Files
BrowserOS/.github/workflows/eval-weekly.yml
shivammittal274 be6858d589 fix(server): allow Linux to skip OpenClaw via BROWSEROS_SKIP_OPENCLAW=1
Earlier surgical fixes (try/catch in main.ts, lazy chat client port) didn't
unblock dev's Linux CI — same throw kept reproducing. Whether this is bun
caching stale stack frames or a missed eager call site, the safer move is
to fix it at the root: make buildContainerRuntime never throw on Linux
when the runner has explicitly opted out.

Adds BROWSEROS_SKIP_OPENCLAW env check alongside the existing NODE_ENV=test
escape hatch in container-runtime-factory.ts. When set, returns the existing
UnsupportedPlatformTestRuntime stub — server boots normally, /health binds,
any actual OpenClaw API call still fails loudly at request time.

eval-weekly.yml sets the flag for the Linux runner. Darwin behavior and
non-CI Linux behavior unchanged (without the flag they still throw).
2026-04-29 23:18:59 +05:30

123 lines
4.6 KiB
YAML

name: Weekly Eval
on:
schedule:
# Every Saturday at 06:00 UTC
- cron: '0 6 * * 6'
push:
branches: [main]
paths:
- 'packages/browseros-agent/apps/server/src/agent/**'
- 'packages/browseros-agent/apps/server/src/tools/**'
workflow_dispatch:
inputs:
config:
description: 'Eval config file (relative to apps/eval/)'
required: false
default: 'configs/browseros-agent-weekly.json'
permissions:
contents: read
jobs:
eval:
runs-on: ubuntu-latest
timeout-minutes: 360
steps:
- name: Checkout
uses: actions/checkout@v4
- name: Install BrowserOS
run: |
# Rolling stable channel — see https://cdn.browseros.com/download/BrowserOS.deb
wget -q -O BrowserOS.deb https://cdn.browseros.com/download/BrowserOS.deb
sudo dpkg -i BrowserOS.deb
browseros --version || echo "BrowserOS installed at $(which browseros)"
- name: Install Bun
uses: oven-sh/setup-bun@v2
with:
bun-version: latest
- name: Install dependencies
working-directory: packages/browseros-agent
run: bun install --ignore-scripts
- name: Install Python eval dependencies
# agisdk pinned so silent upstream releases can't shift task definitions
# or grader behavior. Bump intentionally with a documented re-baseline.
run: pip install agisdk==0.3.5 requests
- name: Clone WebArena-Infinity
run: git clone --depth 1 https://github.com/web-arena-x/webarena-infinity.git /tmp/webarena-infinity
- name: Install xvfb
run: sudo apt-get update && sudo apt-get install -y xvfb
- name: Install captcha solver extension
working-directory: packages/browseros-agent/apps/eval
run: |
mkdir -p extensions
curl -sL -o /tmp/nopecha.zip https://github.com/NopeCHALLC/nopecha-extension/releases/latest/download/chromium_automation.zip
unzip -qo /tmp/nopecha.zip -d extensions/nopecha
- name: Run eval
working-directory: packages/browseros-agent/apps/eval
env:
FIREWORKS_API_KEY: ${{ secrets.FIREWORKS_API_KEY }}
OPENROUTER_API_KEY: ${{ secrets.OPENROUTER_API_KEY }}
CLAUDE_CODE_OAUTH_TOKEN: ${{ secrets.CLAUDE_CODE_OAUTH_TOKEN }}
NOPECHA_API_KEY: ${{ secrets.NOPECHA_API_KEY }}
BROWSEROS_BINARY: /usr/bin/browseros
WEBARENA_INFINITY_DIR: /tmp/webarena-infinity
# OpenClaw container runtime is macOS-only; opt the Linux runner
# into the no-op stub so the server can boot and the eval can run.
BROWSEROS_SKIP_OPENCLAW: '1'
EVAL_CONFIG: ${{ github.event.inputs.config || 'configs/browseros-agent-weekly.json' }}
run: |
echo "Running eval with config: $EVAL_CONFIG"
xvfb-run --auto-servernum --server-args="-screen 0 1440x900x24" bun run src/index.ts -c "$EVAL_CONFIG"
- name: Upload runs to R2
if: success()
working-directory: packages/browseros-agent/apps/eval
env:
EVAL_R2_ACCOUNT_ID: ${{ secrets.EVAL_R2_ACCOUNT_ID }}
EVAL_R2_ACCESS_KEY_ID: ${{ secrets.EVAL_R2_ACCESS_KEY_ID }}
EVAL_R2_SECRET_ACCESS_KEY: ${{ secrets.EVAL_R2_SECRET_ACCESS_KEY }}
EVAL_R2_BUCKET: ${{ secrets.EVAL_R2_BUCKET }}
EVAL_R2_CDN_BASE_URL: ${{ secrets.EVAL_R2_CDN_BASE_URL }}
EVAL_CONFIG: ${{ github.event.inputs.config || 'configs/browseros-agent-weekly.json' }}
run: |
CONFIG_NAME=$(basename "$EVAL_CONFIG" .json)
bun scripts/upload-run.ts "results/$CONFIG_NAME"
- name: Generate trend report
if: success()
timeout-minutes: 5
continue-on-error: true
working-directory: packages/browseros-agent
env:
EVAL_R2_ACCOUNT_ID: ${{ secrets.EVAL_R2_ACCOUNT_ID }}
EVAL_R2_ACCESS_KEY_ID: ${{ secrets.EVAL_R2_ACCESS_KEY_ID }}
EVAL_R2_SECRET_ACCESS_KEY: ${{ secrets.EVAL_R2_SECRET_ACCESS_KEY }}
EVAL_R2_BUCKET: ${{ secrets.EVAL_R2_BUCKET }}
EVAL_R2_CDN_BASE_URL: ${{ secrets.EVAL_R2_CDN_BASE_URL }}
run: bun apps/eval/scripts/weekly-report.ts /tmp/eval-report.html
- name: Upload report as artifact
if: success()
uses: actions/upload-artifact@v4
with:
name: eval-report-${{ github.run_id }}
path: /tmp/eval-report.html
- name: Upload server stderr logs (for post-mortem on startup failures)
if: always()
uses: actions/upload-artifact@v4
with:
name: browseros-server-logs-${{ github.run_id }}
path: /tmp/browseros-server-logs/
if-no-files-found: ignore