Merge main (custom features) into release/custom-0.1.95 (upstream v0.1.95).
New upstream features: group subscription binding, multi-dimension quota (daily/weekly/total),
allow_messages_dispatch, default_mapped_model, recover state API.
Our customizations: simulate_claude_max_enabled, usage status detection, 403 validation handling.
The ChatGPT backend-api codex/responses endpoint requires `input` to be
an array, but the OpenAI Responses API spec allows it to be a plain string.
When a client sends a string input, sub2api now converts it to the expected
message array format. Empty/whitespace-only strings become an empty array
to avoid triggering a 400 "Input must be a list" error.
- VALIDATION_REQUIRED now marks account as error (permanent) instead of temp unschedulable (10min)
- When usage fetch fails with 401 but account has 403 error, show correct forbidden/banned status instead of needs_reauth
- Add AdminResetQuota service method to reset daily/weekly usage windows
- Add POST /api/v1/admin/subscriptions/:id/reset-quota handler and route
- Add resetQuota API function in frontend subscriptions client
- Add reset quota button, confirmation dialog, and handlers in SubscriptionsView
- Add i18n keys for reset quota feature in zh and en locales
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
When the Antigravity quota API returns non-403 errors (e.g. 401
UNAUTHENTICATED), return a UsageInfo with an error field instead of
propagating the error as HTTP 500. This matches Antigravity Manager's
behavior of gracefully displaying errors.
Two SSE scanners in openai_gateway_messages.go were hardcoded to 1MB
while all other scanners use defaultMaxLineSize (500MB) with config
override. This caused Responses API streams to fail on large payloads.
Two SSE scanners in openai_gateway_messages.go were hardcoded to 1MB
while all other scanners use defaultMaxLineSize (500MB) with config
override. This caused Responses API streams to fail on large payloads.
Add IsOpenAIPassthroughEnabled() bypass in two places:
- SelectAccountWithLoadAwareness Layer 2 candidate filtering (line 1285)
- resolveFreshSchedulableOpenAIAccount model check (line 1446)
Without these checks, passthrough accounts are filtered out for models
like gpt-5.4 that aren't in the account's model_mapping, causing 503.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The isModelSupportedByAccount function lost the OpenAI passthrough
check during the upstream merge, causing GPT-5.4 and other models
to be rejected for passthrough accounts.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
When the service restarts, concurrency slots from the old process
remain in Redis, causing phantom occupancy. On startup, scan all
concurrency sorted sets and remove members with non-current process
prefix, then clear orphaned wait queue counters.
Uses Go-side SCAN to discover keys (compatible with Redis client
prefix hooks in tests), then passes them to a Lua script for
atomic member-level cleanup.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
4K image base64 data can exceed 40MB limit, causing "bufio.Scanner:
token too long" errors. Scanner is adaptive (starts at 64KB, grows
as needed), so increasing the cap has no impact on normal responses.
4K image base64 data can exceed 40MB limit, causing "bufio.Scanner:
token too long" errors. Scanner is adaptive (starts at 64KB, grows
as needed), so increasing the cap has no impact on normal responses.