Taskfile.yml

---
# https://taskfile.dev
version: '3'
env :
  RENDER_DIR: .rendered
tasks:
  default:
    desc: "Show all available tasks"
    cmds:
      - task --list
    silent: true
  install:
    desc: "Install development dependencies for all agents into .venv_dev directory"
    cmds:
      - echo "Installing development dependencies for all agents"
      - uv sync --all-extras
      - mv pyproject.toml pyproject.toml.bak
      - uv run jinja -D agent_app_name DockerContext -D agent_template_framework docker template/\{\{agent_app_name\}\}/pyproject.toml.jinja -o pyproject.toml
      - UV_PROJECT_ENVIRONMENT=.venv_dev uv sync --all-extras
      - rm pyproject.toml
      - mv pyproject.toml.bak pyproject.toml
      - uv sync --all-extras

  install-e2e:
    desc: "Install dependencies needed to run E2E tests (prefers active venv when activated)"
    cmds:
      - echo "Installing E2E dependencies"
      - |
          if [ -n "${VIRTUAL_ENV:-}" ]; then
            uv sync --all-extras --active
          else
            uv sync --all-extras
          fi

  test-e2e:
    desc: "Install + run full deployment E2E"
    cmds:
      - task: install-e2e
      - task: e2e

  render-template:
    desc: "Render template"
    cmds:
      - echo "Rendering template {{.AGENT}}"
      # Clean up the existing rendered directory except for .venv to speed up rendering and avoid conflict checks
      - cd ./$RENDER_DIR/agent_{{.AGENT}}/ && [ -d . ] && find . -mindepth 1 -maxdepth 1 ! -name '.venv' -exec rm -rf {} + || true
      - mkdir -p ./$RENDER_DIR/agent_{{.AGENT}}/.datarobot/answers/
      - cp ./fixtures/.datarobot/answers/llm-llm.yml ./$RENDER_DIR/agent_{{.AGENT}}/.datarobot/answers/llm-llm.yml
      - cp ./fixtures/.datarobot/answers/drmcp-mcp_server.yml ./$RENDER_DIR/agent_{{.AGENT}}/.datarobot/answers/drmcp-mcp_server.yml
      - uvx copier copy . ./$RENDER_DIR/agent_{{.AGENT}} --defaults --data agent_template_framework={{.AGENT}} --vcs-ref=HEAD
      - cp ./fixtures/Taskfile.yml ./$RENDER_DIR/agent_{{.AGENT}}/Taskfile.yml
      - mkdir -p ./$RENDER_DIR/agent_{{.AGENT}}/infra/infra
      - cp ./fixtures/infra/Taskfile.yml ./$RENDER_DIR/agent_{{.AGENT}}/infra/Taskfile.yml
      - cp ./fixtures/infra/pyproject.toml ./$RENDER_DIR/agent_{{.AGENT}}/infra/pyproject.toml
      - cp ./fixtures/infra/__init__.py ./$RENDER_DIR/agent_{{.AGENT}}/infra/infra/__init__.py
      - cp ./fixtures/infra/llm.py ./$RENDER_DIR/agent_{{.AGENT}}/infra/infra/llm.py

  render-template-e2e:
    desc: "Render template + inject test scaffolding used by E2E (Taskfiles + minimal infra entrypoint/files)"
    cmds:
      - task: render-template
        vars:
          AGENT: "{{.AGENT}}"
      # Add base component scaffolding used by our tests (Taskfiles + minimal infra entrypoint/files).
      - cp ./fixtures/Taskfile.yml ./$RENDER_DIR/agent_{{.AGENT}}/Taskfile.yml
      - mkdir -p ./$RENDER_DIR/agent_{{.AGENT}}/infra/infra
      - cp ./fixtures/infra/Taskfile.yml ./$RENDER_DIR/agent_{{.AGENT}}/infra/Taskfile.yml
      # Inject minimal infra scaffolding so `pulumi up` can run in this component repo.
      # (In a real AF project this is provided by the base/llm/mcp components.)
      - cp ./fixtures/infra/Pulumi.yaml ./$RENDER_DIR/agent_{{.AGENT}}/infra/Pulumi.yaml
      - cp ./fixtures/infra/__main__.py ./$RENDER_DIR/agent_{{.AGENT}}/infra/__main__.py
      - cp ./fixtures/infra/pyproject.toml ./$RENDER_DIR/agent_{{.AGENT}}/infra/pyproject.toml
      - cp ./fixtures/infra/__init__.py ./$RENDER_DIR/agent_{{.AGENT}}/infra/infra/__init__.py
      - cp ./fixtures/infra/llm.py ./$RENDER_DIR/agent_{{.AGENT}}/infra/infra/llm.py
      - cp ./fixtures/infra/mcp_server.py ./$RENDER_DIR/agent_{{.AGENT}}/infra/infra/mcp_server.py

  install-rendered:
    desc: "Install dev"
    cmds:
      - echo "Installing dev for {{.AGENT}}"
      - task: render-template
        vars:
          AGENT: "{{.AGENT}}"
      - task agent:install --dir $RENDER_DIR/agent_{{.AGENT}}
      - task infra:install --dir $RENDER_DIR/agent_{{.AGENT}}

  docker_update_reqs:
    desc: "Update Docker requirements pyproject.toml from jinja template"
    vars:
      genai_agents_path: "{{.AGENT_PATH}}"
    cmds:
      - |
        if [ -z "{{.AGENT_PATH}}" ]; then
          echo "AGENT_PATH is not set. Please set AGENT_PATH to the path of the genai-agents repository."
          echo "task docker_update_reqs AGENT_PATH=/path/to/datarobot-user-models/public_dropin_environments/python311_genai_agents"
          exit 1
        fi
      - uv run jinja -D agent_app_name DockerContext -D agent_template_framework docker template/\{\{agent_app_name\}\}/pyproject.toml.jinja -o {{.AGENT_PATH}}/pyproject.toml
      - cd {{.AGENT_PATH}} && uv lock --no-upgrade && uv sync
      # This command is temporary. It is using pip compile to produce a truncated requirements.txt without dependencies to avoid overloading the HHI installer.
      - cd {{.AGENT_PATH}} && uv pip compile --python=3.11 --python-platform=x86_64-manylinux_2_28 --no-header --no-annotate --no-deps --extra agentic_playground --output-file=requirements.txt pyproject.toml
      # Need to extend support for larger requirements.txt files to enable this.
      # - cd {{.AGENT_PATH}} && uv pip freeze > {{.AGENT_PATH}}/requirements.txt
      # - cat {{.AGENT_PATH}}/requirements.txt

  docker_build_local:
    desc: "Build docker image locally (Check container builds)"
    cmds:
      - cd template/\{\{agent_app_name\}\}/docker_context && docker build -f Dockerfile . -t test --platform linux/amd64

  test-agent:
    desc: "Run tests"
    cmds:
      - task: install-rendered
        vars:
          AGENT: "{{.AGENT}}"
          SKIP_LINT: "{{.SKIP_LINT| default 0}}"
          SKIP_INFRA: "{{.SKIP_INFRA| default 0}}"
      - echo "Running tests for {{.AGENT}}"
      - |
          if [ "{{.SKIP_LINT}}" != "1" ]; then
            task agent:lint-check --dir $RENDER_DIR/agent_{{.AGENT}}
          else
            echo "Skipping lint check for agent: {{.AGENT}} because SKIP_LINT=1"
          fi
      - task agent:test-coverage --dir $RENDER_DIR/agent_{{.AGENT}}
      - |
          if [ "{{.SKIP_INFRA}}" != "1" ]; then
            task infra:lint-check --dir $RENDER_DIR/agent_{{.AGENT}}
            task infra:test-coverage --dir $RENDER_DIR/agent_{{.AGENT}}
          else
            echo "Skipping lint check for agent: {{.AGENT}} because SKIP_INFRA=1"
          fi

  test-base:
    desc: "Run tests for base agent"
    cmds:
      - task: test-agent
        vars:
          AGENT: base
          SKIP_LINT: "{{.SKIP_LINT| default 0}}"
          SKIP_INFRA: "{{.SKIP_INFRA| default 0}}"

  test-nat:
    desc: "Run tests for base agent"
    cmds:
      - task: test-agent
        vars:
          AGENT: nat
          SKIP_LINT: "{{.SKIP_LINT| default 0}}"
          SKIP_INFRA: "{{.SKIP_INFRA| default 0}}"

  test-crewai:
    desc: "Run tests for crewai agent"
    cmds:
      - task: test-agent
        vars:
          AGENT: crewai
          SKIP_LINT: "{{.SKIP_LINT| default 0}}"
          SKIP_INFRA: "{{.SKIP_INFRA| default 0}}"

  test-langgraph:
    desc: "Run tests for langgraph agent"
    cmds:
      - task: test-agent
        vars:
          AGENT: langgraph
          SKIP_LINT: "{{.SKIP_LINT| default 0}}"
          SKIP_INFRA: "{{.SKIP_INFRA| default 0}}"

  test-llamaindex:
    desc: "Run tests for llamaindex agent"
    cmds:
      - task: test-agent
        vars:
          AGENT: llamaindex
          SKIP_LINT: "{{.SKIP_LINT| default 0}}"
          SKIP_INFRA: "{{.SKIP_INFRA| default 0}}"

  validate-windows-compatibility:
    desc: "Validate that file paths don't contain characters illegal on Windows"
    cmds:
      - cmd: |
          echo "Scanning for Windows-incompatible file names (<>:\"|?*)..."
          total=$(find ./template -type f | wc -l | tr -d ' ')
          failures=$(find ./template -type f | while IFS= read -r f; do
            if echo "$f" | grep -q '[<>:"|?*]'; then
              echo "FAIL: $f" >&2
              echo "$f"
            else
              echo "  ok: $f" >&2
            fi
          done)
          bad=0
          if [ -n "$failures" ]; then
            bad=$(echo "$failures" | wc -l | tr -d ' ')
            echo ""
            echo "Offending files:"
            echo "$failures"
          fi
          echo ""
          echo "Checked $total files, $bad incompatible."
          [ "$bad" -gt 0 ] && exit 1 || true
        platforms: [linux, darwin]

  test:
    desc: "Run tests for all agents"
    cmds:
      - task: validate-windows-compatibility
      - task: test-base
      - task: test-crewai
      - task: test-langgraph
      - task: test-llamaindex
      - task: test-cli
      - task: test-cli-json
      - task: test-cli-streaming

  run-cli:
    desc: "Run CLI"
    cmds:
      - task: install-rendered
        vars:
          AGENT: "base"
      - echo "Running tests for {{.AGENT}}"
      - cd $RENDER_DIR/agent_base && echo "DATAROBOT_API_TOKEN=secret" > .env && echo "DATAROBOT_ENDPOINT=https://test.com/api/v2" >> .env
      - task agent:cli --dir $RENDER_DIR/agent_base -- execute --show_output --user_prompt "Artificial Intelligence" > $RENDER_DIR/agent_base/agent/output.log 2>&1

  test-cli:
    desc: "Test CLI"
    cmds:
      - task: run-cli
      - task: check-output
        vars:
          ASSERTION: "success"

  check-output:
      - if [ $(wc -l < $RENDER_DIR/agent_base/agent/output.log) -lt 13 ]; then exit 1; fi
      - if ! cat $RENDER_DIR/agent_base/agent/output.log | grep -q "Running CLI execute"; then exit 1; fi
      - if ! cat $RENDER_DIR/agent_base/agent/output.log | grep -q "INFO - Parsing args"; then exit 1; fi
      - if ! cat $RENDER_DIR/agent_base/agent/output.log | grep -q "INFO - Starting DRUM runner"; then exit 1; fi
      - if ! cat $RENDER_DIR/agent_base/agent/output.log | grep -q "Running agent with user prompt"; then exit 1; fi
      - 'if ! cat $RENDER_DIR/agent_base/agent/output.log | grep -q "\"content\": \"{{.ASSERTION}}\""; then exit 1; fi'

  run-cli-json:
    desc: "Run CLI with JSON"
    cmds:
      - task: install-rendered
        vars:
          AGENT: "base"
      - echo "Running tests for {{.AGENT}}"
      - cd $RENDER_DIR/agent_base && echo "DATAROBOT_API_TOKEN=secret" > .env && echo "DATAROBOT_ENDPOINT=https://test.com/api/v2" >> .env
      - task agent:cli --dir $RENDER_DIR/agent_base -- execute --show_output --completion_json example-completion.json > $RENDER_DIR/agent_base/agent/output.log 2>&1

  test-cli-json:
    desc: "Test CLI with JSON"
    cmds:
      - task: run-cli-json
      - task: check-output
        vars:
          ASSERTION: "success"

  run-cli-streaming:
    desc: "Run CLI with Streaming"
    cmds:
      - task: install-rendered
        vars:
          AGENT: "base"
      - echo "Running tests for {{.AGENT}}"
      - cd $RENDER_DIR/agent_base && echo "DATAROBOT_API_TOKEN=secret" > .env && echo "DATAROBOT_ENDPOINT=https://test.com/api/v2" >> .env
      - task agent:cli --dir $RENDER_DIR/agent_base -- execute --show_output --user_prompt "Artificial Intelligence" --stream > $RENDER_DIR/agent_base/agent/output.log 2>&1

  test-cli-streaming:
    desc: "Test CLI with Streaming"
    cmds:
      - task: run-cli-streaming
      - task: check-output
        vars:
          ASSERTION: "streaming success"

  build-docker:
    desc: "Build docker image"
    cmds:
      - task: install-rendered
        vars:
          AGENT: "{{.AGENT}}"
      - echo "Building docker image for {{.AGENT}}"
      - task agent:build-docker --dir $RENDER_DIR/agent_{{.AGENT}}

  build-docker-base:
    desc: "Build docker for base agent"
    cmds:
      - task: build-docker
        vars:
          AGENT: base

  build-docker-crewai:
    desc: "Build docker for crewai agent"
    cmds:
      - task: build-docker
        vars:
          AGENT: crewai

  build-docker-langgraph:
    desc: "Build docker for langgraph agent"
    cmds:
      - task: build-docker
        vars:
          AGENT: langgraph

  build-docker-llamaindex:
    desc: "Build docker for llamaindex agent"
    cmds:
      - task: build-docker
        vars:
          AGENT: llamaindex

  build-docker-nat:
    desc: "Build docker image for NAT"
    cmds:
      - task: build-docker
        vars:
          AGENT: nat

  e2e:
    desc: "Run full deployment E2E for all agents (requires DATAROBOT_ENDPOINT + DATAROBOT_API_TOKEN)"
    cmds:
      - echo "Running full deployment E2E (all agent frameworks)"
      - |
          # Load local repo `.env` (gitignored) if present.
          # Keep it scoped to E2E so credentials don't leak into unrelated tasks.
          set -a
          if [ -f ".env" ]; then
            . ".env"
          fi
          set +a

          if [ -n "${VIRTUAL_ENV:-}" ]; then
            RUN_E2E=1 uv run --all-extras --active pytest --import-mode=importlib -vv -s ./tests/e2e/test_agent_e2e.py
          else
            RUN_E2E=1 uv run --all-extras pytest --import-mode=importlib -vv -s ./tests/e2e/test_agent_e2e.py
          fi