Files
windmill/.github/workflows/ai-agent-tests.yml
centdix b0ddcf31e4 ci: add path-gated AI agent + ai_evals smoke workflows (#9640)
* ci: add path-gated AI agent integration tests workflow

Runs integration_tests/ai_agent_tests against real LLM providers
(Anthropic/OpenAI/Google) only when AI-agent backend code or the tests
change, since runs make paid LLM calls. Adds a conftest fixture that
skips provider-parametrized cases whose API keys are absent, so CI
exercises only the providers it has secrets for.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* ci: add path-gated ai_evals global-mode smoke workflow

Runs the global AI chat eval (global-test1) across one cheap model per
provider (Anthropic/OpenAI/Google/DeepSeek) only when the eval harness or
copilot chat code change, since runs make paid LLM calls. Builds Windmill
CE from source as the AI proxy; global tools/drafts run in the Vitest
bridge. Gates on the deterministic draft pipeline (run succeeded +
produced a draft + used write_script), not the variable LLM judge score.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* ci: run AI smokes on PR ready-for-review instead of every push

Switch the pull_request trigger from `synchronize` (every commit) to
`ready_for_review`, with a job guard skipping draft PRs, so the paid LLM
runs only fire when a PR is marked ready to merge (plus push-to-main and
manual dispatch).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(ai_evals): lazily load cli mode so non-cli evals skip the cli toolchain

The entrypoint eagerly imported modes/cli, which pulls the wmill CLI
guidance modules and their JSR deps (@cliffy/*). Global/flow/script/app
runs then crashed with "Cannot find module '@cliffy/ansi/colors'" when
the cli workspace deps were not installed. Import createCliModeRunner
dynamically inside runCliBenchmark instead.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* test(ai_agent): raise low max_completion_tokens to OpenAI's 16 minimum

OpenAI's /v1/responses rejects max_output_tokens < 16 with a 400, failing
test_low_max_tokens for openai. 16 still exercises a truncated response.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* ci: run ai_evals workflow on Node 22 for the frontend undici 8.x dep

The Vitest bridge loads frontend/node_modules/undici@8.x, which requires
Node >=22.19; Node 20 failed with "webidl.util.markAsUncloneable is not a
function" when loading vitest.config.ts.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(ai_evals): run frontend evals autonomously + give global-test1 more turns

Frontend evals (flow/script/app/global) ran the production chat prompt, which
assumes an interactive human — so cheaper models burned their turn budget
asking for confirmation, waiting for approval, or presenting a plan, sometimes
hitting maxTurns without producing a draft. Append a shared autonomy note in
baseEvalRunner (the path all frontend modes share, mirroring cli mode): act
directly on clear requests; only ask on genuinely ambiguous ones (preserving
the askUserQuestion cases). Also raise global-test1's maxTurns 8 -> 10 so a
model that over-explores still converges.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* ci(ai_evals): watch draft/prompt deps outside copilot/

The global eval runs production frontend code in-process, so the smoke's
behavior depends on files outside frontend/src/lib/components/copilot/**:
the draft model (userDraft.svelte.ts, userDraftDbSyncer.svelte.ts), script
inference (infer.ts), and the chat system prompts ($system_prompts ->
system_prompts/auto-generated). Add them to both push and PR path filters so
a change there actually triggers the smoke that gates on draft production.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix: skip direct provider tests without credentials

* feat: add ai evals skip judge flag

* fix: simplify ai evals ci gate

* fix: simplify ai evals smoke gate

* fix: handle ai eval workflow triggers

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
2026-06-22 12:41:31 +02:00

133 lines
4.9 KiB
YAML

name: AI Agent Integration Tests
# Exercises the AI agent flow path (preview_flow with `aiagent` modules) against
# real LLM providers. Runs only when AI-agent backend code or the tests change,
# because each run makes real (paid) LLM calls. To avoid spending on every commit,
# the PR side triggers only when a PR is marked ready for review (out of draft) —
# not on `synchronize` — plus push to main and manual dispatch.
on:
workflow_dispatch:
push:
branches: [main]
paths:
- "integration_tests/ai_agent_tests/**"
- "backend/windmill-ai/**"
- "backend/windmill-api/src/ai.rs"
- "backend/windmill-worker/src/ai_executor.rs"
- "backend/windmill-worker/src/ai/**"
- "backend/windmill-worker/src/memory_common.rs"
- "backend/windmill-common/src/flow_conversations.rs"
- ".github/workflows/ai-agent-tests.yml"
pull_request:
types: [opened, reopened, ready_for_review]
paths:
- "integration_tests/ai_agent_tests/**"
- "backend/windmill-ai/**"
- "backend/windmill-api/src/ai.rs"
- "backend/windmill-worker/src/ai_executor.rs"
- "backend/windmill-worker/src/ai/**"
- "backend/windmill-worker/src/memory_common.rs"
- "backend/windmill-common/src/flow_conversations.rs"
- ".github/workflows/ai-agent-tests.yml"
concurrency:
group: ai-agent-tests-${{ github.ref }}
cancel-in-progress: true
jobs:
ai_agent_e2e:
# Skip draft PRs; the `opened`/`reopened` types would otherwise fire while
# still a draft. `ready_for_review` always arrives non-draft.
if: github.event_name != 'pull_request' || github.event.pull_request.draft == false
runs-on: ubicloud-standard-16
services:
postgres:
image: postgres:16
ports:
- 5432:5432
env:
POSTGRES_DB: windmill
POSTGRES_PASSWORD: changeme
options: >-
--health-cmd pg_isready --health-interval 10s --health-timeout 5s
--health-retries 5
steps:
- uses: actions/checkout@v4
- uses: actions-rust-lang/setup-rust-toolchain@v1
with:
cache-workspaces: backend
toolchain: 1.93.0
- uses: oven-sh/setup-bun@v2
with:
bun-version: 1.3.10
- uses: actions/setup-node@v4
with:
node-version: "20"
- uses: actions/setup-python@v5
with:
python-version: "3.11"
# CE build (no enterprise/license needed for AI agents). `quickjs` powers
# flow input-transform JS eval; `mcp` is required by the deepwiki MCP tool
# test. Bun tool scripts run via the always-on worker (BUN_PATH).
- name: Build Windmill
working-directory: ./backend
env:
SQLX_OFFLINE: true
CARGO_BUILD_JOBS: 12
RUSTFLAGS: ""
run: cargo build --features quickjs,mcp
- name: Start Windmill
working-directory: ./backend
env:
DATABASE_URL: postgres://postgres:changeme@localhost:5432/windmill
BUN_PATH: bun
NODE_BIN_PATH: node
RUST_LOG: info
run: |
mkdir -p ../integration_tests/logs
./target/debug/windmill > ../integration_tests/logs/windmill.log 2>&1 &
echo "Waiting for Windmill to be ready..."
for i in $(seq 1 60); do
if curl -sf http://localhost:8000/api/version > /dev/null 2>&1; then
echo "Windmill is ready"
break
fi
sleep 2
done
curl -sf http://localhost:8000/api/version > /dev/null || { echo "Windmill failed to start"; tail -50 ../integration_tests/logs/windmill.log; exit 1; }
- name: Run AI agent integration tests
timeout-minutes: 20
working-directory: ./integration_tests/ai_agent_tests
env:
WINDMILL_URL: http://localhost:8000
# Only the providers we have org secrets for. Other providers
# (Azure, Bedrock, OpenRouter) are skipped by conftest when their
# keys are absent — see skip_provider_without_credentials.
ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
GOOGLE_AI_API_KEY: ${{ secrets.GOOGLE_API_KEY }}
run: |
python -m venv .venv
.venv/bin/pip install -r requirements.txt
# The S3/vision-attachment tests need MinIO large-file storage and
# image-capable provider setup; out of scope for this cost-controlled
# smoke. Add MinIO secrets + a storage service to enable them.
.venv/bin/python -m pytest -v \
--ignore=test_user_attachments.py \
--ignore=test_user_images.py \
--ignore=test_image_output.py
- name: Archive Windmill logs
uses: actions/upload-artifact@v4
if: always()
with:
name: ai-agent-tests-windmill-logs
path: integration_tests/logs