- Added logic to reset `pinnedAuthID` and replay transcript on unauthorized, forbidden, or throttling errors.
- Enhanced error handling in `forwardResponsesWebsocket` with detailed status inspection.
- Introduced `shouldReleaseResponsesWebsocketPinnedAuth` to determine auth reset conditions.
- Updated state management to preserve prior request and response data during forced replay.
Fixed: #2230
- Deleted the `LoggerPlugin` along with associated usage tracking and in-memory statistics logic.
- Removed all related tests (`logger_plugin_test.go`, `usage_tab_test.go`) and external-facing handler (`usage.go`) for usage statistics export/import.
- Cleaned up TUI integration by deleting `usage_tab.go`.
- Introduced `Success` and `Failed` fields in auth records to track request outcomes.
- Updated `/v0/management/auth-files` and `/v0/management/api-key-usage` responses to include success and failure counts.
- Enhanced tests to validate tracking logic and API responses.
- Implemented `GetAPIKeyUsage` to expose recent request data grouped by provider and API key.
- Added supporting function `mergeRecentRequestBuckets` for bucket aggregation.
- Registered new endpoint `/v0/management/api-key-usage` in the management API.
- Included extensive unit tests for provider and key-based grouping validation.
- Updated `formatRecentRequestBucketLabel` to support configurable bucket duration.
- Implemented `RecentRequestsSnapshot` in `Auth` to capture bucketed recent request data.
- Added new fields and methods to `Auth` for tracking request success and failure counts over time.
- Updated `/v0/management/auth-files` response to include recent request data for each auth record.
- Introduced unit tests to validate request tracking and snapshot generation logic.
- Introduced reusable utilities in `requestmeta` to manage endpoint and response status in request contexts.
- Refactored plugins and handlers to use context-based metadata, removing direct dependency on `gin`.
- Updated tests to validate new context utilities and replaced `gin`-based context handling.
Fixed: #3166
- Introduced `DisableImageGenerationMode` with support for `false`, `true`, and `chat` values.
- Updated payload handling to preserve `image_generation` on images endpoints when `chat` mode is enabled.
- Modified OpenAI image handlers (`ImagesGenerations`, `ImagesEdits`) to respect tri-state logic.
- Added unit tests for `DisableImageGenerationMode` behavior and endpoint-specific handling.
- Enhanced configuration diff logging to support `DisableImageGenerationMode`.
- Added `disable-image-generation` configuration flag to disable the `image_generation` tool globally.
- Updated payload handling to remove `image_generation` tools from request payload arrays when the flag is enabled.
- Modified OpenAI image handlers (`ImagesGenerations`, `ImagesEdits`) to return 404 when the feature is disabled.
- Enhanced configuration diff logging to track changes for the `disable-image-generation` flag.
- Added accompanying unit tests for the new feature in payload helpers and image handler logic.
- Introduced tests to ensure unsupported models are rejected in `/images/generations` and `/images/edits`.
- Added `isSupportedImagesModel` and `rejectUnsupportedImagesModel` functions for consistent model validation.
- Enhanced image handler logic to apply validation checks for model compatibility.
- Introduced a `Disabled` flag to OpenAI compatibility configurations.
- Updated routing, auth selection, and API handling logic to respect the `Disabled` state.
- Extended relevant APIs, YAML configurations, and data structures to include the `Disabled` field.
- Adjusted all relevant loops and filters to skip disabled providers.
Closes: #3060#3059#2977
- Updated `buildImagesResponsesRequest` to derive `model` dynamically based on `toolJSON`.
- Adjusted streaming execution to handle dynamic model resolution across multiple contexts.
Closes: #2965
- Added `protocol_multiplexer.go`, enabling support for both HTTP and Redis protocols on a single listener.
- Introduced `redis_queue_protocol.go` to handle Redis-compatible RESP commands for queue management.
- Integrated `redisqueue` package, supporting in-memory queuing with expiration pruning.
- Updated server initialization to manage a shared listener and multiplex connections.
- Adjusted `Handler` to adopt `AuthenticateManagementKey` for modular key validation, supporting both HTTP and Redis flows.
Replace antigravityCreditsAvailableForModel with inline known/unknown
split. Auths whose credit hints are not yet populated are kept as
lower-priority candidates instead of being rejected, breaking the
chicken-and-egg deadlock at cold start.
- findAllAntigravityCreditsCandidateAuths now filters by PinnedAuthMetadataKey
to prevent credential isolation violations during credits fallback
- Release deferredBody reference on success path to avoid holding large
payloads in memory for the lifetime of the gin context
CountTokens upstream API does not support enabledCreditTypes, so
remove the dead credits fallback path from ExecuteCount and delete
the unused tryAntigravityCreditsExecuteCount method. Fix gofmt on
credits test file.
Move credits handling from executor-level retry to conductor-level
orchestration. When all free-tier auths are exhausted (429/503), the
conductor discovers auths with available Google One AI credits and
retries with enabledCreditTypes injected via context flag.
Key changes:
- Add AntigravityCreditsHint system for tracking per-auth credits state
- Conductor tries credits fallback after all auths fail (Execute/Stream/Count)
- Executor injects enabledCreditTypes only when conductor sets context flag
- Credits fallback respects provider scope (requires antigravity in providers)
- Add context cancellation check in credits fallback to avoid wasted requests
- Remove executor-level attemptCreditsFallback and preferCredits machinery
- Restructure 429 decision logic (parse details first, keyword fallback)
- Expand shouldAbort to cover INVALID_ARGUMENT/FAILED_PRECONDITION/500+UNKNOWN
- Support human-readable retry delay parsing (e.g. "1h43m56s")
- Added `GPT-Image-2` as a built-in model to avoid dependency on remote updates for Codex.
- Updated model tier functions (`CodexFree`, `CodexTeam`, etc.) to include built-in models via `WithCodexBuiltins`.
- Introduced new handlers for image generation and edit operations under `OpenAIAPIHandler`.
- Extended tests to validate 503 response for unsupported image model requests.
- Introduced `refreshIneffectiveBackoff` to prevent tight-looping in auto-refresh when token refresh fails to update expiry.
- Adjusted refresh logic to apply backoff when `shouldRefresh` evaluates true.
Closes: #2830
- Deleted `iflow` provider implementation, including thinking configuration (`apply.go`) and authentication modules.
- Removed iFlow-specific tests, executors, and helpers across SDK and internal components.
- Updated all references to exclude iFlow functionality.
- Refactored `requestExecutionMetadata` to handle empty `Idempotency-Key` gracefully.
- Added test to validate metadata inclusion of execution session without idempotency key.
When multiple auth credentials are configured, requests from the same
session are now routed to the same credential, improving upstream prompt
cache hit rates and maintaining context continuity.
Core components:
- SessionAffinitySelector: wraps RoundRobin/FillFirst selectors with
session-to-auth binding; automatic failover when bound auth is
unavailable, re-binding via the fallback selector for even distribution
- SessionCache: TTL-based in-memory cache with background cleanup
goroutine, supporting per-session and per-auth invalidation
- StoppableSelector interface: lifecycle hook for selectors holding
resources, called during Manager.StopAutoRefresh()
Session ID extraction priority (extractSessionIDs):
1. metadata.user_id with Claude Code session format (old
user_{hash}_session_{uuid} and new JSON {session_id} format)
2. X-Session-ID header (generic client support)
3. metadata.user_id (non-Claude format, used as-is)
4. conversation_id field
5. Stable FNV hash from system prompt + first user/assistant messages
(fallback for clients with no explicit session ID); returns both a
full hash (primaryID) and a short hash without assistant content
(fallbackID) to inherit bindings from the first turn
Multi-format message hash covers OpenAI messages, Claude system array,
Gemini contents/systemInstruction, and OpenAI Responses API input items
(including inline messages with role but no type field).
Configuration (config.yaml routing section):
- session-affinity: bool (default false)
- session-affinity-ttl: duration string (default "1h")
- claude-code-session-affinity: bool (deprecated, alias for above)
All three fields trigger selector rebuild on config hot reload.
Side effect: Idempotency-Key header is no longer auto-generated with a
random UUID when absent — only forwarded when explicitly provided by the
client, to avoid polluting session hash extraction.
- Deleted `QwenAuthenticator`, internal `qwen_auth`, and `qwen_executor` implementations.
- Removed all Qwen-related OAuth flows, token handling, and execution logic.
- Cleaned up dependencies and references to Qwen across the codebase.
- Introduced `auth-auto-refresh-workers` config option to override default concurrency.
- Updated `authAutoRefreshLoop` to support customizable worker counts.
- Enhanced token refresh scheduling flexibility by aligning worker pool with runtime configurations.
- Replaced `filepath.Walk` with `os.ReadDir` for cleaner directory traversal.
- Fixed `isAuthJSON` check to use `filepath.Dir` for directory comparison.
- Updated auth hash cache generation and file synthesis to improve readability and maintainability.
- Introduced `authAutoRefreshLoop` to handle token refresh scheduling.
- Replaced semaphore-based refresh logic in `Manager` with the new loop.
- Added unit tests to verify refresh schedule logic and edge cases.
After a Codex CLI compact, the client sends a full conversation
transcript (with compaction items or assistant messages) as input.
Previously, normalizeResponseSubsequentRequest() unconditionally
merged this with stale lastRequest/lastResponseOutput, breaking
function_call/function_call_output pairings and causing 400 errors
("No tool output found for function call").
Add inputContainsFullTranscript() heuristic that detects compaction
items (type=compaction/compaction_summary) or assistant messages in
the input array, and bypasses the merge when a full transcript is
present.
Fixes#2207
Luis Pater