prefix, the CLI version, and the usage limits first; they caused every
failure so far.
* A flash-tier model gains a `#medium` variant and beats the current
model on the PR #14 eval set — the present pick is a cost/quality
compromise, not a permanent one.
* Go Plus or pay-as-you-go balance becomes relevant.
+22 -4
View File
@@ -22,9 +22,10 @@ Every pull request is checked by three layers, ordered by cost:
1. **Pre-commit hooks** (local, seconds, deterministic): gitleaks, file
hygiene, conventional-commit check, actionlint, eslint --fix on staged
files. A commit that fails here never exists.
2. **Gitea Actions** (per PR push, minutes): ESLint, vue-tsc typecheck,
gitleaks over full history, pnpm audit (blocking), droast Dockerfile
lint, Snyk (currently disabled).
2. **Gitea Actions** (per PR push, minutes): ESLint + vue-tsc typecheck
in one job (one install, cached pnpm store), gitleaks over full
history, pnpm audit (blocking), droast Dockerfile lint, Snyk
(currently disabled).
3. **AI code review** (per PR push): opencode with a plan-covered model,
verdict-based, posted as a PR comment.
@@ -42,6 +43,23 @@ sees intent, contracts and cross-file consequences that no linter can. The
ordering also protects the budget: the AI review only runs for code that
already passed the deterministic gates.
# Caching
Two caches, both served by the runner's own cache server (no external
service, entries stay on the home server):
* **pnpm store**, keyed on the store layout version (v11) and a `sha256sum`
of `pnpm-lock.yaml`. Installs hardlink from the store instead of
downloading tarballs again. The key is built in a shell step because
`hashFiles()` evaluates empty in this act runner.
* **opencode binary**, keyed on the pinned CLI version (`~/.opencode/bin`).
Small next to the model call, but it removes a network fetch per review.
Lint and typecheck also share one job: same image, same lockfile, same
install. Splitting them duplicated the install and made both compete for
the runner's two vCPUs. A future `pnpm build` job reuses the same store
cache key.
# What the pipeline bought us in practice
Recorded so the reasoning survives: during the first week the pipeline
@@ -57,7 +75,7 @@ exposure in the review setup — each fixed where it was cheapest.
|---|---|
| Error logs posted as review comments | Posting gate is the output contract (a `VERDICT:` line), not the exit code |
| Vendor container images without arm64 manifests | Binaries installed arch-aware in `node:22-bookworm`, never vendor CI images |
| `hashFiles()` silently empty in the act runner | File discovery via `find` in a step, not workflow expressions |
| `hashFiles()` silently empty in the act runner | File discovery via `find` and cache keys via `sha256sum`, both in shell steps - never workflow expressions |
| Silent advisory failures | `continue-on-error` removed everywhere; a failed check is red |
| Supply-chain (curl-piped binaries) | Pinned versions, checksums pinned in the workflow |
+3
View File
@@ -1,5 +1,8 @@
# Update Log
## 2026-10-20 (cheaper review model, cached installs, merged checks)
* **Update**: [pr-checker.yml](../.gitea/workflows/pr-checker.yml) - the review model is now `opencode-go/claude-haiku-5-5#medium`: cheap per token and, unlike `deepseek-v4.1-flash`, it has a real `#medium` effort level (flash exposes only low/high/max). ESLint and typecheck are merged into one `checks` job with a single `pnpm install`, and the pnpm store plus the opencode binary are restored with `actions/cache@v4.1.0` (v1 cache API, so it also works on a v1-only runner). Cache keys are built from `sha256sum pnpm-lock.yaml`, the pnpm store layout (`v11`) and the opencode version in shell steps, because `hashFiles()` evaluates empty in this act runner. See [AI review setup](decisions/ai-review-setup.md) and [PR pipeline layering](decisions/pr-pipeline-layering.md).
## 2026-10-20 (first flash review validated, two findings applied)
* **Update**: PR #14 review (flash model, `VERDICT: APPROVE`) was fact-checked finding by finding. Applied: `<NuxtRouteAnnouncer />` restored in `app.vue` (this PR's diff had dropped it with the NuxtWelcome swap), and the review job now requires `github.event_name == 'pull_request'` (manual dispatch - the new snyk trigger - would otherwise launch a paid review with no PR to post to). Rejected with evidence: the reviewer's `postcss-selector-parser` 6.1.5 override suggestion - the 6.1.5 backport exists on npm, but the blocking audit still flags it (`<7.1.6` covers 6.x), verified empirically; 7.1.6 stays. Deferred as design decisions: mobile nav, aria-current via NuxtLink, sprite-to-SVG, font fallback stacks, `borderRadius.full` token, the fall-endpoint budget in the API contract.
+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)~~ - **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 | — |
| ~~Whether the Go plan's per-model usage windows survive real review frequency, and which model the review should use~~ - **decided** (2026-10-20): review model is `opencode-go/claude-haiku-5-5#medium` in [pr-checker.yml](../.gitea/workflows/pr-checker.yml). Cheap per token, and it has a real `#medium` effort level; `deepseek-v4.1-flash` only exposes low/high/max, so it always paid for maximum reasoning. The earlier pro -> flash switch (two consecutive `Endpoint is unavailable` failures) is in the [log](../log.md) | 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