ci.yml

name: CI

on:
  pull_request:
  push:
    branches: [main]
  # Callable, so publishing a release runs these same checks against the release commit rather than
  # trusting that they ran somewhere earlier.
  workflow_call:

# Least privilege by default. Jobs that need more must declare it locally.
permissions:
  contents: read

# The newest push on a branch wins; main runs are retained for badge and release history.
concurrency:
  group: ci-${{ github.ref }}
  cancel-in-progress: ${{ github.ref != 'refs/heads/main' }}

jobs:
  # Formatting, linting, typing and tests are independent checks, so each reports its own result.
  static:
    name: format, lint, types
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          # Checkout leaves a usable credential in the runner otherwise, which every later step and
          # every action it calls can read. Nothing here pushes.
          persist-credentials: false
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      - run: bun install --frozen-lockfile
      - run: bun run format:check
      - run: bun run lint
      - run: bun run typecheck
      # `typecheck` is `bun run --filter '*' typecheck`, and each workspace's `tsc --noEmit` reads
      # its own tsconfig, none of which includes a test tree. The suite is therefore the one part
      # of this repository that no type-checker has ever read: bun erases types rather than
      # checking them, so a test could pass an options object where a connection string belongs,
      # or import a type from a path that does not exist and silently get `any`, and every check
      # here stayed green. This reads `server/tests` under the server's own compiler options and
      # fails on any error that is not in `server/tests/typecheck-baseline.txt` — the errors that
      # were already there when the check was turned on. New ones are new defects.
      - name: server/tests types (baselined)
        run: bun run typecheck:tests
      - name: Desktop version tracks the release version
        run: bun desktop/scripts/desktop-version.ts check
      # The chart's default image tag is its appVersion, and a release must move both together. They
      # drifted once, appVersion left pointing at a tag that was never published, and a default
      # `helm install` sat in ImagePullBackOff. This refuses a tree where package.json (the last
      # version published) and the chart's appVersion have come apart.
      - name: Chart appVersion tracks the release version
        run: |
          pkg=$(bun -e 'console.log(require("./package.json").version)')
          chart=$(bun -e 'console.log(require("yaml").parse(require("fs").readFileSync("charts/openbot/Chart.yaml","utf8")).appVersion)')
          if [ "$pkg" != "$chart" ]; then
            echo "::error::package.json is $pkg but charts/openbot/Chart.yaml appVersion is $chart. The release workflow moves them together; bring them back into line."
            exit 1
          fi

  # The two packages that are deployables in their own right rather than root workspaces. Root
  # `typecheck` is `bun run --filter '*' typecheck`, and `--filter '*'` enumerates `workspaces`, which
  # is app, server and worker. So both of these ship a `typecheck` script that nothing has ever run:
  # `agent-computer` holds the only `spawn` in the deployment and `supervisor` is the only thing
  # holding a Docker socket, which is a poor pair to leave untyped. Both are clean today, so this
  # changes nothing about main and only stops the next change being the first one a type checker sees.
  #
  # Not by adding them to `workspaces`: each is built from its own lockfile and the Dockerfile depends
  # on that, so they are installed separately here for the same reason they are installed separately
  # there. A matrix rather than two jobs, so each package reports its own result and a third one is a
  # line in the list below and nothing in `verify`, which sees the whole matrix as one entry.
  deployables:
    name: types (${{ matrix.package }})
    runs-on: ubuntu-latest
    strategy:
      # One red package must not hide whether the other is red too.
      fail-fast: false
      matrix:
        package: [agent-computer, supervisor]
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      # Two installs, and the root one is load-bearing. Both packages set `types: ["bun"]` in
      # their tsconfig while `@types/bun` is a root devDependency, so tsc resolves it by walking up
      # to the root node_modules. Without it the typecheck fails with TS2688 on a fresh checkout.
      # This is the order the image uses for the same reason: Dockerfile installs at the root
      # before installing agent-computer.
      - run: bun install --frozen-lockfile
      - run: bun install --frozen-lockfile
        working-directory: ${{ matrix.package }}
      - run: bun run typecheck
        working-directory: ${{ matrix.package }}

  mobile:
    name: native types
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: actions/setup-node@949feb2413d6458794dcd2491c4babbbce0c15c1 # v7.1.0
        with:
          node-version: '22'
          cache: npm
          cache-dependency-path: mobile/package-lock.json
      # The native app is an isolated npm package; adding it to Bun workspaces would change the
      # existing desktop/server installation contract. CI installs its own reviewed lockfile.
      - run: npm ci --prefix mobile --ignore-scripts
      - run: npm run --prefix mobile typecheck

  # The agent-computer tests that launch a real Chromium. They are asked for by name because a
  # machine running `bun test` is not required to have a browser, and these are the only tests that
  # would catch a password reaching the snapshot a model reads, a session cookie lost on restart, or
  # WebRTC leaving around the egress filter. The browser is the one the computer image installs,
  # pinned by agent-computer's own lockfile.
  computer-browser:
    name: computer (real browser)
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      - run: bun install --frozen-lockfile
      - run: bun install --frozen-lockfile
        working-directory: agent-computer
      # `--no-shell` as the image does, so the headless run below proves nothing needs the shell.
      - run: bunx playwright install --with-deps --no-shell chromium
        working-directory: agent-computer
      # Headless here: these launch through the profile path without the computer's own X server.
      # The headless run of the Web Bot Auth test is also the one that checks the user agent fix.
      - run: >-
          bun test tests/sign-in.test.ts tests/session-cookies.test.ts tests/follows-popup.test.ts
          tests/webrtc-proxy.test.ts tests/web-bot-auth-browser.test.ts
          tests/handoff-page-restore.test.ts tests/enter-text.test.ts tests/person-browser.test.ts tests/person-browser-computer.test.ts
          tests/bot-browser.test.ts tests/bot-browser-computer.test.ts
          tests/screencast-keys-browser.test.ts tests/browser-memory.test.ts
        working-directory: agent-computer
        env:
          OPENBOT_COMPUTER_BROWSER: '1'
          COMPUTER_BROWSER_MODE: headless
      # Headed is the default for hosted computers, so it is tested as one: the computer starting its
      # own Xvfb, and the signing again with cross-site frames in processes of their own.
      - run: bun test tests/headed-browser.test.ts
        working-directory: agent-computer
        env:
          OPENBOT_HEADED_BROWSER: '1'
      - run: xvfb-run --auto-servernum bun test tests/web-bot-auth-browser.test.ts
        working-directory: agent-computer
        env:
          OPENBOT_COMPUTER_BROWSER: '1'
          COMPUTER_BROWSER_MODE: headed

  chart:
    name: chart (${{ matrix.target }})
    runs-on: ubuntu-latest
    strategy:
      # One red target must not hide whether another is red too.
      fail-fast: false
      matrix:
        target: [self-hosted, eks, eks-sandbox, gke, aks]
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
          # The new-values-key check compares this chart against the last released one, or against
          # main where the chart has not shipped yet. A shallow clone has neither to compare with.
          fetch-depth: 0
      - uses: azure/setup-helm@9bc31f4ebc9c6b171d7bfbaa5d006ae7abdb4310 # v5.0.1
        with:
          version: v3.19.0
      # For the coherence check below, which is a Bun script like everything else here.
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      # Nothing rendered this chart until now, which is how four values files that produce a server
      # unable to start were shipped and stayed shipped. Rendering is the cheap half; the refusals in
      # validation.yaml are the half that catches a missing value before a cluster does.
      - run: helm dependency build charts/openbot
      # Structure only. `helm lint` reports a template `fail` as an INFO line and exits 0 even under
      # `--strict`, which was driven and confirmed, so it cannot gate the refusals below. Rendering
      # does: `helm template` exits non-zero on one.
      - run: helm lint charts/openbot --values charts/openbot/ci/${{ matrix.target }}-values.yaml
      # A key encryption key is a real 32 bytes rather than a placeholder, because the chart checks
      # its shape. Generated here so no example key is ever a literal in this repository.
      - name: Render
        run: |
          helm template ci charts/openbot \
            --values charts/openbot/ci/${{ matrix.target }}-values.yaml \
            --set-string secrets.keyEncryptionKey="$(openssl rand -base64 32)" \
            --api-versions agents.x-k8s.io/v1beta1/Sandbox \
            --api-versions extensions.agents.x-k8s.io/v1beta1/SandboxTemplate \
            > rendered.yaml
      # Rendering proves the templates run. This proves the result is coherent, which is a different
      # question: every secret key a container demands has to be one the chart actually writes.
      # Getting that wrong is invisible until a pod starts, and every shipped target had it wrong.
      - run: bun scripts/check-rendered-chart.ts rendered.yaml
      # No target names a Learning credential, so the setup Job renders nowhere above. Render it with
      # one of each, and hold it to the same coherence check.
      - name: Learning setup Job renders
        run: |
          helm template ci charts/openbot \
            --values charts/openbot/ci/${{ matrix.target }}-values.yaml \
            --set-string secrets.keyEncryptionKey="$(openssl rand -base64 32)" \
            --api-versions agents.x-k8s.io/v1beta1/Sandbox \
            --api-versions extensions.agents.x-k8s.io/v1beta1/SandboxTemplate \
            --set-string config.learning.setup.cliLoginSecret.name=openbot-copilotkit-cli \
            --set-string config.learning.setup.serviceTokenSecret.name=intelligence-service-token \
            --set-string config.learning.setup.projectId=7 \
            > rendered-learning.yaml
          grep -q 'app.kubernetes.io/component: learning-setup' rendered-learning.yaml
          bun scripts/check-rendered-chart.ts rendered-learning.yaml
      # And that the refusals are load-bearing rather than decorative. A chart full of `fail`
      # messages nothing ever triggers is a chart that has never been shown to refuse anything, and
      # every one of these describes a state that shipped in a values file at some point.
      - name: Refusals fire
        run: |
          set -uo pipefail
          refuses() {
            local why="$1"; shift
            if helm template ci charts/openbot \
                 --values charts/openbot/ci/${{ matrix.target }}-values.yaml \
                 --set-string secrets.keyEncryptionKey="$(openssl rand -base64 32)" \
                 --api-versions agents.x-k8s.io/v1beta1/Sandbox \
                 --api-versions extensions.agents.x-k8s.io/v1beta1/SandboxTemplate \
                 "$@" >/dev/null 2>&1; then
              echo "::error::The chart rendered $why, which it is supposed to refuse."
              return 1
            fi
            echo "refused: $why"
          }
          # The public example key, which the server will not start with. Only where this chart
          # holds the secret: with a store, the value is not readable at template time, so the
          # refusal is deliberately not armed and asserting it here would be asserting a bug.
          if ! grep -qE '^ *enabled: true' <(sed -n '/^externalSecrets:/,/^[a-z]/p' charts/openbot/ci/${{ matrix.target }}-values.yaml); then
            refuses "the public example encryption key" \
              --set-string secrets.keyEncryptionKey="$(head -c 32 /dev/zero | base64)"
          else
            echo "skipped: the example-key refusal is not armed when the secret comes from a store"
          fi
          # A Bot endpoint with nothing on the request that says who is calling. Armed whether the
          # secret is this chart's or a store's, because the key list is readable either way.
          refuses "a managed agent URL with no token" \
            --set-string config.managedAgent.url=http://agent.default:8000/ag-ui
          # A browser inside every replica of a replicated API.
          refuses "an embedded browser across several replicas" \
            --set server.embeddedComputer=true --set server.replicaCount=2
          # A Bot's egress proxy on a port the computer's own network policy does not allow. The
          # variables reach the computer through extraEnv, so nothing else notices that the policy
          # then refuses to let it be reached.
          refuses "an egress proxy the network policy blocks" \
            --set networkPolicy.enabled=true \
            --set computers.extraEnv[0].name=EGRESS_PROXY_DEFAULT \
            --set-string computers.extraEnv[0].value=http://proxy.internal:3128
          # The residential proxy setting, whose address is a secret the template cannot read, with
          # network policies on and no path opened to it.
          refuses "a residential proxy with no path through the network policy" \
            --set networkPolicy.enabled=true --set computers.egressProxy.enabled=true \
            --set-string secrets.egressProxyUrl=http://user:pass@gateway.example:8000
          # The same setting switched on with no URL behind it, where this chart holds the secret.
          if ! grep -qE '^ *enabled: true' <(sed -n '/^externalSecrets:/,/^[a-z]/p' charts/openbot/ci/${{ matrix.target }}-values.yaml) \
             && ! grep -qE '^ *existingSecret: [^"]' <(sed -n '/^secrets:/,/^[a-z]/p' charts/openbot/ci/${{ matrix.target }}-values.yaml); then
            refuses "a residential proxy switched on with no URL" \
              --set computers.egressProxy.enabled=true
          else
            echo "skipped: the proxy URL refusal is armed only where this chart writes the Secret"
          fi
          # A delivery or directory switch with no secret behind it. Armed on every target: the chart
          # requires the value when it holds the secret, and requires the store to name the key when
          # a store does.
          refuses "OpenTag switched on with no shared secret" \
            --set config.opentag.enabled=true
          refuses "SCIM switched on with no bearer token" \
            --set config.scim.enabled=true
          refuses "a Learning service token with no project to create the container in" \
            --set-string config.learning.setup.serviceTokenSecret.name=intelligence-service-token
          # The rest are refusals the server makes at boot, mirrored so the value is named at install.
          refuses "an OpenTag URL with the pairing switched off" \
            --set-string config.opentag.url=https://opentag.example.com
          refuses "a partial set of Twilio settings" \
            --set config.sms.enabled=true --set-string secrets.twilioAuthToken=token \
            --set-string config.sms.accountSid=AC0 --set-string config.sms.verifyServiceSid=VA0
          refuses "a push project id that is not a UUID" \
            --set-string config.push.projectId=not-a-uuid
          refuses "a delivery public URL that is not http(s)" \
            --set-string config.deliveryPublicUrl=hooks.example.com
          # A warm pool nothing claims from. Only meaningful where the target asks for per-Bot
          # computers; on the others the mode is not sandbox and the refusal is not armed.
          if grep -qE '^ *mode: sandbox' charts/openbot/ci/${{ matrix.target }}-values.yaml; then
            refuses "a warm pool no Bot can be handed a computer from" \
              --set computers.sandbox.warmPool.enabled=true
          else
            echo "skipped: the warm-pool refusal is only armed for computers.mode: sandbox"
          fi
      # And that a values key this chart did not used to have still renders when it is absent.
      #
      # `helm upgrade --reuse-values` takes the previous release's computed values rather than
      # merging the new chart's defaults, so a key added by the release being installed is missing on
      # every deployment that already exists. Unguarded that is a nil dereference that fails the
      # whole render, or an empty scalar Kubernetes reads as unset. Both shipped: one was found in
      # review, the other by a live upgrade after the first had been fixed one key over.
      - name: A new values key can be absent
        run: bun scripts/check-new-values-keys.ts charts/openbot/ci/${{ matrix.target }}-values.yaml

  test:
    name: tests
    runs-on: ubuntu-latest
    # The suite includes a real database integration test. Without a database it fails on every run,
    # including on main, which trains everyone to read a red CI as normal. pgvector rather than plain
    # postgres because the knowledge schema uses the extension.
    services:
      postgres:
        image: pgvector/pgvector:pg17
        env:
          POSTGRES_DB: openbot_test
          POSTGRES_USER: openbot
          POSTGRES_PASSWORD: openbot
        ports:
          - 5432:5432
        options: >-
          --health-cmd "pg_isready -U openbot -d openbot_test"
          --health-interval 5s
          --health-timeout 5s
          --health-retries 10
    env:
      DATABASE_URL: postgres://openbot:openbot@localhost:5432/openbot_test
      TEST_DATABASE_URL: postgres://openbot:openbot@localhost:5432/openbot_test
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      - run: bun install --frozen-lockfile
      # The Bots are not root workspaces and each keeps its own lockfile. The startup tests spawn
      # the real entrypoint, so its dependencies have to be installed as well.
      - run: bun install --frozen-lockfile
        working-directory: agent-bot
      # Same for the LangGraph Bot: its history tests import @langchain/core, which lives in that
      # Bot's own tree and not in the root one.
      - run: bun install --frozen-lockfile
        working-directory: agent-langgraph
      # Same for the computer: its Web Bot Auth tests import `web-bot-auth`, which lives in
      # agent-computer's own tree and not in the root one.
      - run: bun install --frozen-lockfile
        working-directory: agent-computer
      # Same for the Mastra Bot: root test discovery imports its receiver tests, which load Mastra
      # from that Bot's own dependency tree.
      - run: bun install
        working-directory: agent-mastra
      # Same for the desktop app: root test discovery imports its React tests, whose JSX runtime is
      # pinned by the desktop lockfile rather than the root one.
      - run: bun install --frozen-lockfile
        working-directory: desktop
      # The same migrator a deployment runs, rather than `drizzle-kit migrate`: it is the one that
      # refuses a database holding a stamp newer than the migrations it is about to walk past
      # (server/src/db/migration-reach.ts), and a CI database that is ever reused rather than
      # recreated is one of the places that can happen. Not the db:migrate script, which loads
      # ../.env; that file does not exist in CI, and DATABASE_URL comes from the job env instead.
      - run: bun scripts/migrate.ts
        working-directory: server
      # A passing job must include the expected test floor. Import-time failures can otherwise skip
      # files before their tests are registered.
      - run: bun run test:ci

  startup:
    name: startup (${{ matrix.os }})
    runs-on: ${{ matrix.os }}
    strategy:
      fail-fast: false
      matrix:
        os: [macos-latest, windows-latest]
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      - run: bun install --frozen-lockfile
      - run: bun test app/tests/build-cache.test.ts

  python-harness:
    name: python harness regressions
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: actions/setup-python@6d4fd22b20465ccc0f7354476588e8df7c9ca0e0 # v7.1.0
        with:
          python-version: "3.12"
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      - name: CrewAI provider-boundary regression
        run: |
          set -euo pipefail
          python -m venv .venv-crewai
          . .venv-crewai/bin/activate
          python -m pip install --requirement agent-crewai/requirements.txt --requirement agent-crewai/requirements-test.txt
          python -m pytest agent-crewai/tests -q
      - name: Claude client-tool continuation regressions
        run: |
          set -euo pipefail
          python -m venv .venv-claude-sdk
          . .venv-claude-sdk/bin/activate
          python -m pip install --requirement agent-claude-sdk/requirements.txt --requirement agent-claude-sdk/requirements-test.txt
          python -m pytest agent-claude-sdk/tests -q
      - name: LangGraph AG-UI provider-boundary regressions
        run: |
          set -euo pipefail
          python -m venv .venv-langgraph-agui
          . .venv-langgraph-agui/bin/activate
          python -m pip install --requirement agent-langgraph-agui/requirements.txt --requirement agent-langgraph-agui/requirements-test.txt
          python -m pytest agent-langgraph-agui/tests -q
      - name: LlamaIndex model-choice regression
        run: |
          set -euo pipefail
          python -m venv .venv-llamaindex
          . .venv-llamaindex/bin/activate
          python -m pip install --requirement agent-llamaindex/requirements.txt --requirement agent-llamaindex/requirements-test.txt
          python -m pytest agent-llamaindex/tests -q
      - name: Agno model-choice regression
        run: |
          set -euo pipefail
          python -m venv .venv-agno
          . .venv-agno/bin/activate
          python -m pip install --requirement agent-agno/requirements.txt --requirement agent-agno/requirements-test.txt
          python -m pytest agent-agno/tests -q
      - name: Strands model-choice regression
        run: |
          set -euo pipefail
          python -m venv .venv-strands
          . .venv-strands/bin/activate
          python -m pip install --requirement agent-strands/requirements.txt --requirement agent-strands/requirements-test.txt
          python -m pytest agent-strands/tests -q
      - name: Pydantic AI model-choice regression
        run: |
          set -euo pipefail
          python -m venv .venv-pydantic-ai
          . .venv-pydantic-ai/bin/activate
          python -m pip install --requirement agent-pydantic-ai/requirements.txt --requirement agent-pydantic-ai/requirements-test.txt
          python -m pytest agent-pydantic-ai/tests -q
      - name: Google ADK tool-call regressions
        run: |
          set -euo pipefail
          python -m venv .venv-adk
          . .venv-adk/bin/activate
          python -m pip install --requirement agent-adk/requirements.txt --requirement agent-adk/requirements-test.txt
          python -m pytest agent-adk/tests -q
      - name: Antigravity provider pin and token gate
        run: |
          set -euo pipefail
          python -m venv .venv-antigravity
          . .venv-antigravity/bin/activate
          python -m pip install --requirement agent-antigravity/requirements.txt --requirement agent-antigravity/requirements-test.txt
          python -m pytest agent-antigravity/tests -q
      - name: Langroid model-choice and tool-call regressions
        run: |
          set -euo pipefail
          python -m venv .venv-langroid
          . .venv-langroid/bin/activate
          python -m pip install --requirement agent-langroid/requirements.txt --requirement agent-langroid/requirements-test.txt
          python -m pytest agent-langroid/tests -q
      - name: AG2 model-choice regression
        run: |
          set -euo pipefail
          python -m venv .venv-ag2
          . .venv-ag2/bin/activate
          python -m pip install --requirement agent-ag2/requirements.txt --requirement agent-ag2/requirements-test.txt
          python -m pytest agent-ag2/tests -q
      - name: Microsoft Agent Framework model-choice regression
        run: |
          set -euo pipefail
          python -m venv .venv-microsoft
          . .venv-microsoft/bin/activate
          python -m pip install --requirement agent-microsoft/requirements.txt --requirement agent-microsoft/requirements-test.txt
          python -m pytest agent-microsoft/tests -q
      - run: bun install --frozen-lockfile
      - run: bun test tests/compose.test.ts
      - run: docker compose --env-file /dev/null --profile harness config --format json >/dev/null
        env:
          PICKED_HARNESS_IMAGE: openbot-agent-langgraph-agui:test

  build:
    name: build
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      - run: bun install --frozen-lockfile
      - run: bun run build

  # Migration files and the schema they were generated from, checked against each other. A snapshot
  # that has drifted from the schema produces a migration nobody wrote, applied to somebody's
  # database on their next deploy. Neither command needs a running database, only the config.
  migrations:
    name: migrations
    runs-on: ubuntu-latest
    env:
      # drizzle.config.ts refuses to load without it. Nothing here connects.
      DATABASE_URL: postgres://openbot:openbot@localhost:5432/openbot
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: oven-sh/setup-bun@0c5077e51419868618aeaa5fe8019c62421857d6 # v2.2.0
        with:
          bun-version: 1.4.3
      - run: bun install --frozen-lockfile
      # Collisions and gaps between the migration files themselves.
      - run: bunx drizzle-kit check --config=drizzle.config.ts
        working-directory: server
      # And the other direction: a schema change with no migration written for it. `generate` writes a
      # file when it finds one, so the tree being dirty afterwards is the failure.
      - name: Schema has no unwritten migration
        working-directory: server
        run: |
          set -euo pipefail
          bunx drizzle-kit generate --config=drizzle.config.ts --name=ci_drift_probe
          if [ -n "$(git status --porcelain drizzle)" ]; then
            echo "::error::The schema has changed without a migration. Run drizzle-kit generate and commit it."
            git status --porcelain drizzle
            exit 1
          fi

  # The image is the artefact people deploy, and almost nothing about whether it works is visible to
  # the checks above. A dangling symlink, a supervised service that exits, a missing binary: all of
  # them typecheck, lint and test perfectly.
  image:
    name: image
    runs-on: ubuntu-latest
    # The only slow job: it builds the container and boots it, ~14 minutes. The four checks above are
    # seconds and stay on every push, because they catch most things cheaply. This one is the cost, so
    # on a pull request it runs only when an admin adds the `full-ci` label — the point at which the PR
    # is actually a merge candidate. On main and through the release workflow_call it always runs, so
    # nothing reaches a release without it. `verify` treats a skipped job as passing, so an unlabelled
    # PR is green on the cheap checks alone.
    if: >-
      github.event_name != 'pull_request' ||
      contains(github.event.pull_request.labels.*.name, 'full-ci')
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: docker/setup-buildx-action@f87e5991a6d7451dcb8d9637bfbc97413f497069 # v4.4.1
      # Loaded rather than pushed: this runs on pull requests, including from forks, and it proves
      # the image builds without granting anything the ability to publish one.
      - uses: docker/build-push-action@c3c9e263c25d99ce0380d002d59b67737d91b0dc # v7.4.0
        with:
          context: .
          load: true
          tags: openbot:ci
          cache-from: type=gha
          cache-to: type=gha,mode=max
      # Building it says the layers resolve. Running it says the supervision tree comes up and stays
      # up, which is a different claim and the one that has broken before.
      - name: The image boots and serves
        run: |
          set -euo pipefail
          # `OPENBOT_SINGLE_USER=true` because this boots a deployment with no identity provider, and
          # the image sets NODE_ENV=production, where that combination refuses to start rather than
          # serve an open deployment. Saying so is the point: the flag is how somebody declares they
          # meant it. This was `OPENBOT_DEV_NO_AUTH=1` before, which the code never accepted at all,
          # since it compares against the exact string "true"; it did nothing and nothing noticed.
          #
          # Placeholders, not secrets. `loadConfig` refuses to start without Intelligence and a
          # licence configured, but it only checks that they are present and well-formed; nothing is
          # contacted at start-up and /api/capabilities reads config alone. So this proves the image
          # boots and serves without needing a licence, which is the part CI cannot have: a licence
          # is bound to the machine it was issued for. Whether Intelligence actually answers is what
          # the smoke journey checks, on a real deployment.
          docker run -d --name openbot-ci -p 3001:3001 \
            -e EMBEDDED_POSTGRES=on \
            -e KEY_ENCRYPTION_KEY="$(openssl rand -base64 32)" \
            -e TRUSTED_ORIGINS=http://localhost:3001 \
             -e OPENBOT_SINGLE_USER=true \
             -e INTELLIGENCE_API_URL=https://api.intelligence.copilotkit.ai \
            -e INTELLIGENCE_GATEWAY_WS_URL=wss://realtime.intelligence.copilotkit.ai \
            -e INTELLIGENCE_API_KEY=ci-not-a-real-key \
            openbot:ci
          for attempt in $(seq 1 150); do
            if curl -fsS http://localhost:3001/api/capabilities >/dev/null 2>&1; then
              echo "answered after ${attempt}s"
              exit 0
            fi
            if [ -z "$(docker ps -q -f name=openbot-ci)" ]; then
              echo "::error::The container exited before it answered."
              docker logs openbot-ci
              exit 1
            fi
            sleep 1
          done
          echo "::error::No answer on /api/capabilities within 150s."
          docker logs openbot-ci
          exit 1
      - name: Supervision tree is stable, not respawning
        run: |
          set -euo pipefail
          # A supervised service that exits is restarted forever. That looks healthy from outside for
          # as long as something else is answering, so the log is where it shows.
          sleep 15
          if docker logs openbot-ci 2>&1 | grep -Eic 'restarting|respawn' | grep -qv '^0$'; then
            echo "::error::A supervised service is restarting."
            docker logs openbot-ci 2>&1 | grep -Ei 'restarting|respawn' | head -20
            exit 1
          fi
          test -n "$(docker ps -q -f name=openbot-ci)" || {
            echo "::error::The container is no longer running after 15s."
            docker logs openbot-ci
            exit 1
          }
      - name: A Bot's shell cannot reach the deployment secrets
        run: |
          set -uo pipefail
          if docker exec -u pwuser openbot-ci env -i PATH=/usr/bin:/bin sh -c '
                for e in /proc/[0-9]*/environ; do tr "\0" "\n" < "$e" 2>/dev/null; done
              ' | grep -q "^KEY_ENCRYPTION_KEY="; then
            echo "::error::A pwuser shell reads KEY_ENCRYPTION_KEY from a sibling process /proc/<pid>/environ."
            exit 1
          fi
          if docker exec -u pwuser openbot-ci env -i PATH=/usr/bin:/bin sh -c \
                'cat /run/s6/container_environment/KEY_ENCRYPTION_KEY 2>/dev/null' | grep -q .; then
            echo "::error::A pwuser shell reads KEY_ENCRYPTION_KEY from /run/s6/container_environment."
            exit 1
          fi
          echo "pwuser cannot reach KEY_ENCRYPTION_KEY"
      - if: always()
        run: docker rm -f openbot-ci >/dev/null 2>&1 || true

  # The five images the release publishes alongside the one above. Nothing else here builds them:
  # they are built by `docker compose up --build`, which only the smoke journey runs and which
  # cannot run in CI. So a broken `supervisor/Dockerfile` used to surface during a release, after
  # the `openbot` image had already been pushed, which is the one failure this repository calls out
  # as leaving a published image with nothing pointing at it.
  #
  # Built, not pushed, and one architecture: this answers whether the Dockerfile resolves. Whether
  # it resolves on arm64 too is answered at release time, on an arm64 runner.
  #
  # No registry cache, so a release builds these a second time. `type=gha` needs ACTIONS_* in the
  # environment, which `build-push-action` injects and a plain `run:` step does not have, and
  # reaching for another action to export them would buy minutes in a job that runs rarely at the
  # price of a dependency in the one workflow whose inputs are a security boundary.
  component-dockerfiles:
    name: component dockerfiles
    runs-on: ubuntu-latest
    # The same cost gate as `image` above, and for the same reason: `agent-computer` pulls
    # Playwright's base and `server` builds the app, so this is minutes rather than seconds. On main
    # and through the release's workflow_call it always runs.
    if: >-
      github.event_name != 'pull_request' ||
      contains(github.event.pull_request.labels.*.name, 'full-ci')
    steps:
      - uses: actions/checkout@3d3c42e5aac5ba805825da76410c181273ba90b1 # v7.0.1
        with:
          persist-credentials: false
      - uses: docker/setup-buildx-action@f87e5991a6d7451dcb8d9637bfbc97413f497069 # v4.4.1
      # A Dockerfile the release does not publish is one a machine would build from source, so the
      # list and the tree have to agree. Checked here rather than trusted, because the failure is
      # silent: the release succeeds and a laptop somewhere compiles the app.
      - name: Every buildable service is published
        run: |
          set -euo pipefail
          found="$(find . -mindepth 2 -maxdepth 2 -name Dockerfile -not -path './node_modules/*' \
            | sed -e 's|^\./||' -e 's|/Dockerfile$||' | sort)"
          published="$(jq -r '.[]' .github/published-images.json | sort)"
          if [ "$found" != "$published" ]; then
            echo "::error::.github/published-images.json does not match the Dockerfiles in the tree."
            diff <(echo "$published") <(echo "$found") || true
            exit 1
          fi
      - name: They build
        run: |
          set -euo pipefail
          while read -r image; do
            echo "::group::$image"
            docker buildx build --file "$image/Dockerfile" --platform linux/amd64 .
            echo "::endgroup::"
          done < <(jq -r '.[]' .github/published-images.json)
      # Building proves the Dockerfile resolves, not that the image starts. A runtime import of a file
      # the Dockerfile never copied (../../shared/file-download, from #698) built cleanly and crashed
      # on start. The computer image has no dependencies to start, so it is booted and asked for health.
      - name: The computer image starts
        run: |
          set -euo pipefail
          docker buildx build --file agent-computer/Dockerfile --platform linux/amd64 --load \
            --tag openbot-agent-computer:ci .
          docker run -d --name computer-ci -e COMPUTER_TOKEN=ci-token-0123456789abcdef \
            -p 127.0.0.1:4100:4100 openbot-agent-computer:ci
          for _ in $(seq 1 60); do
            if curl -fsS http://127.0.0.1:4100/health >/dev/null; then
              echo "agent-computer answered /health"
              exit 0
            fi
            if [ "$(docker inspect -f '{{.State.Running}}' computer-ci)" != "true" ]; then break; fi
            sleep 1
          done
          echo "::error::agent-computer did not answer /health"
          docker logs computer-ci || true
          exit 1

  # One check for branch protection to require. A new job above is covered by this without anybody
  # remembering to add it to a list, and a job that was skipped for the wrong reason is not a pass.
  verify:
    name: verify
    runs-on: ubuntu-latest
    if: always()
    needs: [static, deployables, mobile, computer-browser, chart, test, startup, python-harness, build, migrations, image, component-dockerfiles]
    steps:
      - name: Require every check
        env:
          RESULTS: ${{ join(needs.*.result, ' ') }}
        run: |
          set -euo pipefail
          echo "$RESULTS"
          for result in $RESULTS; do
            case "$result" in
              success|skipped) ;;
              *) echo "::error::A required check reported $result"; exit 1 ;;
            esac
          done