Parse OpenAI's native cache_write_tokens (chat prompt_tokens_details /
responses input_tokens_details), bill it at the cache-creation ratio, and
clamp the uncached prompt remainder at zero since cached + cache-write can
exceed prompt_tokens. Propagate the field through chat/responses/claude
format conversions and tiered expression billing (cc variable).
* refactor: consolidate relay protocol converters
* refactor relayconvert text converters
* feat: refine relay converters and advanced custom routing
* refactor: enhance logging and add thought signature handling for Gemini requests
* refactor: enhance channel cache and pricing endpoint handling for advanced custom models
* feat: preserve billing usage semantics
* feat: add protocol-aware billing usage
* Delete useless files
* chore: update action versions in workflow files
* chore: update Docker action versions in workflow files
* fix: harden billing usage settlement and hot-path route matching
- estimate Gemini completion tokens locally when billable usageMetadata is
prompt-only but output content was received (e.g. client aborts the stream
before the final chunk), and rebuild the attached billing_usage as estimated
so settlement does not bill zero output tokens
- guard NewClaudeMessagesBillingUsage against all-zero ClaudeUsage, matching
the OpenAI/Gemini constructors, so a zero billing_usage cannot override a
non-zero top-level usage during settlement
- cache compiled advanced-custom route model regexes; they run on the request
hot path and were recompiled per request
- move the effectiveBillingUsage remap to PostTextConsumeQuota only, and
document that calculateTextQuotaSummary expects remapped usage
- document the updatePricingLock -> channelSyncLock lock ordering that
InitChannelCache/CacheUpdateChannel rely on, and the aux-struct pitfall in
GeminiChatResponse.UnmarshalJSON
Revert 6869cd94b (perf(web): align table badge spacing).
Restores px padding on status badges so usage-log timing/duration
pills are not flush against the edges.
The custom home page URL iframe only allows allow-forms/allow-popups/
allow-popups-to-escape-sandbox/allow-scripts, without allow-top-navigation*.
As a result, target="_top" nav/menu links inside the (admin-configured,
trusted) embedded page cannot navigate the top-level window on desktop
browsers, while some mobile browsers still allow it via allow-popups —
causing inconsistent behavior rather than an intended restriction.
Add allow-top-navigation-by-user-activation so user-clicked top-level
links work consistently across devices. This token only permits
user-activated top navigation and does NOT grant same-origin access,
so it avoids the security concern of allow-same-origin.
Co-authored-by: 贺. <kuang@M1.local>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Replace collapsible card details with always-visible label/value rows, add badge-list full display for cards, and polish pricing/channel toolbars and mobile cards.
- remove built-in horizontal padding from status badges so table columns align with headers.
- drop legacy negative-margin compensation from badge cells and affected table renderers.
- clean touched table components to satisfy lint rules around type imports, keys, and JSX flow.
- keep input focus and invalid rings inside the shared Input component to avoid clipping in constrained containers.
- make the sync channel selector table use a fixed header, internal body scrolling, and fixed pagination inside the dialog.
- align status cells with their header and translate channel status labels through existing i18n keys.
- batch select and unselect operations into single state updates to reduce lag with large model lists.
- align current price and upstream preset rows by rendering the same ordered field set.
- switch the table to a fixed header, internal body scrolling, and fixed pagination layout with compact upstream header counts.
- hide synthesized preset internal IDs and count only effective selectable resolutions.
- expose playground controls for sampling, penalty, token, and seed parameters.
- reuse the existing persisted config and parameter-enabled state so requests only send enabled values.
- add localized parameter copy and keep the panel scrollable on short screens.
The ratio column normalized input to number on every keystroke, so
typing 0.05 got stuck at 0.0 which Number() collapses to 0. Keep the
value as a string draft like the top-up ratio column and only convert
to number when serializing.
Fixes#5986
* feat(subscription): add admin quota reset actions
* fix(subscription): keep quota reset in plan row actions
* refactor(subscription): move user subscription actions into menu
Thread int32 saturation clamps from tiered settlement and video task
recompute into the consume/task logs under admin_info, so oversized or
malformed billing inputs stay auditable. Clamp negative audio duration
before token conversion and gate the saturation UI markers on admin.
- replace per-model test text buttons with icon-only actions matching the channel list.
- let the action column shrink to its icon content to reduce table width.
- split model test status from result details so failures and latency no longer crowd one column.
- move batch progress into toast updates to keep the dialog height stable during tests.
- consolidate the channel title and model actions to reduce vertical churn.
* fix(openai): harden Chat-to-Responses compatibility
Add a shared Responses-to-Chat stream state machine and use it from the OpenAI relay path. Preserve assistant text alongside tool calls, bind tool argument deltas by output_index, map incomplete finish reasons, support reasoning/custom tool events, and buffer upstream SSE for non-stream Chat clients.
Add deterministic service tests and relay SSE tests for the conversion path.
Related to #5745.
* refactor: rename openaicompat to relayconvert for improved clarity
* feat(gemini): support responses request conversion
* feat: add responses to chat conversion support
* fix: harden responses chat conversion edge cases
* fix(markdown): render announcement markdown consistently
- support soft line breaks for announcement markdown without changing the default parser behavior.
- add explicit markdown element styles so lists, tables, code blocks, and quotes render correctly when typography styles are unavailable.
- apply the announcement markdown mode in both the popover and detail dialog for consistent display.
* refactor(markdown): simplify fallback markdown styles
- remove duplicate typography utility classes now covered by explicit markdown element fallbacks.
- keep the markdown renderer behavior unchanged while reducing class noise.
- modernize small helper expressions to satisfy targeted lint checks.
* fix(content): render custom HTML consistently
- add shared rich content rendering so custom HTML and Markdown use the same path across public pages and announcements.
- reuse common URL and HTML detection instead of duplicating content format checks per page.
- keep custom home content inside the standard public layout while preserving full-page iframe rendering for external URLs.
feitianbubu