Skip to content

TokenHub v0.10.0

Choose a tag to compare

@astaxie astaxie released this 20 Oct 05:34
· 135 commits to main since this release

TokenHub v0.10.0

TokenHub v0.10.0 brings image, audio, and video workflows into the enterprise AI gateway, adds multiple upstream keys per Provider, simplifies daily administration, and improves usage reporting, billing, and resource control.

This release covers all 172 commits and 20 merged pull requests since v0.9.0, including contributor updates and release-version synchronization. The complete PR and commit inventories are included below.

Highlights

  • One gateway for text and media: published models can use image generation/editing/variations, speech synthesis, transcription, translation, and video workflows with shared authentication, model access, quotas, policies, and accounting.
  • Durable asynchronous video and image tasks: submit through /v1/videos/generations or supported Images routes and poll /v1/tasks/{task_id} with opaque, owner-bound task IDs that survive restarts and stay pinned to the original provider account.
  • Multiple API keys per Provider: rotate and manage upstream keys with independent priority, weight, RPM, TPM, and concurrency limits while retaining resource routing and failover.
  • Simpler operations: searchable provider onboarding, compact model and route management, background provider monitoring, and Today / 7 days / 30 days / All usage reports.
  • Practical cost and quota controls: independent default allowances for every user, configurable daily/monthly quota timezones, audio-second billing, and statement reporting in USD, CNY, JPY, or RUB using manual rates.
  • Bounded resource use: SQL-based usage aggregation, separate SQLite critical/analytics readers, smaller audit snapshots, and configurable application memory budgets.

Detailed changes

Media APIs, transcription, and retrieval

  • Extend ordinary published upstream models to image generation, JSON/multipart edits, variations, speech, transcription, and translation. Preserve supported masks, repeated multipart fields, vendor options, multiple results, binary audio, subtitles, and large numeric identifiers. Native Gemini media retains generation configuration, inline content, thought signatures, and supported SSE payloads. (#361, #390)
  • Support documented provider-specific video, music, lyrics, voice upload/clone, and drawing workflows through existing Chat/Responses surfaces. Media generation and task queries bypass response caches; background Responses retain their local job ID and expose validated vendor response IDs separately. (#361)
  • Add POST /v1/videos/generations and GET /v1/tasks/{task_id}. Recognized Images/Videos receipts receive opaque gateway handles backed by encrypted, persistent bindings. Polling validates owner, model access, and the original resource/account; supported source/draft/extension references are resolved under the same restrictions. Polling consumes request quota without charging generation tokens again. (#392)
  • Apply text policies, request/provider/response/stream hooks, guarded outbound transport, authorization, quotas, and usage attribution across media workflows. Validate JSON, SSE completion/error envelopes, binary output, MIME types, Base64, and size limits. Ambiguous generation failures do not cause automatic duplicate submissions. Preserve valid usage after delivery or hook failures and redact credentials without corrupting task/media fields. (#361, #390, #392)
  • Complete native multipart transcription capability and endpoint checks. Validate file signatures and upload limits after hooks and safety masking, accept valid MP4/M4A padding, and keep uploaded media and transcript text out of direct-media audit/trace payloads. (#390)
  • Add fractional audio-second metering with independent public-model and provider rates via metadata.audio_price_usd_per_second. Retain duration and independent token/cache/reasoning details through streams, quota settlement, durable evidence, statements, and failed delivery. Missing costs remain unknown; duration is not converted into invented token charges. (#390)
  • Extend the existing rerank API to string documents, {id,text} documents, and mixed inputs. Preserve IDs/indexes through hooks, cache, routing, and optional document returns; enforce bounds and retain usage through cancellation and partial failure. Add transcription/rerank SDK smoke clients and fake-provider regression coverage. (#390)
  • Restore previously published local embedding/rerank calls rejected solely for unconfirmed supplier costs. Recognize custom catalogs for default Jina-compatible reranking, preserve explicit protocols and same-type provider metadata, and add actionable route errors, endpoint previews, vector-space guidance, and explicit free-price hints. (#365)

Providers, models, and routing

  • Add an upstream API-key pool editor with rotation, enable/disable, deletion, priority/weight, and RPM/TPM/concurrency controls. Re-enabling clears cooldown/failure state. Preserve resource groups, route bindings, legacy priority, saved secrets, and recoverable account drafts during edits. New connections receive distinct identities; duplicate explicit create IDs return 409 provider_exists. (#394)
  • Redesign provider setup around searchable service cards grouped by subscriptions/accounts, providers, aggregators, and local deployments. Combine connection/model selection, add consistent icons, preserve discovery selections and failed drafts, and rename Provider Channels to Providers. (#366)
  • Simplify the model directory and replace the route configuration wall with per-model summaries and an on-demand editor. Retain all seven existing strategies, resource/project bindings, ordering, and plugin fields. Improve keyboard access, mobile layouts, overflow menus, row actions, and unsaved-draft handling. (#366)
  • Move monitoring checks into the background and refresh saved health/account snapshots every five minutes. Separate enabled status from observed health, keep unobserved providers neutral, and preserve health when checks are canceled. (#366)
  • Separate Jev protocol compatibility, model capability declarations, and request-dependent admission. Validate reasoning, sampling/tool combinations, denied fields, endpoints, and continuations before selection. Preserve valid shared request controls and record safe administrator reason codes without exposing candidate inventory. (#384)
  • Complete GPT-6.1 Sol / GPT-6 Sol / GPT-6 Luna Chat and Responses capability declarations. Update the Codex subscription client fingerprint to 0.159.3 for current model discovery. (#382, #373)
  • Refresh 72 standard model templates and 97 provider-specific offers across 27 catalog entries; add Voyage AI catalog integration and lifecycle/reference-price evidence. Guard new active routes against retired or unsupported offers and require confirmation for conditional or unverified costs. Preserve existing routes/prices and honor confirmed zero provider costs. (#383)

Usage, billing, quotas, and notifications

  • Add Today, 7 days, 30 days, and All filters to Global Usage with permission-scoped metrics, trends, attribution, and timezone-aware hourly/daily/monthly buckets. Include failed requests in request counts, refresh open reports, distinguish loading/empty/error states, and show visible project names instead of raw IDs. (#366, #374)
  • Add quota_timezone with UTC default and IANA timezone support for daily/monthly API-key and aggregate-user quotas. Preserve admission-period attribution for settlement, refunds, and recovery when timezone settings change. Minute limits and billing/budget periods keep their existing boundaries. (#358)
  • Add the opt-in scope: user, scope_id: all_users policy template. Each current or future attributed user receives an independent allowance across their keys/projects; individual policies combine with the strictest positive limits. Only platform administrators can manage the template. (#375)
  • Add USD/CNY/JPY/RUB reporting to billing statements using explicit manual exchange rates, locale-aware decimal input, exact arithmetic, half-even rounding, conversion evidence in CSV exports, and per-currency browser preferences. Preserve original amounts and unknown costs. This reporting conversion does not change historical charges, USD accounting, pricing cards, or budgets. (#391)
  • Add delivery tests for saved email, webhook, DingTalk, Feishu, WeCom, Slack, Discord, Telegram, and WhatsApp channels, including disabled channels. Verify destination acceptance, record redacted results without creating an operational alert, and fix encrypted/legacy credential handling and SMTP secret redaction. (#360)

Security, reliability, and maintenance

  • Fix the API-key creation wizard so selected models persist as restricted access. Extend existing model-access editing regression coverage and documentation for deny-all/inheritance, permissions, quotas, and failed-save drafts. (#359, #370)
  • Preserve OIDC/OAuth callback failures on the login page with translated, allowlisted failure-stage codes while retaining PKCE, one-time codes, browser/tab binding, and callback cleanup. (#381)
  • Harden credential redaction across Gemini headers, provider/plugin failures, media task results, invalid-hook diagnostics, SMTP errors, logs, traces, and audits while retaining valid accounting evidence. (#360, #361, #390, #392)
  • Aggregate daily usage, breakdowns, and UTC trends in SQL under scoped windows and deadlines. Add API-key/time and attributed-user/time indexes; use a shared breakdown snapshot while retaining the requested all-history scope for summary/breakdown views. Bound file-backed SQLite to one writer, two critical WAL readers, and one analytics reader with isolated caches, while preserving private/shared in-memory and memdb behavior. Add protected pool and database/WAL/SHM diagnostics. (#395)
  • Cap redacted request/response audit snapshots at 8 KiB and omit oversized/complex snapshots early. Add configurable 4 GiB bundled-container hard limits without swap and a 2 GiB Go soft target; start.sh also sets the Go target. Document resource diagnostics and offline WAL maintenance. (#395)
  • Move reconciliation persistence and administrative HTTP adapters into dedicated modules while preserving API, billing, attribution, and cost/audit evidence. Add PostgreSQL coverage and architecture/refactoring guidance. (#296)
  • Synchronize affected English, Simplified Chinese, and Japanese documentation, localized UI coverage, and contributor lists; extend backend integration-test time allowances. (#358 and contributor updates)

Upgrade notes

  1. Database: back up the database and server secret. Startup applies migration v8 (media task bindings) and v9 (usage indexes); large histories need time and disk/I/O headroom for index creation. The new manifest targets schema 9 and accepts states 8–9. v0.9.0 declares schema 7 only, so do not assume a binary-only downgrade after migration; follow the documented compatibility checks and backup recovery procedure.
  2. Persistent tasks: keep TOKENHUB_SECRET_KEY and the database stable. Use the submitting project key and returned TokenHub task handle. Bindings expire after 30 days, and changing pinned credentials/connection settings can invalidate old handles. Bound media tasks, managed image jobs, and background Responses use distinct polling mechanisms. Provider-specific Chat/Responses auxiliary task IDs remain provider-managed and do not receive this ownership binding.
  3. Jev inventory: empty parameter declarations no longer accept arbitrary optional model parameters. Review existing declarations; imported records are not silently rewritten. Preserve configured prices and unrelated metadata when updating inventory.
  4. Existing key scopes: the wizard fix changes new saves, not previously stored keys. Review old wizard-created keys where selected-model restriction was intended and explicitly save the correct model access.
  5. Quotas: UTC remains the default. Upgrade all backends before changing quota timezone; historical counters are not redistributed. Older versions do not enforce all_users templates. Any rollback must first satisfy database compatibility and account for timezone/policy differences.
  6. Memory: size TOKENHUB_CONTAINER_MEMORY_LIMIT=4g and GOMEMLIMIT=2GiB for the host/workload. The Go target is not an RSS cap; export it before direct/native startup rather than placing it only in backend/.env. Remote-PostgreSQL multi-instance and Helm deployments need their own resource budgets.
  7. Media and pricing: endpoint support requires an accessible upstream, published model, compatible route, and key permission. Direct media defaults to a 32 MiB request bound and buffers responses up to 128 MiB; transcription files have a separate 25 MiB limit. Buffered SSE is not low-latency speech delivery. Multipart transcription requires a native adapter with explicit transcription capability; generic media capability or provider-call hooks alone are insufficient. Configure public/provider audio-second rates separately; unsupported units and missing prices remain explicit.
  8. Provider API: use PATCH to update existing Providers; duplicate explicit POST IDs now return 409. Eligible key-pool resources replace the default connection key unless it is explicitly added as an inheriting resource. Disabling preserves bindings; deleting a resource clears its route bindings.
  9. Reporting currencies: exchange rates are manually supplied browser preferences, not a live FX feed or a change to settlement currency. Historical evidence remains unchanged.

See the multimodal API guide, media operations guide, deployment guide, and database evolution guide.

Validation

  • Release commit: aae5ca4a04a945616e5864f89236366844ea0fff.

  • CI passed on this exact commit: backend formatting/vet/golangci-lint/tests and migration manifest, Plugin DevKit, PostgreSQL integration, database N-1 compatibility, frontend lint/typecheck/tests/build, browser smoke, deployment/Compose/Helm validation, and repository gates.

  • Migration CI passed: migration unit tests and LiteLLM migration end-to-end tests.

  • Local checks passed: version/upgrade/rollback regressions, 183 repository tests, translation/environment/source-size gates, Compose rendering, release-tag tests, and git diff --check.

  • Native release publication passed. Both downloaded archives match checksums.txt, contain VERSION=0.10.0 and the required bundled files, and have matching amd64/arm64 backend, Node.js, and Sharp binaries.

  • Container publication passed. Registry inspection confirms 0.10.0, 0.10, and latest all resolve to sha256:22fd47cffe49a99d431c03746bb6b08792ce4dce2a91a40b7711c93f3482060a, with both linux/amd64 and linux/arm64 images.

Live, credential-dependent vendor generation was not exercised for this release; media regression coverage uses synthetic inputs and local fake upstreams.

Release artifacts

  • tokenhub_0.10.0_linux_amd64.tar.gz
  • tokenhub_0.10.0_linux_arm64.tar.gz
  • checksums.txt with SHA-256 checksums
  • Combined backend/console container: ghcr.io/astaxie/tokenhub-backend:0.10.0, with :0.10 and :latest aliases for linux/amd64 and linux/arm64

Complete pull request inventory

  • #358 fix(quota): respect configurable daily and monthly timezones — @astaxie
  • #359 fix(api-keys): persist wizard model allowlist mode — @astaxie
  • #360 feat(alerts): add notification channel test delivery — @astaxie
  • #361 feat(gateway): support image, audio, and video model APIs — @astaxie
  • #365 fix(retrieval): restore local calls and clarify provider setup — @deepjerry-ai
  • #373 fix(codex): claim client version 0.159.3 for model discovery — @stoner-byte
  • #374 fix(usage): display project names in attribution tables — @ACwisdom
  • #370 test(api-keys): cover existing model access editing — @ACwisdom
  • #366 feat(admin): simplify provider, model, and route management — @astaxie
  • #383 feat(catalog): refresh models and guard retired upstream offers — @deepjerry-ai
  • #382 fix(models): complete gpt-6 request capability declarations — @deepjerry-ai
  • #375 feat(quota): add default per-user quota policies — @ACwisdom
  • #381 fix(auth): preserve OAuth callback errors on the login page — @astaxie
  • #296 refactor(reconciliation): move persistence and admin adapters — @bailu-ZZ
  • #384 feat(routing): separate Jev protocol and model compatibility — @deepjerry-ai
  • #391 feat(billing): add statement reporting currencies — @astaxie
  • #390 feat(gateway): complete transcription and rerank support — @astaxie
  • #392 feat(gateway): add async video APIs and multimodal documentation — @astaxie
  • #394 feat(providers): manage multiple upstream keys per provider — @astaxie
  • #395 fix(sqlite): bound usage queries and isolate critical reads — @astaxie
Complete commit inventory (172 commits since v0.9.0)
  • aae5ca4a0 chore(release): prepare v0.10.0
  • 779bc46e5 Merge pull request #395 from astaxie/codex/sqlite-resource-bounds-393-378
  • 6e5d9c331 fix(sqlite): isolate wal readers from shared cache locks
  • f02a8ba7f fix(sqlite): preserve memdb vfs connection semantics
  • ea5e427eb fix(sqlite): preserve private memory stores in analytics
  • 7653607b5 fix(sqlite): decode file uri paths in storage diagnostics
  • 228be431c fix(db): align compatibility manifest with usage indexes
  • 1c0afd1c9 fix(sqlite): bound usage queries and isolate critical reads
  • dfb45438b Merge pull request #394 from astaxie/codex/provider-multiple-keys-377
  • 11873cbc8 fix(providers): preserve key scope and reuse account drafts
  • 37247422c fix(providers): preserve legacy key priority during edits
  • 75cbbf397 fix(providers): validate pool priority and restore key health
  • 794721ec0 feat(providers): manage multiple upstream keys per provider
  • 6ee041490 Merge pull request #392 from astaxie/codex/apimart-multimodal-api
  • 56053c8a4 fix(media): preserve task protocol during credential redaction
  • f8f217378 test(media): close fixture database pools
  • 2db2b049c docs(media): align localized video task contracts
  • 359838e4b docs(media): remove provider branding from guides
  • c736d3faa fix(gateway): bind versioned media task receipts
  • 703c1665a Merge remote-tracking branch 'origin/main' into codex/apimart-multimodal-api
  • ba35d1a92 fix(gateway): resolve video task references safely
  • 52aafeb29 Merge pull request #390 from astaxie/codex/complete-audio-rerank
  • 614bb93b4 fix(gateway): keep task cleanup aligned with batch limits
  • b85f194bc fix(gateway): preserve usage on invalid media outputs
  • 2cdd8ac56 feat(gateway): add async video APIs and multimodal documentation
  • f3c433c12 fix(gateway): retain cache-only usage evidence
  • 8def03e7e fix(gateway): harden audio and rerank usage failures
  • 3afdfed66 fix(rerank): preserve usage across hook failures
  • cd58b07df fix(audio): preserve duration billing and enforce period quotas
  • 3e236de2b Merge remote-tracking branch 'origin/main' into codex/complete-audio-rerank
  • db48c83a1 fix(audio): retain independent and durable token usage
  • 81cfc0fd1 Merge pull request #391 from astaxie/codex/multi-currency-billing-386
  • 41951f582 fix(audio): retain token details and bound complete uploads
  • 471f68c05 fix(billing): preserve rate edits and saved currency preferences
  • 21f270885 fix(gateway): enforce audio endpoints and retain safe usage
  • 9784656ac fix(billing): accept locale decimal separators in manual rates
  • b72ab52eb fix(audio): enable publication and preserve native upload contracts
  • 69952d864 fix(billing): wrap long converted amounts on mobile
  • 725fa3b70 feat(billing): add statement reporting currencies
  • 5edfab835 fix(audio): preserve metering privacy and local rejection health
  • f9fe2e7d4 fix(rerank): preserve neutral health on client cancellation
  • 6418286d9 fix(gateway): enforce request limits after native masking
  • 094993764 fix(audio): enforce transformed upload limits and retain evidence
  • 32450a0fa fix(gateway): close review gaps in hooks and audio statements
  • 1e975f5a0 fix(gateway): preserve usage and harden plugin boundaries
  • 7195a86f6 feat(gateway): complete transcription and rerank support
  • f5926f394 docs(contributors): refresh contributor lists
  • 51d2529fc Merge pull request #384 from astaxie/codex/jev-request-compatibility
  • 9a7dc7f7f merge: sync Jev compatibility with main
  • f017e563b Merge pull request #296 from bailu-ZZ/refactor/169-reconciliation-adapters
  • c0e5d2025 docs(contributors): refresh contributor lists
  • bc6171957 Merge pull request #381 from astaxie/codex/fix-oidc-login-379
  • c30fae2b9 Merge pull request #375 from ACwisdom/codex/default-user-quota
  • df844e491 docs: translate backend refactoring plan
  • 101f912ff docs(architecture): document modular boundaries and migration plan
  • 3d66fdbff Merge remote-tracking branch 'upstream/main' into refactor/169-reconciliation-adapters
  • 17ddd801a test(quota): cover team leader user-target permissions
  • 7056a9336 Merge branch 'main' of https://github.com/astaxie/TokenHub into codex/default-user-quota
  • 5923d6935 Merge pull request #382 from astaxie/codex/update-openai-gpt6-models
  • 3956f9f08 fix(routing): audit bound rejections and honor disabled tools
  • 06bf14e67 merge: preserve catalog policy and Jev contract updates
  • e864944be fix(models): admit shared OpenAI request controls
  • 90cf715ff merge: preserve catalog refresh and gpt-6 admission fixes
  • eceb62293 Merge pull request #383 from astaxie/codex/model-catalog-refresh-20261010
  • ef4b188b8 fix(routing): normalize Jev metadata lists
  • aa5cd357c feat(routing): separate Jev protocol and model compatibility
  • a9e0a75c8 fix(models): allow gpt-6 responses output controls
  • 95f94413e fix(models): complete gpt-6 capability parameter declarations
  • 44073a24d fix(catalog): scope policy updates and reasoning admission
  • 0f49b7ef6 fix(catalog): scope lifecycle and preserve deepseek temperature
  • 0c50bfdcc fix(billing): honor confirmed zero provider costs
  • b756ce953 fix(catalog): require confirmation for conditional reference prices
  • c5ed2e1da feat(catalog): refresh models and guard retired upstream offers
  • e74e62345 fix(models): declare top_p support for gpt-6 sol and luna
  • 144862d8c feat(models): add openai gpt-6 sol and luna catalogs
  • c00c4233e fix(auth): preserve OAuth callback errors on the login page
  • a23084074 Merge pull request #366 from astaxie/codex/simplify-provider-model-routing
  • 9a2aca623 docs(contributors): refresh contributor lists
  • 62c4048f2 fix(models): show row actions when space permits
  • 9b37a8059 fix(providers): keep overflow actions visible at viewport edges
  • 1795e30b9 Merge origin/main and preserve range-based usage regressions
  • 71d2de136 fix(providers): show actions inline when space permits
  • 61188f7a6 Merge pull request #370 from ACwisdom/repo-assist/test-key-model-access
  • e1943f464 docs(contributors): refresh contributor lists
  • 0a0219a21 Merge pull request #374 from ACwisdom/codex/fix-usage-project-attribution
  • e3c18de6c fix(usage): propagate report cursor close errors
  • 59bbaf6c5 feat(usage): add time range usage dashboard
  • bdd63ed65 feat(admin): categorize provider selection and harden connections
  • 1d652216b fix: preserve provider state when monitoring is canceled
  • a6aa14106 feat(quota): add default per-user quota policies
  • ab3c4ef0b fix(admin): keep provider health unknown without snapshots
  • 2a8ecdec9 test(admin): update localized route controls
  • 79079cae4 fix(admin): isolate route editor errors and deep links
  • 7f0ca930f test(admin): cover provider monitoring scheduler revisions
  • bf6255d2c fix(admin): close billing popover and refresh route labels
  • a2487b37a test(admin): cover unobserved degraded providers
  • 2eb41edc8 fix(admin): keep unobserved provider health neutral
  • fc91a8dde fix(admin): move provider monitoring checks to background
  • a49ef0dcf test(admin): follow provider navigation rename
  • 98efbe3e4 fix(admin): rename provider navigation label
  • 748c77431 fix(admin): style provider icon fallback
  • 1c64b3730 fix(admin): keep llama provider branding distinct
  • 1cbf1976c fix(admin): use provider brand icons in onboarding
  • 47d00eb9c fix(admin): prevent provider menu clipping
  • 61d615596 fix(admin): keep provider actions in a floating menu
  • 9a86ad0a5 fix(usage): display project names in attribution tables
  • eece1a461 fix(admin): preserve manual discovery selections
  • a129c0cd3 fix(admin): preserve drafts across async management flows
  • c398415d3 fix(admin): preserve drafts and explicit routing choices
  • a2b87e2a4 fix(admin): address follow-up management review findings
  • 0c6b337ce test(admin): correct route label accessibility queries
  • 3ada06061 Merge origin/main into provider management review fixes
  • cd1da4530 fix(admin): address provider and routing review feedback
  • b73632010 Merge pull request #373 from keepcrazy/fix/codex-client-version-0159
  • 1056de9f5 Merge origin/main into simplify provider model routing
  • 4c559070b fix(codex): claim client version 0.159.3 for model discovery
  • 8750058f5 test(api-keys): cover existing model access editing
  • 6a982c0b2 Merge pull request #365 from astaxie/codex/fix-retrieval-upgrade
  • da53cab6c feat(admin): simplify provider, model, and route management
  • 48d9396f9 fix(admin): guide retrieval setup and preserve provider protocols
  • 4c0ca7ce5 fix(billing): keep unpriced native retrieval costs unknown
  • a68f9feaa fix(rerank): recognize custom catalogs in automatic selection
  • a2326c611 Merge remote-tracking branch 'origin/main' into codex/fix-retrieval-upgrade
  • 02d652de4 fix(retrieval): restore published local model calls after upgrade
  • 62ef5d4e1 fix(embeddings): explain rejected route configuration
  • cbe50e64c Merge pull request #361 from astaxie/codex/media-api-support
  • dd50a8787 fix(gateway): retain Seedream output usage and costs
  • 9067d18c1 fix(plugins): redact Gemini client keys from hook headers
  • f44da518b fix(gateway): redact failed media status events
  • 1d41fc5c6 fix(gateway): bound media hook base64 by decoded bytes
  • 151e341d7 test(frontend): await plugin detail redirect effects
  • db8c67380 fix(gateway): retain media hook usage and complete vendor schemas
  • 2c545c495 fix(gateway): reject Gemini stream errors and allow Qwen outputs
  • f103ccfee fix(gateway): complete Gemini media and validate MIME base types
  • a6268b6ac fix(gateway): recognize structured JSON media responses
  • 3288585dc fix(gateway): reject media errors and retain overflow usage
  • 4ba891e5d test(gateway): use JSON in media hook scope fixtures
  • 6ce3dacca fix(gateway): validate direct media streams and preserve large audio
  • 985299421 fix(gateway): validate media events and align public schemas
  • a34d29a6e fix(gateway): preserve media failure accounting and vendor formats
  • 38ee4681b fix(gateway): enforce media hook protocols and JSON output
  • 6bc7ed3d3 Merge remote-tracking branch 'origin/main' into codex/media-api-support
  • 909e8be34 fix(gateway): validate media JSON and preserve attributed usage
  • fb50d54b7 Merge pull request #360 from astaxie/codex/test-notification-channels
  • 9dfc540f0 fix(gateway): validate media binary output and nested Wan prompts
  • 578c47048 fix(gateway): honor speech streams and multishot text policies
  • 218bdacde fix(alerts): redact Go-quoted SMTP credentials
  • 4d0878879 fix(gateway): close media publication and response handling gaps
  • f4993a1b2 fix(alerts): redact password reset SMTP errors
  • 08d35ff07 fix(gateway): inspect repeated media instructions
  • 74ce3d7f7 fix(alerts): preserve literal notification credentials
  • 15444168b fix(gateway): handle vendor media stream completion safely
  • b0709ea25 fix(gateway): preserve media stream policies and task ids
  • de644e332 fix(alerts): reveal encrypted notification credentials
  • a78b0fdc5 fix(gateway): enforce media policies and safe retries
  • 46ac4a1d1 fix(alerts): verify notification destination acceptance
  • 031d009cd feat(gateway): support compatible media model APIs
  • 251ea0f03 fix(alerts): reject failed notification test responses
  • bb77e8bb0 feat(alerts): add notification channel delivery tests
  • 736f0471c Merge pull request #359 from astaxie/codex/fix-api-key-model-allowlist
  • 40ad7a9ab test(ui): cover restricted key wizard scope
  • 605c66556 fix(api-keys): persist wizard model allowlist mode
  • f2d3781f2 Merge pull request #358 from astaxie/codex/fix-quota-timezone-335
  • 630684f59 ci: allow sufficient time for backend integration tests
  • a4a419fb3 docs(quota): align user guides with configurable quota periods
  • 2acccf8d5 fix(quota): respect configurable daily and monthly timezones
  • 08db06008 docs(contributors): refresh contributor lists
  • 768de8690 Merge remote-tracking branch 'upstream/main' into refactor/169-reconciliation-adapters
  • a4fdda576 fix(reconciliation): preserve bounded evidence compatibility
  • 2a3225c63 fix(reconciliation): restore attribution and cost evidence
  • 6291731f6 fix(server): remove unused codex subscription field
  • da1ab929e refactor(reconciliation): move persistence and admin adapters

Full Changelog: v0.9.0...v0.10.0