q792602257/agentic-mobile-control
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

feat: surface task execution progress across Host Agent and Cloud

Browse Source 
Host Agent now persists step-level execution detail locally (via a real
TaskMetadataStore/Timeline wired into TaskRunner) and reports a bounded
in-progress snapshot piggybacked on lease renewal. Cloud persists that
snapshot per active assignment and exposes it through the existing task
list/detail query path; Cloud Console renders it as a live badge. Host
Agent's local console gains authenticated, read-only task list and
detail/timeline pages (same-origin, server-rendered) with inlined
screenshots.

Also fixes a pre-existing gap in the shared Timeline: the actual
per-step LLM prompt is now recorded instead of the task goal, benefiting
both Runtime and Host Agent consoles. When a host uses the cloud planner
transport, each decide call's prompt and resulting tool decision are
durably logged in a new planner_decision_log table (with bounded
retention) and browsable from Cloud Console; direct-transport hosts
explicitly surface a "not reported" state.

Includes Alembic migrations 0008 (progress columns on scheduled_tasks)
and 0009 (planner_decision_log), bounded Host-Agent-local retention,
dual-backend repository parity, and Vitest + pytest coverage. Task 6.5
(manual end-to-end device verification) remains.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
This commit is contained in:
q792602257
2026-07-14 12:47:49 +08:00
co-authored by Claude Opus 4.6
parent c049c3c1b1
commit ec261d57c2
59 changed files with 3801 additions and 122 deletions
Download Patch File Download Diff File Expand all files Collapse all files
tests/test_cloud_planner_decision_endpoint.py
+163 -9
View File
@@ -1,10 +1,14 @@
from __future__ import annotations
from datetime import UTC, datetime, timedelta
from fastapi import FastAPI
from fastapi.testclient import TestClient
from sqlalchemy.orm import Session
from cloud.auth import BearerCredential, ConfiguredBearerAuthProvider
from cloud.config import CloudConfig
from cloud.db_models import TaskAttemptRow
from cloud.internal_api.api import create_internal_router
from cloud.pool import DevicePool
from cloud.store import CloudStore
@@ -49,7 +53,7 @@ class _FakeToolCallingClient:
def _build_client(
tmp_path, *, fake_client: _FakeToolCallingClient
) -> tuple[TestClient, _FakeToolCallingClient]:
) -> tuple[TestClient, _FakeToolCallingClient, DevicePool]:
pool = DevicePool(
CloudStore(tmp_path / "internal.sqlite3"),
CloudConfig(stale_after_seconds=60),
@@ -68,7 +72,7 @@ def _build_client(
planner_client_factory=lambda: fake_client,
)
)
return TestClient(app), fake_client
return TestClient(app), fake_client, pool
def _decision_payload(**overrides: object) -> dict[str, object]:
@@ -94,7 +98,7 @@ def test_authenticated_host_resolves_planner_decision(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={"x": 1, "y": 2})
)
client, fake_client = _build_client(tmp_path, fake_client=fake_client)
client, fake_client, _pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
@@ -113,7 +117,7 @@ def test_planner_decision_decodes_screenshot_base64(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client = _build_client(tmp_path, fake_client=fake_client)
client, fake_client, _pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
@@ -129,7 +133,7 @@ def test_invalid_screenshot_base64_is_rejected(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client = _build_client(tmp_path, fake_client=fake_client)
client, fake_client, _pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
@@ -145,7 +149,7 @@ def test_unauthenticated_request_is_rejected(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client = _build_client(tmp_path, fake_client=fake_client)
client, fake_client, _pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
@@ -160,7 +164,7 @@ def test_foreign_host_token_cannot_request_another_hosts_decision(tmp_path) -> N
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client = _build_client(tmp_path, fake_client=fake_client)
client, fake_client, _pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
@@ -176,7 +180,7 @@ def test_path_and_payload_host_id_mismatch_is_rejected(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client = _build_client(tmp_path, fake_client=fake_client)
client, fake_client, _pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
@@ -190,7 +194,7 @@ def test_path_and_payload_host_id_mismatch_is_rejected(tmp_path) -> None:
def test_provider_failure_returns_structured_error_without_crashing(tmp_path) -> None:
fake_client = _FakeToolCallingClient(error="anthropic timed out")
client, fake_client = _build_client(tmp_path, fake_client=fake_client)
client, fake_client, _pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
@@ -203,3 +207,153 @@ def test_provider_failure_returns_structured_error_without_crashing(tmp_path) ->
"code": "planner_unavailable",
"detail": "anthropic timed out",
}
def _seed_attempt(pool: DevicePool, task_id: str, host_id: str = "host-a") -> None:
"""Insert a task_attempts row so _validate_planner_context passes."""
now = datetime.now(tz=UTC)
with Session(pool.store.engine) as session, session.begin():
session.add(
TaskAttemptRow(
task_id=task_id,
attempt=1,
lease_id="lease-x",
host_id=host_id,
device_id="device-x",
status="dispatched",
lease_expires_at=(now + timedelta(minutes=5)).isoformat(),
created_at=now.isoformat(),
completed_at=None,
failure_reason=None,
result_json=None,
)
)
def test_successful_decision_is_persisted_with_correct_fields(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={"x": 10, "y": 20})
)
client, fake_client, pool = _build_client(tmp_path, fake_client=fake_client)
_seed_attempt(pool, "task-log-1")
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
headers={"Authorization": "Bearer token-a"},
json=_decision_payload(
task_id="task-log-1",
attempt=1,
lease_id="lease-x",
system_prompt="you are a test planner",
user_prompt="tap the button now",
),
)
assert response.status_code == 200
decisions = pool.store.list_planner_decisions(task_id="task-log-1", attempt=1)
assert len(decisions) == 1
row = decisions[0]
assert row.step_index == 1
assert row.system_prompt == "you are a test planner"
assert row.user_prompt == "tap the button now"
assert row.tool_name == "tap"
assert '"x": 10' in row.arguments_json
assert '"y": 20' in row.arguments_json
def test_failed_decision_persists_nothing(tmp_path) -> None:
fake_client = _FakeToolCallingClient(error="model is down")
client, fake_client, pool = _build_client(tmp_path, fake_client=fake_client)
_seed_attempt(pool, "task-log-fail")
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
headers={"Authorization": "Bearer token-a"},
json=_decision_payload(
task_id="task-log-fail",
attempt=1,
lease_id="lease-x",
),
)
assert response.status_code == 502
assert pool.store.list_planner_decisions(task_id="task-log-fail", attempt=1) == []
def test_screenshot_bytes_are_never_persisted_to_decision_log(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client, pool = _build_client(tmp_path, fake_client=fake_client)
_seed_attempt(pool, "task-log-screenshot")
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
headers={"Authorization": "Bearer token-a"},
json=_decision_payload(
task_id="task-log-screenshot",
attempt=1,
lease_id="lease-x",
screenshot_base64="aGVsbG8gd29ybGQ=",
),
)
assert response.status_code == 200
decisions = pool.store.list_planner_decisions(
task_id="task-log-screenshot", attempt=1
)
assert len(decisions) == 1
# The row has no screenshot column at all — verify the raw row text
# does not contain the screenshot bytes.
row_repr = repr(decisions[0])
assert "hello world" not in row_repr
assert "aGVsbG8" not in row_repr
def test_decision_without_task_context_is_not_persisted(tmp_path) -> None:
"""When task_id/attempt are absent, the log insert is skipped."""
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client, pool = _build_client(tmp_path, fake_client=fake_client)
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
headers={"Authorization": "Bearer token-a"},
json=_decision_payload(), # no task_id / attempt / lease_id
)
assert response.status_code == 200
# No task_id to query by — verify no rows exist at all via direct SQL.
from sqlalchemy import func, select
from cloud.db_models import PlannerDecisionLogRow
with Session(pool.store.engine) as session:
count = session.scalar(select(func.count()).select_from(PlannerDecisionLogRow))
assert count == 0
def test_multiple_decisions_get_incrementing_step_index(tmp_path) -> None:
fake_client = _FakeToolCallingClient(
decision=ToolCallDecision(tool_name="tap", arguments={})
)
client, fake_client, pool = _build_client(tmp_path, fake_client=fake_client)
_seed_attempt(pool, "task-log-multi")
for i in range(3):
response = client.post(
"/internal/v1/hosts/host-a/planner/decide",
headers={"Authorization": "Bearer token-a"},
json=_decision_payload(
task_id="task-log-multi",
attempt=1,
lease_id="lease-x",
user_prompt=f"step {i}",
),
)
assert response.status_code == 200
decisions = pool.store.list_planner_decisions(task_id="task-log-multi", attempt=1)
assert [d.step_index for d in decisions] == [1, 2, 3]
assert [d.user_prompt for d in decisions] == ["step 0", "step 1", "step 2"]