Jaymr/new-api
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity
6,227 Commits 2 Branches 0 Tags
2a0ce3475c2df51ef5fd725f1eb0249822eb1c35
Commit Graph
10 Commits
Author SHA1 Message Date
Calcium-Ion 0ab0202060 Feat/auto group (#6590) 
* feat(token): support custom auto group order

* feat(keys): enhance auto group presentation

* fix(keys): rework Auto flow border and compact inherited order

The Auto group highlight previously tinted the whole control surface
with a gradient and animated only a 1px top sweep, which read as a
background color rather than a flowing border. Replace it with a
border-only effect: an aria-hidden, pointer-events-none overlay whose
conic gradient is masked down to a thin ring hugging the rounded
perimeter, so the highlight travels around all four edges and corners
every 3.2s. The interior stays neutral with a restrained static
primary border and glow; prefers-reduced-motion hides the moving
layer while keeping the static emphasis.

The inherited global Auto order also rendered as spacious two-line
rows with circular sequence markers, wasting drawer space. Render it
as a compact wrapping strip of one-line chips (index, name, ratio
badge) with descriptions kept accessible via title and sr-only text,
scrolling only past a much smaller max height.

Custom add/remove/reorder editing, empty-array inheritance semantics,
and the submit payload are unchanged.

* fix(keys): preserve Auto inheritance and unify effects

* refactor(keys): temporarily disable AutoGroupBadge in api-key-group-cell
 2026-08-01 23:19:01 +08:00
Calcium-Ion 86ac0f7745 refactor: extract protocol conversion layer into standalone relaykit module (#6369) 
* test(relayconvert): add golden snapshot matrix and relaykit boundary guard

Phase 0 of the relaykit extraction plan: pin byte-level output of every
registered (from,to) request/response/stream conversion route, and
forbid kit-bound packages from growing host-only imports.

* wip(relayconvert): drop gin.Context from converter signatures; add convmeta draft

Phase 1 in progress: relayconvert now takes context.Context; host media
resolver adapts gin.Context back at the service boundary.

* refactor(relayconvert): decouple converters from RelayInfo, gin, and settings

Phase 1 of the relaykit extraction plan:
- converters now depend on convmeta.Meta (implemented by RelayInfo) instead
  of *relaycommon.RelayInfo; ClaudeConvertInfo and the format guesser move
  to convmeta with aliases left behind
- host settings reach converters via a convmeta.Options snapshot built in
  RelayInfo.ConvOptions; no more model_setting/reasoning global reads inside
  the conversion layer
- effort-suffix helpers move to service/relayconvert/reasoning (old package
  forwards); chat-to-responses upgrade policy moves to service (host routing
  logic, not conversion)
- golden conversion matrix unchanged

* test(relayconvert): tighten boundary — kit packages now free of gin/setting imports

* refactor(dto): drop gin and logger dependencies

Phase 2 (part 1): dto.Request.IsStream now takes *http.Request instead of
*gin.Context (Gemini's impl reads query/path off the std request); dto's
three logger calls become common.SysError. Boundary test allowlist is now
empty — kit-bound packages import no gin/setting/logger/model.

* refactor(kit): extract dependency-free kitutil; dto/types/relayconvert stop importing common

Phase 2 of the relaykit extraction plan:
- new service/relayconvert/kitutil holds the pure helpers the kit needs
  (JSON wrappers, pointer/string/uuid/timestamp utils, MaskSensitiveInfo,
  pluggable LogInfo/LogError hooks, Debug flag)
- dto, types, and all relayconvert packages now use kitutil; their only
  remaining internal deps are dto/types/constant
- common keeps every original symbol (MaskSensitiveInfo delegates to
  kitutil) so host code is untouched; main.go routes kit logging into
  common.SysLog/SysError and mirrors DebugEnabled
- golden conversion matrix unchanged

* refactor(kit): move EndpointType/FinishReason to types; OpenRouter dialect via Options

Kit packages (dto/types/relayconvert/reasonmap) no longer import constant:
- EndpointType and finish-reason values live in types; constant re-exports
- the OpenRouter special-case in claude->openai request conversion reads
  Options.OpenRouterDialect, set by the host from the channel type;
  InitChannelMeta invalidates the cached snapshot on channel switch

* refactor: extract relaykit submodule (dto/types/relayconvert/reasonmap)

Phase 3 of the relaykit extraction plan:
- new go module github.com/QuantumNous/new-api/relaykit containing dto
  (minus task family), types, relayconvert (with convmeta/kitutil/reasoning),
  and reasonmap; host consumes it via require + replace, go.work for dev
- task-family dto (task/suno/midjourney/video) stays in the host dto
  package; dual-consumer host files alias it as taskdto
- relaykit builds and tests standalone (GOWORK=off): no host imports,
  no gin, no DB, no settings
- golden conversion matrix unchanged

* build(docker): copy relaykit/go.mod before go mod download

The local-replace submodule's go.mod must exist inside the build context
for the main module graph to resolve.

* fix: address relaykit extraction regressions

* fix: address relaykit review regressions

* docs: document Meta nil receiver contract

* fix(relaykit): fail OpenAI→Claude conversion without max_tokens; reject negative default_max_tokens

The Claude Messages API requires max_tokens (omitting it is a 400
"Field required"), but with a nil Options.Claude.DefaultMaxTokens hook
the converters silently emitted a request the upstream is guaranteed to
reject. Both OpenAI Chat and Responses → Claude conversions now return
sharedclaude.ErrMissingMaxTokens when no path (client value, default
hook, thinking-adapter floor) supplied one. Unreachable in the host,
which always configures the hook.

Host side, claude.default_max_tokens now rejects negative values at the
option API before persisting — they would wrap into huge unsigned values
during conversion. Zero stays allowed: the current API treats
max_tokens: 0 as cache pre-warming.

* fix: make Gemini safety settings read path race-free
 2026-07-27 15:56:21 +08:00
QuentinHsu 4aa08f917e fix(playground): resolve auto group model listing (#6163) 
* fix(playground): resolve auto group model listing

- merge and deduplicate available models in configured auto group order.
- reuse special usable group rules and add model filtering regression coverage.

* refactor: extract GetGroupsEnabledModels to dedupe group model expansion
 2026-07-20 18:21:24 +08:00
Calcium-Ion 31d70fca39 refactor(auth): replace dashboard sessions with stateless tokens and session control (#6329) 
* refactor(auth): replace dashboard sessions with stateless tokens

* feat(auth): harden session issuance and distributed enforcement

* fix(proxy): preserve trusted proxy compatibility defaults

* refactor: address dashboard auth review feedback

* refactor: remove classic frontend and flatten web app
 2026-07-20 16:48:43 +08:00
Calcium-Ion c36418c863 feat: enhance text protocol conversion and advanced custom routing (#5825) 
* 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
 2026-07-11 20:44:12 +08:00
CaIon 4a64b87072 test(user): cover self-service password update guard 
Pin down the checkUpdatePassword contract:

- changing a password requires the correct current password
- OAuth/passwordless accounts (empty password hash) cannot set a
  password through the self-service endpoint and must use the password
  reset flow instead
- setupLogin never writes back the password column

These guard against regressing the password-change path back into a
short-circuit that lets passwordless accounts set a password directly.
 2026-07-05 13:53:25 +08:00
QuentinHsu 966af88ec3 feat(playground): improve Playground chat experience and Markdown rendering (#5217) 
* refactor(playground): streamline chat request state

- extract conversation actions from the page component to keep message flow logic reusable.
- unify streaming and non-streaming generation state, including abort support for non-stream requests.
- simplify message rendering and payload construction while localizing Playground prompts.

* fix(playground): validate persisted chat state

- wrap saved Playground state with a storage version while still reading legacy values.

- validate config, parameter toggles, and messages before restoring them from localStorage.

- cap stored chat history to the latest messages to avoid oversized or stale state.

* refactor(playground): centralize message content access

- route chat rendering, copy actions, and error display through shared message helpers.

- reuse the current-version update helper for non-streaming assistant responses.

- keep message version details behind utility functions to reduce future model churn.

* refactor(playground): split storage schemas

- move Playground storage validation schemas into a dedicated module.

- keep storage read and write logic focused on migration, trimming, and persistence.

- preserve the existing storage envelope and validation behavior.

* refactor(playground): extract options loading hook

- move model and group queries into a dedicated hook so the page component stays focused on layout wiring.
- preserve existing fallback selection and error toast behavior while reusing the hook through the playground barrel export.

* refactor(playground): extract prompt suggestions

- move static prompt suggestion rendering into a focused component so the input stays centered on compose controls.
- preserve translated suggestion submission behavior while isolating icon metadata from the input form.

* refactor(playground): extract input tools

- move attachment and search controls into a dedicated component so the prompt input stays focused on compose state.
- keep existing development toast behavior and disabled handling while centralizing tool metadata.

* refactor(playground): extract input controls

- move model, group, send, and stop controls into a focused component so the input only manages compose state.
- preserve existing disabled states and generation button behavior while isolating control rendering.

* refactor(playground): extract message content display

- move sources, reasoning, loading, error, and response rendering into a dedicated message content component.
- keep the chat list focused on message iteration, edit state, and action wiring without changing display behavior.

* refactor(playground): extract message editor

- move inline message editing controls into a dedicated editor component so the chat list stays focused on rendering flow.
- preserve save, save-and-submit, cancel, and disabled-state behavior for edited messages.

* refactor(playground): extract stream error parsing

- move SSE error payload parsing into a reusable stream utility so the request hook stays focused on lifecycle handling.
- preserve existing error message, error code, and fallback behavior for raw or empty stream errors.

* refactor(playground): extract request error parsing

- move non-stream request error extraction into a shared utility so the chat handler stays focused on request flow.
- preserve the existing response message, error code, and fallback priority for failed chat completions.

* refactor(playground): extract streaming chunk updates

- move reasoning and content chunk application into a message utility so the chat handler only wires stream events.
- preserve error-state skipping, reasoning accumulation, and content streaming behavior for assistant messages.

* refactor(playground): extract message reasoning parser

- move think tag parsing into a dedicated playground message utility.
- export the parser through the shared playground lib barrel for consistent imports.

* refactor(playground): extract message streaming utilities

- move stream chunk application and message finalization into a dedicated utility.
- keep stored message sanitization with the streaming lifecycle helpers.

* refactor(playground): extract message update utilities

- move assistant message update helpers into a focused playground utility.
- keep error-state message updates separate from core message construction helpers.

* refactor(playground): extract completion choice handling

- move non-streaming choice application into the message streaming utilities.
- keep the chat handler focused on request orchestration and message updates.

* refactor(playground): centralize assistant completion state

- add a helper for finalizing assistant messages with complete status.
- reuse the helper in stream completion and stop-generation paths.

* refactor(playground): extract stream message parsing

- move SSE delta parsing into a shared stream utility.
- keep the stream request hook focused on lifecycle handling and update dispatch.

* refactor(playground): extract stream ready state checks

- move SSE ready-state status handling into stream utilities.
- keep weak source status typing outside the stream request hook.

* refactor(playground): extract conversation message helpers

- move send, regenerate, and edit message list construction into focused utilities.
- keep the conversation hook focused on edit state and update dispatch.

* refactor(playground): extract state initialization helpers

- move playground initial state loading into focused utility helpers.
- centralize message state updater resolution outside the React state hook.

* refactor(playground): extract option fallback helpers

- move model and group fallback selection into focused playground utilities.
- keep the options hook focused on query results, toasts, and config updates.

* refactor(playground): extract message action helpers

- move message action state derivation into focused utilities.

- keep the action component focused on guarded handlers and rendering.

* refactor(playground): extract input control state

- move submit, stop, and selector state derivation into a pure helper.

- keep input controls focused on rendering model selectors and action buttons.

* refactor(playground): extract message content state

- move source, reasoning, loader, and body visibility checks into a pure helper.

- use a discriminated state shape so rendered reasoning content stays type-safe.

* refactor(playground): extract message editor state

- move save eligibility and submit visibility checks into a pure helper.

- keep the editor component focused on textarea and button rendering.

* refactor(playground): extract message error state

- move error kind, fallback content, and admin visibility checks into a pure helper.

- centralize the model pricing settings path used by the error action.

* refactor(playground): extract chat render state

- move editing content lookup and per-message render flags into conversation helpers.

- keep the chat component focused on mapping messages to editor and content views.

* refactor(playground): extract suggestion display state

- move suggestion class selection into a pure helper.

- keep the suggestions component focused on translation and rendering.

* refactor(playground): extract assistant message state checks

- move final and pending assistant status checks into streaming utilities.

- keep the chat handler focused on request lifecycle updates.

* refactor(playground): extract input tool state

- move attachment action metadata and development notices into input tool utilities.

- keep the input tools component focused on menu and button rendering.

* refactor(playground): extract stream protocol checks

- move SSE done-message and closed-ready-state checks into stream utilities.

- keep the stream request hook focused on event handling flow.

* refactor(playground): extract message removal helper

- move delete-message filtering into conversation message utilities.

- keep the conversation hook focused on action orchestration.

* refactor(playground): extract option error messages

- move option load error message selection into playground option utilities
- keep the options hook focused on query effects and fallback updates

* refactor(playground): extract input submit text helper

- move prompt submit text validation into input control utilities
- let the input component submit only when a concrete text value is available

* refactor(playground): centralize error message checks

- add a shared helper for identifying error messages
- remove direct status string checks from message content rendering

* refactor(playground): extract message content display checks

- move loader and content visibility decisions into local helper functions
- keep message content state assembly focused on composing render state

* refactor(playground): replace raw message role checks

- use shared message role constants in conversation edit handling
- avoid raw assistant role literals when validating API messages

* refactor(playground): extract non-stream response handling

- move chat completion response choice handling into message streaming utilities
- keep the chat handler focused on request lifecycle and error routing

* refactor(playground): centralize stream cleanup

- reuse one stream cleanup path for completion, errors, startup failures, and manual stops
- preserve the current-source guard when closing SSE streams

* refactor(playground): extract pending assistant check

- centralize pending assistant message detection in streaming utilities
- reuse the helper when sanitizing stored playground messages

* perf(playground): improve mobile input controls

- split mobile input controls into selector and action rows
- keep the desktop input footer compact while reducing mobile control crowding

* perf(playground): add starter empty state

- show starter prompts in the empty playground chat area
- wire empty-state prompt selection into the existing send flow
- add localized copy for the new empty state

* perf(playground): improve mobile message actions

- collapse mobile message actions into a touch-friendly dropdown menu
- keep the desktop hover action strip unchanged for pointer workflows
- share one action list between desktop buttons and the mobile menu

* perf(playground): add error recovery actions

- show retry, edit, and delete actions inside error message alerts
- route edit recovery to the previous user prompt when available
- keep recovery controls touch-friendly on mobile layouts

* perf(playground): refine message editing experience

- present message edits in a focused bordered editor panel
- add unsaved-change state, reset, and cancel confirmation flows
- improve mobile touch targets and keyboard shortcuts for editing

* perf(playground): improve markdown code blocks

- render fenced markdown code with syntax highlighting, line numbers, and fallback plain text
- add copy, download, and collapse controls for playground AI responses
- tighten code block layout and theme token styles for responsive markdown rendering

* fix(playground): constrain markdown code block height

- collapse long playground code blocks after a short preview instead of waiting for very large snippets
- cap expanded code blocks so long responses scroll inside the code block
- keep generic code block usage unconstrained unless a caller opts in

* feat(playground): add chat history clearing

- add a toolbar action that is enabled only when saved playground messages exist.
- confirm destructive clears before removing browser-stored conversation state.
- add localized strings for the action, dialog, and completion toast.

* perf(playground): improve chat markdown rendering

- refine assistant and user message surfaces so chat content matches the app UI.
- normalize markdown typography, tables, images, lists, blockquotes, and details rendering.
- add indentation cues for collapsible reasoning and source sections.

* style: format code block component

* style: format playground frontend files

* feat(playground): render markdown with stream parser

- replace Streamdown with stream-markdown-parser for project-owned markdown rendering and styling.
- split response rendering into focused block, inline, table, alert, details, and footnote modules.
- pass message final state into response parsing so streaming content can be parsed incrementally.

* fix(playground): localize reasoning and chat feedback

- translate reasoning status, message actions, playground errors, and response renderer fallbacks across supported locales.
- keep reasoning duration numeric and tighten the collapsible layout to prevent trigger jitter.
- register dynamic keys so i18n sync keeps runtime labels covered.

* refactor(playground): group files by functional area

- move chat, input, and message components into focused subdirectories to make the UI structure easier to scan.
- split playground helpers into input, message, streaming, storage, options, state, and suggestions modules.
- update barrel exports and imports so existing feature entry points continue to work.

* fix(playground): prevent history replay from freezing page

- defer saved conversation loading so route entry no longer blocks on localStorage parsing and markdown rendering.
- limit initial history rendering and skip expensive markdown parsing for oversized responses.
- normalize corrupted streaming snapshots and cumulative chunks to keep saved playground history bounded.
- add message timing metadata and layout alignment groundwork without introducing live timers.

* feat(playground): allow regenerating from user messages

- show regenerate actions on user messages with saved content.
- truncate following conversation state before starting a fresh assistant response.

* feat(playground): add raw response source view

- add a per-message source toggle for assistant responses.
- render raw response content with the existing code block viewer.
- localize the new source and preview action labels.

* feat(playground): render code with unified editor

- replace Shiki HTML rendering with a read-only CodeMirror view for code blocks and raw responses.
- reuse the same CodeMirror frame for message editing so source and edit modes stay visually aligned.
- add lightweight CodeMirror dependencies while keeping language support scoped to Markdown.

* perf(playground): streamline chat input controls

- combine model and group selection into one compact picker for faster context switching.
- switch playground action buttons to icon-first controls with tooltips to reduce toolbar width.
- refresh input footer styling and submit states so active and destructive actions are clearer.
- bump dompurify lockfile entry to keep the frontend dependency current.

* fix(playground): filter models by selected group

- query user models by the selected playground group instead of reusing the cross-group model union.
- clear unavailable model selections and block sending when the active group has no models.
- align model selector and error action controls with the existing playground interaction style.

* perf(playground): remove input suggestion chips

- remove the prompt suggestion row below the playground input to reduce visual noise.
- delete the now-unused suggestion component and display helper.

* perf(playground): stabilize reasoning trigger layout

- use fixed icon slots around the reasoning label so the left content stays still when toggling.
- limit the open state animation to the chevron rotation for a smoother collapse interaction.

* perf(playground): smooth reasoning expansion

- use the collapsible panel height animation for vertical reasoning reveals.
- sync inner content opacity and position with the panel state.
 2026-06-27 17:23:25 +08:00
Calcium-Ion 6dc4030fdf feat: support ClickHouse log database (#5663) 
* feat: support ClickHouse log database

* feat(log): optimize log deletion process for ClickHouse
 2026-06-22 18:41:26 +08:00
Seefs 8d87d5fd8b test: clean up reward-hacking backend tests (#5563) 2026-06-17 16:47:52 +08:00
yyhhyyyyyy e3d64cb76d Merge pull request #4431 from yyhhyyyyyy/fix/tiered-billing-model-list 
fix: include tiered billing models in model listing
 2026-04-24 16:24:36 +08:00