ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
Python ai-agentsamdcomfyuidockerllama-cppllmlocal-ain8nnvidiaopen-webuiragself-hostedspeech-to-textstrix-halotext-to-speechworkflow-automation
"""Public envelopes and host controller boundary, not installed proof."""
import copy
import importlib.util
from pathlib import Path
import sys
import pytest
ROOT = Path(__file__).resolve().parents[2]
sys.path.insert(0, str(ROOT / "bin"))
from pixel_settings import host_api, public
from pixel_provider.store import StoreError
def response(status="not-applied"):
caps = {"providerContextTokens": 65536, "providerMaxOutputTokens": 8192,
"activeContextTokens": 32768, "activeMaxOutputTokens": 4096, "backendContextTokens": None,
"capacitySource": "owner-declared", "supportedThinkingLevels": [], "samplingSupported": False,
"pixelOnlyRuntime": True}
value = {"schemaVersion": 1, "status": status, "revision": "a" * 64, "settingsRevision": 3,
"appliedRevision": None, "capabilities": caps, "pending": False, "lastVerifiedAt": None, "reason": None}
if status in ("applied", "saved-changes"):
value.update(appliedRevision=3 if status == "applied" else 2, lastVerifiedAt="2026-09-08T16:00:00Z")
if status == "restored": value.update(lastVerifiedAt="2026-09-08T16:00:00Z")
if status == "pending": value.update(pending=True, capabilities=None)
if status == "unavailable": return public.unavailable("settings-controller-unavailable")
return value
def dashboard_public():
spec = importlib.util.spec_from_file_location("_settings_dashboard_public_parity", ROOT / "extensions/services/dashboard-api/pixel_settings_public.py")
result = importlib.util.module_from_spec(spec)
spec.loader.exec_module(result)
return result
@pytest.mark.parametrize("state", ["not-applied", "applied", "saved-changes", "restored", "pending", "unavailable"])
def test_host_and_dashboard_runtime_contract_agree(state):
value = response(state)
assert public.normalize_runtime(value) == dashboard_public().normalize_runtime(value) == value
@pytest.mark.parametrize("state,field,value", [
("applied", "schemaVersion", True), ("applied", "settingsRevision", True), ("applied", "appliedRevision", 2),
("applied", "lastVerifiedAt", "2026-99-08T00:00:00Z"), ("applied", "pending", True),
("saved-changes", "appliedRevision", 3), ("not-applied", "lastVerifiedAt", "2026-09-08T16:00:00Z"),
("pending", "pending", False), ("pending", "appliedRevision", 3),
("unavailable", "pending", False), ("unavailable", "revision", "a" * 64),
("unavailable", "reason", "secret/key/with/path"), ("not-applied", "revision", "bad"),
("not-applied", "private", "do-not-echo"),
("restored", "lastVerifiedAt", None), ("restored", "appliedRevision", 3),
])
def test_inconsistent_runtime_metadata_is_rejected_in_both_boundaries(state, field, value):
document = response(state)
document[field] = value
for normalize in (public.normalize_runtime, dashboard_public().normalize_runtime):
with pytest.raises(ValueError): normalize(document)
@pytest.mark.parametrize("field,value", [("activeContextTokens", True), ("providerContextTokens", 0),
("supportedThinkingLevels", ["high", "high"]), ("supportedThinkingLevels", [True]),
("backendContextTokens", -1), ("capacitySource", "backend-observed"), ("samplingSupported", 1),
("activeMaxOutputTokens", 999999), ("credential", "do-not-echo")])
def test_capability_checks_match_without_upgrading_declared_limits(field, value):
document = response()
document["capabilities"][field] = value
for normalize in (public.normalize_runtime, dashboard_public().normalize_runtime):
with pytest.raises(ValueError): normalize(document)
def test_public_projection_does_not_share_mutable_capability_list():
value = response()
for normalize in (public.normalize_runtime, dashboard_public().normalize_runtime):
result = normalize(value)
result["capabilities"]["supportedThinkingLevels"].append("high")
assert value["capabilities"]["supportedThinkingLevels"] == []
def test_host_passes_actual_data_dir_not_http_selected_path(tmp_path):
body = {"operation": "apply", "revision": "b" * 64, "settingsRevision": 3}
calls = []
def request(*args, **kwargs):
calls.append((args, kwargs))
return 200, {"outcome": "applied", "appliedRevision": 3}
assert host_api.runtime_change(tmp_path, body, request=request)["outcome"] == "applied"
assert calls == [(("settings-change", body), {"settings_data_dir": tmp_path})]
@pytest.mark.parametrize("field,value", [("capabilities", {}), ("data_dir_id", "a" * 64), ("path", "/etc"),
("operation", "exec"), ("settingsRevision", True), ("revision", "bad")])
def test_invalid_change_does_not_call_controller(tmp_path, field, value):
body = {"operation": "apply", "revision": "b" * 64, "settingsRevision": 3, field: value}
with pytest.raises(StoreError, match="invalid-request"):
host_api.runtime_change(tmp_path, body, request=lambda *_args, **_kwargs: pytest.fail("controller called"))
def test_lost_controller_reply_is_unknown_not_false_success_or_retry(tmp_path):
calls = []
def failed(*args, **_kwargs):
calls.append(args)
raise OSError("private failure detail")
value = host_api.runtime_status(tmp_path, request=failed)
assert value["status"] == "unavailable" and value["pending"] is None and value["revision"] is None
body = {"operation": "apply", "revision": "b" * 64, "settingsRevision": 3}
with pytest.raises(OSError): host_api.runtime_change(tmp_path, body, request=failed)
assert len(calls) == 2
def test_fresh_store_code_survives_both_public_status_boundaries(tmp_path):
value = host_api.runtime_status(tmp_path, request=lambda *_args, **_kwargs:
(409, {"error": "settings-store-not-initialized"}))
expected = public.unavailable("settings-store-not-initialized")
assert value == dashboard_public().normalize_runtime(value) == expected
assert not (tmp_path / "pixel-providers").exists()
def test_wrong_completed_revision_cannot_be_reported_as_success(tmp_path):
body = {"operation": "apply", "revision": "b" * 64, "settingsRevision": 3}
with pytest.raises(ValueError, match="revision-mismatch"):
host_api.runtime_change(tmp_path, body, request=lambda *_args, **_kwargs: (200, {"outcome": "applied", "appliedRevision": 4}))
def test_saved_marker_is_honest_and_older_host_still_readable(tmp_path):
value = host_api.get_settings(tmp_path)
assert value["runtime"] == {"status": "not-inspected", "reason": "runtime-status-separate"}
assert dashboard_public().normalize_response(value) == value
old = copy.deepcopy(value)
old["runtime"] = {"status": "not-applied", "reason": "settings-runtime-not-integrated"}
assert dashboard_public().normalize_response(old) == old