feat(cloud): add targeted task governance foundation
Tests / Test passed: 659

This commit is contained in:
2026-07-13 22:21:12 +08:00
parent a3ba94be04
commit 2cd314b183
41 changed files with 2099 additions and 15 deletions
@@ -0,0 +1,113 @@
## 1. Preconditions and dependency reconciliation
- [x] 1.1 Confirm `cloud-planner-proxy` is fully implemented, passes its
strict validation and relevant runtime tests, and its deployed schema/API
contract is the base for this change.
- [ ] 1.2 Complete and reconcile `cloud-console-user-authentication`, including
the documented Console Users experience and its remaining PostgreSQL and
HTTPS/Host-Agent verification tasks.
- [ ] 1.3 Rebase this change's delta specs against the then-current canonical
`platform-sdk`, `cloud-console-ui`, `cloud-planner-proxy`, and
`agent-runtime` specifications before implementation begins.
## 2. Governance domain, persistence, and migration
- [ ] 2.1 Define target-selector, user-submission-policy, Host-governance-
policy, token-usage-event, and token-reservation domain models with
non-secret representations and revision semantics.
- [ ] 2.2 Extend the Cloud repository port with transactional CRUD/query
operations for policies, active Host capacity, budget reservations,
settlement, expiry cleanup, and bounded usage summaries/events.
- [ ] 2.3 Add SQLAlchemy rows, indexes, conversion helpers, and concurrency-
safe PostgreSQL/SQLite implementations for governance policies, usage
events, reservations, and safe policy audits.
- [ ] 2.4 Add an Alembic forward/downgrade revision that preserves existing
users, Hosts, devices, tasks, and attempts; advance schema readiness
checks to the new head.
- [ ] 2.5 Add repository and migration tests for policy revisions, null versus
empty allow-list semantics, active-task capacity, reservation races,
settlement, expiry cleanup, audit redaction, and SQLite/PostgreSQL
parity.
## 3. Targeted scheduling and public governance API
- [x] 3.1 Extend task constraints, SDK request/response models, persistence
serialization, status/list representations, and `CloudClient` for an
optional Host/Device target; reject a Device without its Host.
- [x] 3.2 Enforce target ownership/existence at public submission and filter
scheduler candidates by target before strategy selection; keep an
unavailable targeted task queued with no fallback assignment.
- [x] 3.3 Add `governance:read` and `governance:admin` scopes and policy-aware
public task authorization that combines `tasks:submit` with the
authenticated human user's effective submission policy.
- [ ] 3.4 Add bounded, non-secret public governance routes and `CloudClient`
methods for user policy, Host policy, Host AI-budget summaries, and
paginated usage events; audit every policy mutation.
- [ ] 3.5 Add public API/SDK tests for targeted submission, target-policy
denial, scope enforcement, status exposure, governance CRUD, and
absence of secret fields in responses/audits.
## 4. Host policy synchronization and self-submission
- [x] 4.1 Extend shared internal heartbeat request/response models with the
Host's last policy revision and a revision-aware effective-policy reply.
- [ ] 4.2 Update Host heartbeat synchronization to persist only safe cached
policy state and expose it through local status/history without adding an
inbound Cloud connection.
- [x] 4.3 Add the Host-scoped goal-only task-submission route and
`HostAgentClient` method; derive Host targeting from authenticated
credentials and validate any named local Device ownership.
- [ ] 4.4 Enforce the Host policy's self-submission and active-task limits in
the Cloud service/scheduler, not only in Host-local code.
- [ ] 4.5 Add Cloud API and Host Agent tests for revision convergence,
unchanged-policy replies, self-targeted task creation, foreign target
rejection, disabled self-submission, and outbound-only compatibility.
## 5. Cloud-proxy usage metering and hard budget enforcement
- [x] 5.1 Extend the dual-provider tool-calling result with optional
non-secret provider usage fields while preserving `AIPlanner`'s existing
single-decision behavior and direct transport compatibility.
- [ ] 5.2 Extend planner-proxy request context and Host-Agent-local context
binding so Cloud-proxied calls carry known task/attempt metadata without
importing Host or Cloud concerns into `runtime`.
- [ ] 5.3 Add Cloud proxy preflight reservation, configured conservative
per-call ceiling, provider invocation, actual-usage settlement, and
bounded unknown-usage reservation expiry.
- [ ] 5.4 Record non-secret usage events and expose accurate
used/reserved/remaining UTC-day budget summaries; explicitly report
direct transport as unmetered.
- [ ] 5.5 Add provider-fake, repository concurrency, Cloud API, and Host
Agent tests for usage parsing, over-budget rejection before provider
invocation, concurrent reservations, settlement, timeout uncertainty,
task-attempt attribution, and direct-transport labeling.
## 6. Cloud Console governance experience
- [x] 6.1 Add task composer API bindings and a scope-aware Console form for
goal/workflow submission, Host/Device target selection, and target/policy
validation errors.
- [ ] 6.2 Complete or reconcile the admin Users view, then add user-
submission-policy editing with safe refresh and conflict/error handling.
- [ ] 6.3 Add Host policy administration and AI-usage/budget views, including
revision display, unmetered direct Hosts, and no rendering of prompts,
screenshots, provider credentials, cookies, or lease secrets.
- [ ] 6.4 Add frontend tests for task composer scope/policy failures,
governance navigation, policy mutations, usage rendering, unmetered
status, CSRF writes, and retained session behavior on `403`.
## 7. Documentation, verification, and rollout
- [ ] 7.1 Update deployment and Console documentation with the dependency
order, migration/rollback sequence, policy semantics, Cloud transport
prerequisite, budget reservation behavior, direct-host limitation, and
safe administrator operations.
- [ ] 7.2 Run formatting, Ruff, compile checks, secret/redaction review, the
complete non-integration workspace test suite, Cloud Console unit tests,
type-check, and production build.
- [ ] 7.3 Run real PostgreSQL migration/concurrency coverage and a Compose HTTPS
manual flow covering user restriction, targeted scheduling, Host policy
delivery, Host self-submission, Cloud-proxy budget exhaustion, and a
direct Host labelled unmetered.
- [x] 7.4 Run `openspec validate cloud-console-governance --strict` and resolve
all proposal, design, specification, and task validation errors.