Add OpenCode Zen support, model health/stats dashboard, and gateway fallback fixes

AI gateway:
- Add a generic, admin-selectable `client_profile` field on gateway_providers
  (e.g. "opencode") so a provider needing special request headers (OpenCode
  Zen's client-identity spoofing) is configured like any other provider, not
  hardcoded by name.
- Track per-(provider, model) reliability/speed/latency health in memory,
  seeded from the existing gateway_usage_ledger at startup - purely
  observational, never influences routing.
- New Stats tab on /admin/gateway: request volume, latency, per-model
  breakdowns, and reliability weight, charted with a vendored Chart.js and
  devplace's own theme tokens.
- Record which model a failed request actually fell back to
  (fallback_used_route), surfaced in the Recent Failures table.
- Stop excluding context_length errors from fallback, and skip a primary
  attempt outright when its known context window is already too small for
  the estimated request size, going straight to the fallback.
- gateway_usage_ledger's provider/fallback_used_route columns and indexes
  are ensured centrally in database/schema.py's init_db(), the single point
  of truth for this table's schema.
- Non-OpenAI upstream routing and client-model passthrough; trust only the
  upstream's own X-Gateway-Model header for served-model attribution.

Devii agent:
- Fix a real lockup: plan/verify tools could be individually disabled via
  the admin tool toggles while still being required by the protocol gate,
  permanently bricking any task that needed tools. They can no longer be
  disabled, and the gate now also checks the tool is actually offered.
- Fix compaction being silently calibrated for a 1M-token model while
  running a much smaller one: context budget is now percentage-based and
  the summarizer's own request is sized to fit the real model.
- Give a specific, actionable retry message when plan()'s own arguments get
  cut off by the output limit, and tighten its schema to discourage
  overlong plans.

Other:
- Backup service: offload completed backups to a remote Hetzner Storage Box.
- Container manager: fix orphan blob leaks from sync races, add a two-phase
  plan/execute `system prune` CLI command.
- Admin gateway UI: replace the JS-rendered model/provider tables with
  server-rendered forms and pages.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DhmEkvutuwtzFVcLbTrhdo
This commit is contained in:
2026-09-09 07:38:24 +02:00
co-authored by Claude Sonnet 5
parent b475e7d6ed
commit 569f1dcc64
67 changed files with 6151 additions and 814 deletions
+27 -1
View File
@@ -66,6 +66,32 @@ def test_gateway_page_requires_admin(seeded_db):
def test_gateway_page_renders_for_admin(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(f"{BASE_URL}/admin/gateway", headers={})
response = admin.get(f"{BASE_URL}/admin/gateway", headers={"Accept": "text/html"})
assert response.status_code == 200
assert "Gateway routing" in response.text
def test_gateway_page_tabs_default_and_select_content(seeded_db):
admin = admin_session(seeded_db)
default_page = admin.get(f"{BASE_URL}/admin/gateway", headers={"Accept": "text/html"})
assert "Model routes" in default_page.text
assert "GatewayAdmin" not in default_page.text
providers_page = admin.get(f"{BASE_URL}/admin/gateway?tab=providers", headers={"Accept": "text/html"})
assert "Providers" in providers_page.text
assert "Model routes" not in providers_page.text
quota_page = admin.get(f"{BASE_URL}/admin/gateway?tab=quota", headers={"Accept": "text/html"})
assert "Quota rules" in quota_page.text
assert "Model routes" not in quota_page.text
unknown_tab_page = admin.get(f"{BASE_URL}/admin/gateway?tab=bogus", headers={"Accept": "text/html"})
assert "Model routes" in unknown_tab_page.text
def test_gateway_page_json_reports_active_tab(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(f"{BASE_URL}/admin/gateway?tab=providers")
assert response.status_code == 200
assert response.json()["tab"] == "providers"
+107
View File
@@ -224,6 +224,113 @@ def test_model_route_fallback_rejects_self_reference(seeded_db):
assert response.json()["ok"] is False
def test_model_form_pages_require_admin(seeded_db):
assert (
requests.get(
f"{BASE_URL}/admin/gateway/models/new",
headers=JSON_gateway,
allow_redirects=False,
).status_code
== 401
)
key = member_key()
assert (
requests.get(
f"{BASE_URL}/admin/gateway/models/new",
headers={**JSON_gateway, "X-API-KEY": key},
allow_redirects=False,
).status_code
== 403
)
def test_model_add_edit_delete_via_page(seeded_db):
admin = admin_session(seeded_db)
source = _unique_gateway("modelpage")
new_page = admin.get(f"{BASE_URL}/admin/gateway/models/new")
assert new_page.status_code == 200
assert new_page.json()["is_edit"] is False
created = admin.post(
f"{BASE_URL}/admin/gateway/models/new",
data={
"source_model": source,
"target_model": "vendor/page",
"kind": "chat",
"price_output_per_m": "1.5",
"off_peak_start": "22:30",
"off_peak_end": "06:00",
"off_peak_discount_pct": "10",
"is_active": "1",
},
allow_redirects=False,
)
assert created.status_code == 302
assert created.headers["location"] == "/admin/gateway?tab=models"
edit_page = admin.get(f"{BASE_URL}/admin/gateway/models/{source}/edit")
assert edit_page.status_code == 200
form = edit_page.json()["form"]
assert form["target_model"] == "vendor/page"
assert form["off_peak_start"] == "22:30"
assert form["off_peak_end"] == "06:00"
edit_html = admin.get(f"{BASE_URL}/admin/gateway/models/{source}/edit", headers={"Accept": "text/html"})
assert f'value="{source}"' in edit_html.text
assert "readonly" in edit_html.text
updated = admin.post(
f"{BASE_URL}/admin/gateway/models/{source}/edit",
data={"target_model": "vendor/page2", "kind": "chat", "is_active": "1"},
allow_redirects=False,
)
assert updated.status_code == 302
relisted = admin.get(f"{BASE_URL}/admin/gateway/models").json()
row = next(m for m in relisted["models"] if m["source_model"] == source)
assert row["target_model"] == "vendor/page2"
deleted = admin.post(
f"{BASE_URL}/admin/gateway/models/{source}/delete", allow_redirects=False
)
assert deleted.status_code == 302
assert admin.get(f"{BASE_URL}/admin/gateway/models/{source}/edit").status_code == 404
def test_model_page_fallback_must_be_the_same_kind(seeded_db):
admin = admin_session(seeded_db)
embed_route = _unique_gateway("fbpage-embed")
chat_route = _unique_gateway("fbpage-chat")
admin.post(
f"{BASE_URL}/admin/gateway/models",
json={"source_model": embed_route, "target_model": "vendor/e", "kind": "embed"},
)
response = admin.post(
f"{BASE_URL}/admin/gateway/models/new",
data={
"source_model": chat_route,
"target_model": "vendor/c",
"kind": "chat",
"fallback_model": embed_route,
},
)
assert response.status_code == 400
assert "message" in response.json()["error"]
admin.delete(f"{BASE_URL}/admin/gateway/models/{embed_route}")
def test_model_edit_page_404_for_missing_model(seeded_db):
admin = admin_session(seeded_db)
missing = _unique_gateway("ghostmodel")
assert admin.get(f"{BASE_URL}/admin/gateway/models/{missing}/edit").status_code == 404
assert (
admin.post(f"{BASE_URL}/admin/gateway/models/{missing}/delete").status_code == 404
)
def test_model_route_validation(seeded_db):
admin = admin_session(seeded_db)
missing_target = admin.post(
+120
View File
@@ -0,0 +1,120 @@
# retoor <retoor@molodetz.nl>
import http.server
import json
import socket
import socketserver
import threading
import requests
from tests.conftest import BASE_URL
from tests.api.admin.gateway.index import (
JSON_gateway,
admin_session,
member_key,
_unique_gateway,
)
def _fake_upstream(status_code, payload):
sock = socket.socket()
sock.bind(("127.0.0.1", 0))
port = sock.getsockname()[1]
sock.close()
body = json.dumps(payload).encode() if payload is not None else b""
seen_auth = {}
class Handler(http.server.BaseHTTPRequestHandler):
def do_GET(self):
seen_auth["path"] = self.path
seen_auth["authorization"] = self.headers.get("Authorization")
self.send_response(status_code)
self.send_header("Content-Type", "application/json")
self.end_headers()
self.wfile.write(body)
def log_message(self, *args):
pass
httpd = socketserver.TCPServer(("127.0.0.1", port), Handler)
thread = threading.Thread(target=httpd.serve_forever, daemon=True)
thread.start()
return httpd, port, seen_auth
def test_provider_models_requires_admin(seeded_db):
assert (
requests.get(
f"{BASE_URL}/admin/gateway/provider-models",
headers=JSON_gateway,
allow_redirects=False,
).status_code
== 401
)
key = member_key()
assert (
requests.get(
f"{BASE_URL}/admin/gateway/provider-models",
headers={**JSON_gateway, "X-API-KEY": key},
allow_redirects=False,
).status_code
== 403
)
def test_provider_models_returns_the_list_on_success(seeded_db):
admin = admin_session(seeded_db)
name = _unique_gateway("modelsupstream").lower()
httpd, port, seen = _fake_upstream(
200, {"data": [{"id": "vendor/a"}, {"id": "vendor/b"}]}
)
try:
admin.post(
f"{BASE_URL}/admin/gateway/providers",
json={
"name": name,
"base_url": f"http://127.0.0.1:{port}/v1/chat/completions",
"api_key": "sk-fake",
},
)
response = admin.get(
f"{BASE_URL}/admin/gateway/provider-models", params={"provider": name}
)
assert response.status_code == 200
assert response.json()["models"] == ["vendor/a", "vendor/b"]
assert seen["path"] == "/v1/models"
assert seen["authorization"] == "Bearer sk-fake"
finally:
httpd.shutdown()
admin.delete(f"{BASE_URL}/admin/gateway/providers/{name}")
def test_provider_models_404_when_upstream_has_no_model_listing(seeded_db):
admin = admin_session(seeded_db)
name = _unique_gateway("modelsupstream404").lower()
httpd, port, _ = _fake_upstream(404, {})
try:
admin.post(
f"{BASE_URL}/admin/gateway/providers",
json={
"name": name,
"base_url": f"http://127.0.0.1:{port}/v1/chat/completions",
},
)
response = admin.get(
f"{BASE_URL}/admin/gateway/provider-models", params={"provider": name}
)
assert response.status_code == 404
finally:
httpd.shutdown()
admin.delete(f"{BASE_URL}/admin/gateway/providers/{name}")
def test_provider_models_404_for_unknown_provider(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(
f"{BASE_URL}/admin/gateway/provider-models",
params={"provider": _unique_gateway("ghostprovider")},
)
assert response.status_code == 404
+112
View File
@@ -82,3 +82,115 @@ def test_provider_name_validation(seeded_db):
)
assert bad.status_code == 400
assert bad.json()["ok"] is False
def test_provider_form_pages_require_admin(seeded_db):
name = _unique_gateway("provpage").lower()
assert (
requests.get(
f"{BASE_URL}/admin/gateway/providers/new",
headers=JSON_gateway,
allow_redirects=False,
).status_code
== 401
)
key = member_key()
assert (
requests.get(
f"{BASE_URL}/admin/gateway/providers/new",
headers={**JSON_gateway, "X-API-KEY": key},
allow_redirects=False,
).status_code
== 403
)
assert (
requests.post(
f"{BASE_URL}/admin/gateway/providers/new",
data={"name": name},
headers={**JSON_gateway, "X-API-KEY": key},
allow_redirects=False,
).status_code
== 403
)
def test_provider_add_edit_delete_via_page(seeded_db):
admin = admin_session(seeded_db)
name = _unique_gateway("provpage").lower()
new_page = admin.get(f"{BASE_URL}/admin/gateway/providers/new")
assert new_page.status_code == 200
assert new_page.json()["is_edit"] is False
created = admin.post(
f"{BASE_URL}/admin/gateway/providers/new",
data={
"name": name,
"base_url": "https://page.example/v1/chat/completions",
"api_key": "sk-page",
"is_active": "1",
},
allow_redirects=False,
)
assert created.status_code == 302
assert created.headers["location"] == "/admin/gateway?tab=providers"
edit_page = admin.get(f"{BASE_URL}/admin/gateway/providers/{name}/edit")
assert edit_page.status_code == 200
edit_body = edit_page.json()
assert edit_body["is_edit"] is True
assert edit_body["form"]["base_url"] == "https://page.example/v1/chat/completions"
edit_html = admin.get(f"{BASE_URL}/admin/gateway/providers/{name}/edit", headers={"Accept": "text/html"})
assert f'value="{name}"' in edit_html.text
assert "readonly" in edit_html.text
updated = admin.post(
f"{BASE_URL}/admin/gateway/providers/{name}/edit",
data={"base_url": "https://page2.example/v1/chat/completions", "is_active": "0"},
allow_redirects=False,
)
assert updated.status_code == 302
relisted = admin.get(f"{BASE_URL}/admin/gateway/providers").json()
match = next(p for p in relisted["providers"] if p["name"] == name)
assert match["base_url"] == "https://page2.example/v1/chat/completions"
assert match["is_active"] is False
deleted = admin.post(
f"{BASE_URL}/admin/gateway/providers/{name}/delete", allow_redirects=False
)
assert deleted.status_code == 302
assert admin.get(f"{BASE_URL}/admin/gateway/providers/{name}/edit").status_code == 404
def test_provider_page_validation_error_rerenders_with_message(seeded_db):
admin = admin_session(seeded_db)
response = admin.post(
f"{BASE_URL}/admin/gateway/providers/new",
data={"name": "has spaces!"},
)
assert response.status_code == 400
assert "message" in response.json()["error"]
def test_provider_page_validation_error_shows_banner_in_html(seeded_db):
admin = admin_session(seeded_db)
response = admin.post(
f"{BASE_URL}/admin/gateway/providers/new",
data={"name": "has spaces!"},
headers={"Accept": "text/html"},
)
assert response.status_code == 400
assert "gw-error" in response.text
assert "letters, numbers, hyphen, underscore" in response.text
def test_provider_edit_page_404_for_missing_provider(seeded_db):
admin = admin_session(seeded_db)
missing = _unique_gateway("ghostprov").lower()
assert admin.get(f"{BASE_URL}/admin/gateway/providers/{missing}/edit").status_code == 404
assert (
admin.post(f"{BASE_URL}/admin/gateway/providers/{missing}/delete").status_code
== 404
)
+119
View File
@@ -11,6 +11,7 @@ from tests.api.admin.gateway.index import (
JSON_gateway,
admin_session,
member_key,
_unique_gateway,
)
from tests.conftest import BASE_URL
@@ -160,3 +161,121 @@ def test_spend_recorded_after_a_reset_counts_again(seeded_db):
)
_burn(owner, "appa", 0.5)
assert _spent(owner, "appa") == 0.5
def test_quota_rule_form_pages_require_admin(seeded_db):
assert (
requests.get(
f"{BASE_URL}/admin/gateway/quota-rules/new",
headers=JSON_gateway,
allow_redirects=False,
).status_code
== 401
)
key = member_key()
assert (
requests.get(
f"{BASE_URL}/admin/gateway/quota-rules/new",
headers={**JSON_gateway, "X-API-KEY": key},
allow_redirects=False,
).status_code
== 403
)
def test_quota_rule_add_edit_delete_reset_via_page(seeded_db):
admin = admin_session(seeded_db)
owner = _owner()
app_ref = _unique_gateway("qrapp").lower()
new_page = admin.get(f"{BASE_URL}/admin/gateway/quota-rules/new")
assert new_page.status_code == 200
assert new_page.json()["is_edit"] is False
created = admin.post(
f"{BASE_URL}/admin/gateway/quota-rules/new",
data={
"owner_kind": "user",
"owner_id": owner,
"app_reference": app_ref,
"limit_usd": "2.5",
"is_active": "1",
"label": "page test",
},
allow_redirects=False,
)
assert created.status_code == 302
assert created.headers["location"] == "/admin/gateway?tab=quota"
rules = admin.get(f"{BASE_URL}/admin/gateway/quota-rules").json()["rules"]
rule = next(r for r in rules if r["owner_id"] == owner and r["app_reference"] == app_ref)
uid = rule["uid"]
assert rule["limit_usd"] == 2.5
edit_page = admin.get(f"{BASE_URL}/admin/gateway/quota-rules/{uid}/edit")
assert edit_page.status_code == 200
assert edit_page.json()["form"]["limit_usd"] == "2.5"
updated = admin.post(
f"{BASE_URL}/admin/gateway/quota-rules/{uid}/edit",
data={
"owner_kind": "user",
"owner_id": owner,
"app_reference": app_ref,
"limit_usd": "5.0",
"is_active": "1",
"label": "page test updated",
},
allow_redirects=False,
)
assert updated.status_code == 302
rules = admin.get(f"{BASE_URL}/admin/gateway/quota-rules").json()["rules"]
rule = next(r for r in rules if r["uid"] == uid)
assert rule["limit_usd"] == 5.0
assert rule["label"] == "page test updated"
_burn(owner, app_ref, 1.0)
assert _spent(owner, app_ref) == 1.0
reset_response = admin.post(
f"{BASE_URL}/admin/gateway/quota-rules/{uid}/reset", allow_redirects=False
)
assert reset_response.status_code == 302
assert _spent(owner, app_ref) == 0.0
deleted = admin.post(
f"{BASE_URL}/admin/gateway/quota-rules/{uid}/delete", allow_redirects=False
)
assert deleted.status_code == 302
assert (
admin.get(f"{BASE_URL}/admin/gateway/quota-rules/{uid}/edit").status_code == 404
)
def test_quota_rule_page_requires_a_dimension(seeded_db):
admin = admin_session(seeded_db)
response = admin.post(
f"{BASE_URL}/admin/gateway/quota-rules/new",
data={"limit_usd": "1.0"},
)
assert response.status_code == 400
assert "message" in response.json()["error"]
def test_quota_rule_edit_page_404_for_missing_rule(seeded_db):
admin = admin_session(seeded_db)
from devplacepy.utils import generate_uid
missing = generate_uid()
assert (
admin.get(f"{BASE_URL}/admin/gateway/quota-rules/{missing}/edit").status_code
== 404
)
assert (
admin.post(f"{BASE_URL}/admin/gateway/quota-rules/{missing}/delete").status_code
== 404
)
assert (
admin.post(f"{BASE_URL}/admin/gateway/quota-rules/{missing}/reset").status_code
== 404
)
+82
View File
@@ -0,0 +1,82 @@
# retoor <retoor@molodetz.nl>
import requests
from tests.api.admin.gateway.index import JSON_gateway, admin_session, member_key
from tests.conftest import BASE_URL
def test_stats_tab_renders_for_admin(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(
f"{BASE_URL}/admin/gateway?tab=stats", headers={"Accept": "text/html"}
)
assert response.status_code == 200
assert "Model pool" in response.text
assert "Model routes" not in response.text
def test_stats_data_requires_admin(seeded_db):
assert (
requests.get(
f"{BASE_URL}/admin/gateway/stats/data", headers=JSON_gateway, allow_redirects=False
).status_code
== 401
)
key = member_key()
assert (
requests.get(
f"{BASE_URL}/admin/gateway/stats/data",
headers={**JSON_gateway, "X-API-KEY": key},
allow_redirects=False,
).status_code
== 403
)
def test_stats_data_shape_for_admin(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(f"{BASE_URL}/admin/gateway/stats/data?range=24h", headers=JSON_gateway)
assert response.status_code == 200
data = response.json()
for key in (
"totals",
"per_model",
"per_endpoint",
"per_provider",
"timeseries",
"status_codes",
"streaming_split",
"failure_reasons",
"hourly_distribution",
"recent_failures",
"latency_histogram",
"tokens_per_second_histogram",
"bucket_seconds",
):
assert key in data
def test_stats_data_rejects_unknown_range(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(f"{BASE_URL}/admin/gateway/stats/data?range=nonsense", headers=JSON_gateway)
assert response.status_code == 400
def test_stats_models_returns_a_list(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(f"{BASE_URL}/admin/gateway/stats/models", headers=JSON_gateway)
assert response.status_code == 200
assert isinstance(response.json()["models"], list)
def test_stats_model_detail_for_an_unknown_model(seeded_db):
admin = admin_session(seeded_db)
response = admin.get(
f"{BASE_URL}/admin/gateway/stats/model/default/never-called-model?range=24h",
headers=JSON_gateway,
)
assert response.status_code == 200
data = response.json()
assert data["summary"]["total_requests"] == 0
assert data["health"] is None
+115
View File
@@ -0,0 +1,115 @@
# retoor <retoor@molodetz.nl>
import time
import pytest
import requests
from tests.conftest import BASE_URL
from devplacepy.database import clear_settings_cache, get_table, refresh_snapshot
from devplacepy.services.devii import tool_prefs
JSON = {"Accept": "application/json"}
_counter = [0]
@pytest.fixture(scope="module", autouse=True)
def _settings(app_server):
from devplacepy.database import set_setting
set_setting("rate_limit_per_minute", "1000000")
set_setting("registration_open", "1")
yield
def _admin(seeded_db):
refresh_snapshot()
key = get_table("users").find_one(username="alice_test")["api_key"]
s = requests.Session()
s.headers.update({"X-API-KEY": key, **JSON})
return s
def _member():
_counter[0] += 1
name = f"dtmem{int(time.time() * 1000)}{_counter[0]}"
requests.post(
f"{BASE_URL}/auth/signup",
data={
"username": name,
"email": f"{name}@t.dev",
"password": "secret123",
"confirm_password": "secret123",
"birth_date": "1990-01-01",
"accept_terms": "1",
},
allow_redirects=True,
)
refresh_snapshot()
s = requests.Session()
s.headers.update(
{"X-API-KEY": get_table("users").find_one(username=name)["api_key"], **JSON}
)
return s
def _all_tool_names():
return set(tool_prefs.GROUPS_BY_TOOL_NAME)
def test_devii_service_page_has_tools_tab(seeded_db):
admin = _admin(seeded_db)
r = admin.get(f"{BASE_URL}/admin/services/devii")
assert r.status_code == 200
assert 'data-tab="tools"' in r.text
assert "devii-tools-form" in r.text
def test_other_service_page_has_no_tools_tab(seeded_db):
admin = _admin(seeded_db)
r = admin.get(f"{BASE_URL}/admin/services/news")
assert r.status_code == 200
assert 'data-tab="tools"' not in r.text
def test_admin_can_disable_and_reenable_a_tool(seeded_db):
admin = _admin(seeded_db)
all_names = _all_tool_names()
try:
enabled = sorted(all_names - {"create_post"})
r = admin.post(
f"{BASE_URL}/admin/services/devii/tools",
data=[("enabled", name) for name in enabled],
)
assert r.status_code == 200, r.text[:300]
assert r.json()["ok"] is True
clear_settings_cache()
assert tool_prefs.disabled_tool_names() == frozenset({"create_post"})
r2 = admin.post(
f"{BASE_URL}/admin/services/devii/tools",
data=[("enabled", name) for name in all_names],
)
assert r2.status_code == 200
clear_settings_cache()
assert tool_prefs.disabled_tool_names() == frozenset()
finally:
clear_settings_cache()
tool_prefs.set_disabled_tool_names(set())
def test_member_cannot_save_tool_config(app_server):
member = _member()
r = member.post(
f"{BASE_URL}/admin/services/devii/tools",
data={"enabled": "create_post"},
allow_redirects=False,
)
assert r.status_code in (302, 303, 403)
def test_guest_cannot_save_tool_config(app_server):
r = requests.post(
f"{BASE_URL}/admin/services/devii/tools",
data={"enabled": "create_post"},
allow_redirects=False,
)
assert r.status_code in (302, 303, 401)