aditya
50680612d5
Implemented optional estimation_actor role for cost, risk, and duration estimation during plan lifecycle. Estimates are persisted to plan metadata and surfaced in CLI output. Key implementation details: - EstimationOutput (Pydantic models): CostEstimate with currency, token estimates, and confidence ranges; RiskScore with 0-100 scale and factors; DurationEstimate with min/expected/max seconds. All include confidence levels and validation. EstimationSkipped records when estimation is opted out. - EstimationService: stateless async service that invokes estimation actor (stub implementation for M6). Handles actor output parsing, error recovery, and fallback to EstimationSkipped on failure. - Integration: Plan model gains estimation_output and estimation_skipped fields. LifecyclePlanModel adds JSON columns for persistence. PlanLifecycleService invokes estimation during use_action unless skip_estimation is true. - CLI: --no-estimate flag added to 'agents plan use'. plan status displays cost (USD with token estimates), risk (score/100 with confidence), and duration (seconds with ranges) in rich format. - Database: Alembic migration m6_003_estimation_metadata adds estimation_output_json and estimation_skipped_json columns to v3_plans. - Tests: 44 BDD scenarios (features/estimation.feature) covering validation, lifecycle integration, persistence, CLI display, and edge cases. 18 Robot Framework smoke tests. 19 ASV benchmark suites for schema, serialization, validation, and plan integration performance. - Documentation: docs/reference/estimation.md with schema, configuration, and examples. plan_cli.md updated with --no-estimate flag usage. ISSUES CLOSED: #209
121 lines
5.4 KiB
Gherkin
121 lines
5.4 KiB
Gherkin
Feature: Async Resource Cleanup and Leak Prevention
|
|
As a system operator
|
|
I want async resources to be tracked and closed reliably
|
|
So that the system does not leak connections, tasks, or subscriptions
|
|
|
|
Background:
|
|
Given I have an async resource tracker
|
|
|
|
Scenario: Registering an async resource and closing it
|
|
Given I have a mock async resource named "db-pool"
|
|
When I register the resource with the tracker
|
|
And I close all tracked resources
|
|
Then the resource "db-pool" should be closed
|
|
And the tracker should have zero open resources
|
|
|
|
Scenario: close_all awaits registered resources with timeout
|
|
Given I have 3 mock async resources with varied close times
|
|
When I close all tracked resources with a 5 second timeout
|
|
Then all 3 resources should be closed
|
|
And the tracker should have zero open resources
|
|
|
|
Scenario: Leaked resources are logged by name in finalizer
|
|
Given I have a mock async resource named "leaky-conn"
|
|
And I register the resource with the tracker
|
|
When the tracker finalizer runs without close_all
|
|
Then a warning should be logged mentioning "leaky-conn"
|
|
|
|
Scenario: Graceful cancellation awaits in-flight tasks before cleanup
|
|
Given I have an async task tracked by the bridge
|
|
When I request graceful cleanup with a 2 second timeout
|
|
Then the task should be cancelled
|
|
And the bridge should have no active tasks
|
|
|
|
Scenario: Time-bounded shutdown warns on forced termination
|
|
Given I have an async resource tracker
|
|
Given I have a mock async resource named "slow-resource" that takes 5 seconds to close
|
|
And I register the resource with the tracker
|
|
When I close all tracked resources with a 0.1 second timeout
|
|
Then a warning should be logged about forced termination
|
|
And the tracker should report the timed-out resource
|
|
|
|
Scenario: Cancelled async jobs persist cancellation reason
|
|
Given I have an async task tracked by the bridge
|
|
When I cancel the task with reason "user-requested-shutdown"
|
|
Then the cancellation reason should be "user-requested-shutdown"
|
|
|
|
Scenario: Checkpoint file handles are closed on cleanup
|
|
Given I have a state manager with a temporary checkpoint directory
|
|
When I save a checkpoint and then close the state manager
|
|
Then the checkpoint file should exist and be readable
|
|
And the state manager should be marked as closed
|
|
|
|
Scenario: Subscriptions are disposed on cleanup
|
|
Given I have an A2A event queue with 3 active subscriptions
|
|
When I close the event queue
|
|
Then all subscriptions should be removed
|
|
And the subscription count should be zero
|
|
|
|
Scenario: Tracker rejects duplicate resource names
|
|
Given I have a mock async resource named "unique-res"
|
|
And I register the resource with the tracker
|
|
When I try to register another resource named "unique-res"
|
|
Then a ValueError should be raised mentioning "unique-res"
|
|
|
|
Scenario: Tracker rejects empty resource name
|
|
When I try to register a resource with an empty name
|
|
Then a ValueError should be raised mentioning "name"
|
|
|
|
Scenario: Tracker rejects None resource
|
|
When I try to register a None resource with name "valid-name"
|
|
Then a ValueError should be raised mentioning "resource"
|
|
|
|
Scenario: Tracker context manager closes resources on exit
|
|
Given I have a mock async resource named "ctx-resource"
|
|
When I use the tracker as an async context manager and register the resource
|
|
Then the resource "ctx-resource" should be closed after exiting the context
|
|
|
|
Scenario: close_all is idempotent
|
|
Given I have a mock async resource named "once-resource"
|
|
And I register the resource with the tracker
|
|
When I close all tracked resources twice
|
|
Then the resource "once-resource" should be closed exactly once
|
|
And the tracker should have zero open resources
|
|
|
|
Scenario: Enhanced bridge cleanup awaits tasks with timeout
|
|
Given I have a bridge with 2 slow async tasks
|
|
When I run enhanced cleanup with a 2 second timeout
|
|
Then the bridge should have no active tasks
|
|
And completed tasks should be logged
|
|
|
|
Scenario: Registering a resource after close_all raises RuntimeError
|
|
Given I have a mock async resource named "late-resource"
|
|
And I register the resource with the tracker
|
|
When I close all tracked resources
|
|
And I try to register a resource named "post-close" after close_all
|
|
Then a RuntimeError should be raised mentioning "closed"
|
|
|
|
Scenario: State update after close raises RuntimeError
|
|
Given I have a state manager with a temporary checkpoint directory
|
|
When I close the state manager
|
|
And I try to update state after close
|
|
Then a RuntimeError should be raised mentioning "closed"
|
|
|
|
Scenario: State reset after close raises RuntimeError
|
|
Given I have a state manager with a temporary checkpoint directory
|
|
When I close the state manager
|
|
And I try to reset state after close
|
|
Then a RuntimeError should be raised mentioning "closed"
|
|
|
|
Scenario: Publish after close raises RuntimeError on A2aEventQueue
|
|
Given I have an A2A event queue with 1 active subscriptions
|
|
When I close the event queue
|
|
And I try to publish an event after close
|
|
Then a RuntimeError should be raised mentioning "closed"
|
|
|
|
Scenario: A2aEventQueue exposes is_closed property
|
|
Given I have an A2A event queue with 0 active subscriptions
|
|
Then the event queue is_closed should be False
|
|
When I close the event queue
|
|
Then the event queue is_closed should be True
|