cleveragents/cleveragents-core
4
0
Fork 3
Code Issues 6.0k Pull Requests 115 Actions Packages Projects Releases Wiki Activity

test(e2e): workflow example 10 — full-auto batch formatting and linting (full-auto profile) #786

Merged
HAL9000 merged 2 commits from test/e2e-wf10-batch into master 2026-04-22 02:54:22 +00:00
Conversation 20 Commits 2 Files Changed 2
+405
2 changed files with 405 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
CHANGELOG.md
+13
View File
@@ -7,6 +7,19 @@ The format follows [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
### Added
- Added E2E test for Workflow Example 10: Full-Auto Batch Operations.
Robot Framework test (`robot/e2e/wf10_batch.robot`) exercises batch
formatting across a monorepo with 3 badly-formatted Python packages
using the `full-auto` automation profile. Creates badly-formatted
Python packages, plus a deliberately broken action (non-existent LLM
actor) for error handling. Registers resources and projects with
dynamic branch detection, launches plans in full-auto mode with dynamic
actor selection (OpenAI/Anthropic), and verifies batch results via
`plan list --state applied` and `--state errored` filters.
Demonstrates batch error handling when one plan fails. Zero mocking —
real CLI, real LLM API keys. Tagged `E2E`, skips gracefully when API
keys are absent. (#756)
- **Plan diff shows worktree branch changes** (#9231): `plan diff` now detects
the worktree branch `cleveragents/plan-<id>` created during `plan execute`
and runs `git diff HEAD...<branch>` to display actual file changes. Falls
robot/e2e/wf10_batch.robot
+392
View File
@@ -0,0 +1,392 @@
*** Settings ***
Documentation E2E test for Workflow Example 10: Full-Auto Batch Operations.
...
... A team reformats packages in a monorepo using the ``full-auto``
... automation profile. Multiple plans run without human
... intervention (strategize → execute → apply automatically).
... Includes a deliberately broken action (non-existent LLM actor)
... to demonstrate batch error handling.
...
... Requires real LLM API keys — zero mocking.
Library Collections
Resource common_e2e.resource
Suite Setup WF10 Suite Setup
Suite Teardown E2E Suite Teardown
*** Variables ***
@{PACKAGE_NAMES} pkg_auth pkg_common pkg_billing
${ACTION_NAME} local/format-codebase
${PLAN_TIMEOUT} 180s
*** Keywords ***
WF10 Suite Setup
[Documentation] E2E Suite Setup plus database initialisation.
E2E Suite Setup
# Initialise the database so CLI commands work in all tests.
${init}= Run CleverAgents Command init --force --yes
Should Be Equal As Integers ${init.rc} 0
Create Package Directory
[Documentation] Create a single package with badly-formatted Python files.
[Arguments] ${monorepo} ${pkg_name}
${pkg_dir}= Set Variable ${monorepo}${/}${pkg_name}
Create Directory ${pkg_dir}${/}src
# __init__.py with extra blank lines and bad spacing
${init_content}= Set Variable
... \n\n\n"""${pkg_name} package."""\n\n\nimport os\nimport sys\nimport json\n\n\n__all__=["main"]\n
Create File ${pkg_dir}${/}src${/}__init__.py ${init_content}
# main.py with intentionally bad formatting: unsorted imports, extra spaces, long lines
${main_content}= Set Variable
... import json\nimport os\nimport sys\nfrom pathlib import Path\nimport re\n\n\ndef main( ):\n """Entry point with bad formatting."""\n x=1\n y = 2\n z=x+y\n data = {"key": "value", "another": "item"}\n return z\n\nif __name__=="__main__":\n main( )\n
Create File ${pkg_dir}${/}src${/}main.py ${main_content}
RETURN ${pkg_dir}
Create Temp Monorepo
[Documentation] Create a temporary monorepo with multiple badly-formatted packages.
...
... Each healthy package contains ``src/__init__.py`` and ``src/main.py``
... with intentionally poor formatting (extra spaces, unsorted imports).
... Returns the path to the monorepo root and the detected branch name.
${monorepo}= Create Temp Git Repo wf10-monorepo
FOR ${pkg_name} IN @{PACKAGE_NAMES}
Create Package Directory ${monorepo} ${pkg_name}
END
# Commit all packages so git-checkout resources have content
${add_res}= Run Process git add . cwd=${monorepo} timeout=60s on_timeout=kill
Should Be Equal As Integers ${add_res.rc} 0 git add failed: ${add_res.stderr}
${commit_res}= Run Process git commit -m Add packages with bad formatting cwd=${monorepo} timeout=60s on_timeout=kill
Should Be Equal As Integers ${commit_res.rc} 0 git commit failed: ${commit_res.stderr}
# Detect the actual default branch name (may be main or master)
${branch_result}= Run Process git rev-parse --abbrev-ref HEAD cwd=${monorepo} timeout=60s on_timeout=kill
Should Be Equal As Integers ${branch_result.rc} 0 git rev-parse failed: ${branch_result.stderr}
${branch}= Strip String ${branch_result.stdout}
Log Detected monorepo branch: ${branch}
RETURN ${monorepo} ${branch}
Write Action Config
[Documentation] Write a formatting action YAML config with full-auto profile.
...
... Uses dynamic actor selection based on available API keys.
... Returns the path to the YAML file.
# Pick an actor that matches the available API key
${has_anthropic}= Evaluate bool(__import__('os').environ.get('ANTHROPIC_API_KEY', ''))
IF ${has_anthropic}
${actor}= Set Variable anthropic/claude-sonnet-4-20250514
ELSE
${actor}= Set Variable openai/gpt-4o
END
${yaml_path}= Set Variable ${SUITE_HOME}${/}format_action.yaml
${config}= Catenate SEPARATOR=\n
... name: ${ACTION_NAME}
... description: "Reformat Python source files for consistent style"
... strategy_actor: ${actor}
... execution_actor: ${actor}
... definition_of_done: "All Python files are consistently formatted"
... automation_profile: full-auto
... reusable: true
... state: available
... read_only: false
... invariants:
... ${SPACE}${SPACE}- "Changes must be whitespace-only (no semantic modifications)"
... ${SPACE}${SPACE}- "Every package must pass its own test suite after formatting"
Create File ${yaml_path} ${config}\n
RETURN ${yaml_path}
Write Broken Action Config
[Documentation] Write a deliberately broken action YAML that uses a non-existent
... LLM actor. Plans created with this action will fail during
... execution when the strategy/execution actor cannot be resolved.
${yaml_path}= Set Variable ${SUITE_HOME}${/}broken_action.yaml
${config}= Catenate SEPARATOR=\n
... name: local/broken-format
... description: "Deliberately broken action for error handling testing"
... strategy_actor: nonexistent/model-xyz-404
... execution_actor: nonexistent/model-xyz-404
... definition_of_done: "This action should always fail"
... automation_profile: full-auto
... reusable: true
... state: available
... read_only: false
... invariants:
... ${SPACE}${SPACE}- "No changes expected — action is broken"
Create File ${yaml_path} ${config}\n
RETURN ${yaml_path}
Register Package Resources And Projects
[Documentation] Register git-checkout resources and create projects for the
... specified packages.
...
... Error handling is tested separately via a broken action
... (non-existent LLM actor), not via missing resources.
...
... Resource/project names use fixed ``local/<pkg>`` prefixes without
... UUID suffixes — safe because ``init --force --yes`` in suite setup
... resets the workspace database before each run.
[Arguments] ${monorepo} ${branch} @{healthy_packages}
FOR ${pkg_name} IN @{healthy_packages}
# Register git-checkout resource pointing to the monorepo root
${res_result}= Run CleverAgents Command
... resource add git-checkout local/${pkg_name}
... --path ${monorepo} --branch ${branch}
Log Resource ${pkg_name}: ${res_result.stdout}
Should Not Contain ${res_result.stderr} Traceback
... Resource registration for ${pkg_name} produced Traceback:\n${res_result.stderr}
# Create project linked to resource — must succeed WITHOUT Traceback.
${proj_result}= Run CleverAgents Command
... project create local/${pkg_name}
... --resource local/${pkg_name}
Log Project ${pkg_name}: ${proj_result.stdout}
Should Not Contain ${proj_result.stderr} Traceback
... Project creation for ${pkg_name} produced Traceback:\n${proj_result.stderr}
END
Launch Batch Plans
[Documentation] Launch a plan for each package in full-auto mode.
...
... Returns a list of plan identifiers extracted from output.
... Uses ``expected_rc=None`` because broken packages may fail
... during plan use (which is expected behaviour).
[Arguments] @{all_packages}
@{plan_ids}= Create List
FOR ${pkg_name} IN @{all_packages}
${result}= Run CleverAgents Command
... plan use ${ACTION_NAME} local/${pkg_name}
... --automation-profile full-auto --format plain
... timeout=${PLAN_TIMEOUT} expected_rc=None
Log Plan use ${pkg_name} stdout: ${result.stdout}
Log Plan use ${pkg_name} stderr: ${result.stderr}
Should Not Contain ${result.stderr} Traceback
... plan use for ${pkg_name} produced Traceback:\n${result.stderr}
# Extract plan ID from output — must find a valid ULID (Crockford Base32)
${combined}= Set Variable ${result.stdout} ${result.stderr}
${match}= Get Regexp Matches ${combined} ([0-9A-HJKMNP-TV-Z]{26}) flags=IGNORECASE
${match_count}= Get Length ${match}
IF ${match_count} > 0
${plan_id}= Set Variable ${match}[0]
Append To List ${plan_ids} ${plan_id}
Log Captured plan ID for ${pkg_name}: ${plan_id}
ELSE
Log No plan ID extracted for ${pkg_name} (rc=${result.rc}) — plan creation may have failed WARN
END
END
RETURN @{plan_ids}
Execute Batch Plans
[Documentation] Execute each plan through the strategize→execute pipeline.
...
... ``plan execute`` runs the current plan phase synchronously.
... With the full-auto automation profile, a single ``plan execute``
... call auto-advances through both strategize and execute phases,
... leaving the plan ready for ``plan apply --yes``.
...
... Plans that fail during execution (e.g. broken action with
... non-existent actor) are logged but do not abort the batch —
... the batch continues with remaining plans.
...
... Returns a list of plan IDs that completed execution
... successfully (candidates for apply).
[Arguments] @{plan_ids}
@{executed_ids}= Create List
FOR ${plan_id} IN @{plan_ids}
Log Executing plan ${plan_id} (strategize + execute via full-auto)
${exec}= Run CleverAgents Command
... plan execute ${plan_id} --format plain
... timeout=${PLAN_TIMEOUT} expected_rc=None
Log Execute ${plan_id} rc=${exec.rc}: ${exec.stdout}
IF ${exec.rc} != 0
Log Plan ${plan_id} failed during execution (rc=${exec.rc}): ${exec.stderr} WARN
CONTINUE
END
Append To List ${executed_ids} ${plan_id}
END
RETURN @{executed_ids}
Apply Batch Plans
[Documentation] Apply each successfully-executed plan via ``plan apply --yes``.
...
... After ``plan execute`` with full-auto profile, plans are in
... the ``execute/complete`` state. ``plan apply --yes`` performs
... the actual application (e.g. committing changes to the repo)
... and completes the apply phase, transitioning the plan to the
... ``applied`` processing state.
...
... Note: ``plan lifecycle-apply`` only transitions the plan INTO
... the apply phase (``apply/queued``) without completing it.
... ``plan apply --yes`` is required to actually run and complete
... the apply step.
...
... Plans that fail during apply are logged but do not abort
... the batch. Returns a list of plan IDs that were applied.
[Arguments] @{executed_ids}
@{applied_ids}= Create List
FOR ${plan_id} IN @{executed_ids}
Log Applying plan ${plan_id}
${apply}= Run CleverAgents Command
... plan apply --yes ${plan_id} --format plain
... timeout=${PLAN_TIMEOUT} expected_rc=None
Log Apply ${plan_id} rc=${apply.rc}: ${apply.stdout}
IF ${apply.rc} != 0
Log Plan ${plan_id} failed during apply (rc=${apply.rc}): ${apply.stderr} WARN
CONTINUE
END
Append To List ${applied_ids} ${plan_id}
END
RETURN @{applied_ids}
*** Test Cases ***
Workflow 10 Full-Auto Batch Formatting
[Documentation] End-to-end test for full-auto batch formatting across
... multiple packages in a monorepo, including error handling
... when one plan fails due to a broken action.
...
... 1. Creates a monorepo with 3 badly-formatted Python packages
... 2. Registers a reusable formatting action with full-auto profile
... 3. Registers a broken action (non-existent LLM actor) for error testing
... 4. Registers resources and projects for healthy packages
... 5. Creates plans in full-auto mode via ``plan use``
... 6. Executes all plans via ``plan execute`` (strategize + execute)
... 7. Applies successful plans via ``plan apply --yes``
... 8. Verifies batch results via ``plan list`` with state filters
... 9. Verifies error handling — broken action's plan fails during execution
[Tags] E2E
[Timeout] 25 minutes
[Teardown] Run CleverAgents Command config set core.automation-profile manual expected_rc=None
Skip If No LLM Keys
# --- Step 1: Create temp monorepo with badly-formatted packages ---
${monorepo} ${branch}= Create Temp Monorepo
Log Monorepo created at: ${monorepo} (branch: ${branch})
Directory Should Exist ${monorepo}${/}pkg_auth${/}src
Directory Should Exist ${monorepo}${/}pkg_common${/}src
Directory Should Exist ${monorepo}${/}pkg_billing${/}src
# --- Step 2: Create the formatting action with full-auto profile ---
${yaml_path}= Write Action Config
File Should Exist ${yaml_path}
${action_result}= Run CleverAgents Command
... action create --config ${yaml_path}
Log Action create output: ${action_result.stdout}
Should Not Contain ${action_result.stderr} Traceback
Output Should Contain ${action_result} format-codebase
# Also create a broken action with a non-existent actor for error handling
${broken_action}= Write Broken Action Config
${broken_action_result}= Run CleverAgents Command
... action create --config ${broken_action}
Log Broken action create output: ${broken_action_result.stdout}
Output Should Contain ${broken_action_result} broken-format
# --- Step 3: Register resources and projects for all packages ---
Register Package Resources And Projects ${monorepo} ${branch} @{PACKAGE_NAMES}
# --- Step 4: Create plans — healthy + broken ---
# 4a: Launch plans for healthy packages with the good action
@{plan_ids}= Launch Batch Plans @{PACKAGE_NAMES}
${plan_count}= Get Length ${plan_ids}
Log Healthy plan IDs (${plan_count}): @{plan_ids}
Should Be True ${plan_count} == 3
... Expected 3 healthy plan IDs but got ${plan_count}
# 4b: Launch plan for first healthy package but with BROKEN action
# (non-existent actor will cause execution to fail)
${broken_result}= Run CleverAgents Command
... plan use local/broken-format local/pkg_auth
... --automation-profile full-auto --format plain
... timeout=${PLAN_TIMEOUT} expected_rc=None
Log Broken action plan use rc=${broken_result.rc}: ${broken_result.stdout}
Log Broken action plan use stderr: ${broken_result.stderr}
Should Not Contain ${broken_result.stderr} Traceback
${broken_plan_failed_at_creation}= Evaluate ${broken_result.rc} != 0
${broken_plan_id}= Set Variable ${EMPTY}
IF not ${broken_plan_failed_at_creation}
# Extract plan ID for the broken plan
${combined}= Set Variable ${broken_result.stdout} ${broken_result.stderr}
${match}= Get Regexp Matches ${combined} ([0-9A-HJKMNP-TV-Z]{26}) flags=IGNORECASE
${match_count}= Get Length ${match}
IF ${match_count} > 0
${broken_plan_id}= Set Variable ${match}[0]
Append To List ${plan_ids} ${broken_plan_id}
Log Broken plan ID captured: ${broken_plan_id}
END
ELSE
Log Broken plan failed at creation (rc=${broken_result.rc})
END
# --- Step 5: Execute all plans (strategize + execute phases) ---
@{executed_ids}= Execute Batch Plans @{plan_ids}
${executed_count}= Get Length ${executed_ids}
${total_plan_count_at_execute}= Get Length ${plan_ids}
Log Plans that completed execution: ${executed_count} / ${total_plan_count_at_execute}
# --- Step 6: Apply successfully-executed plans ---
@{applied_ids}= Apply Batch Plans @{executed_ids}
${applied_count}= Get Length ${applied_ids}
Log Plans that reached applied state: ${applied_count} / ${executed_count}
# --- Step 7: Verify batch results via plan list ---
# 7a: Unfiltered listing — smoke test and verify healthy plan IDs appear
${list_result}= Run CleverAgents Command
... plan list --format plain
... timeout=30s
Log Final plan list stdout: ${list_result.stdout}
Log Final plan list stderr: ${list_result.stderr}
Should Not Contain ${list_result.stderr} Traceback
FOR ${pid} IN @{applied_ids}
Should Contain ${list_result.stdout} ${pid}
... Applied plan ID ${pid} not found in plan list output
END
# 7b: Filtered listing — verify --state applied returns successful plans
${applied_list}= Run CleverAgents Command
... plan list --state applied --format plain
... timeout=30s
Log Applied plan list stdout: ${applied_list.stdout}
${applied_matches}= Get Regexp Matches ${applied_list.stdout}
... processing_state:\\s*applied
${success_count}= Get Length ${applied_matches}
Log Plans in 'applied' state: ${success_count}
# At least 2 of 3 healthy packages must reach 'applied' state.
# Using >= 2 (not == 3) because LLM-generated changes can occasionally fail
# during apply (e.g. merge conflicts, empty changesets from the LLM producing
# no edits). Requiring >= 2 verifies the batch mechanism works while
# tolerating one transient apply failure.
Should Be True ${success_count} >= 2
... Expected at least 2 plans (of 3 healthy) to reach 'applied' state but found ${success_count}
# --- Step 8: Verify error handling for the broken action ---
# Count how many plans failed during execution (didn't make it to executed_ids)
${total_plan_count}= Get Length ${plan_ids}
${failed_execution_count}= Evaluate ${total_plan_count} - ${executed_count}
Log Plans that failed during execution: ${failed_execution_count}
# Also check for errored plans via --state errored filter
${errored_list}= Run CleverAgents Command
... plan list --state errored --format plain
... timeout=30s
Log Errored plan list stdout: ${errored_list.stdout}
${errored_matches}= Get Regexp Matches ${errored_list.stdout}
... processing_state:\\s*errored
${error_count}= Get Length ${errored_matches}
Log Plans in 'errored' state: ${error_count}
# The broken action (non-existent actor) should cause at least one failure.
# Error handling is demonstrated if ANY of the following hold:
# (a) broken action's plan use failed (rc != 0)
# (b) the broken plan's ID is NOT in executed_ids (failed during execution)
# (c) the broken plan's ID appears in the errored list
IF ${broken_plan_failed_at_creation}
Log Error handling demonstrated: broken plan failed at creation (rc != 0)
ELSE IF "${broken_plan_id}" != "${EMPTY}"
# Verify the specific broken plan ID either did NOT complete execution
# or appears in the errored list
${broken_in_executed}= Evaluate """${broken_plan_id}""" in ${executed_ids}
${broken_in_errored}= Evaluate """${broken_plan_id}""" in """${errored_list.stdout}"""
${broken_failed}= Evaluate not ${broken_in_executed} or ${broken_in_errored}
Should Be True ${broken_failed}
... Broken plan ${broken_plan_id} should have failed but was found in executed_ids and not in errored list
Log Error handling demonstrated: broken plan ${broken_plan_id} failed (not in executed=${broken_in_executed}, in errored=${broken_in_errored})
ELSE
# Fallback: general failure count check
${broken_demonstrated}= Evaluate
... ${error_count} >= 1 or ${failed_execution_count} >= 1
Should Be True ${broken_demonstrated}
... Error handling not demonstrated: 0 errored, 0 failed execution
END