What's Changed

• fix(openai): tolerate object-shaped usage cost by @looplj in #2373
• Pass the caller's context through the v0.3.0 and v0.4.0 migrations by @omlahore in #2348
• fix(openai): preserve request contracts across transports by @techblack in #2374
• fix(openai): prevent zero-status errors from returning HTTP 200 by @Solaris-star in #2372
• fix(images): accept application/json in passthrough mode for /images/edits (#2336) by @remote-controlled-man in #2370
• fix(anthropic): preserve client cache_control breakpoints to keep prompt cache hits by @Azhc in #2342
• fix(trace): stop persisting traces for embedding requests by @trim21 in #2378
• fix(frontend): silence transient channel polling errors by @llc1123 in #2382
• fix(frontend): handle empty channel error messages by @llc1123 in #2375
• fix(api): report server stream deadlines to SSE clients by @llc1123 in #2362
• feat: add ZenMux channels and quota-aware UI by @llc1123 in #2381
• fix: window deploy scripts, close #2376 by @looplj in #2386
• fix(openai): complete cross-format request conversion by @techblack in #2377
• fix(frontend): localize model developer selector by @llc1123 in #2385
• docs: move deployment to axonhub-skills by @looplj in #2388
• fix: include hy4 models in Tencent catalog by @imzrme in #2387
• fix(openai): merge multiple system messages for strict upstreams by @itachaaa in #2384
• feat: archive personal api keys when delete user, close #2272 by @looplj in #2391
• feat: fire webhook when api key disabled, close #2380 by @looplj in #2393
• feat: remove cc billing system message for non official channel, clos… by @looplj in #2394
• feat: add Command Code channel support by @kaiwishc in #2392
• fix: propagate OpenCode sessions to custom endpoints by @imzrme in #2389
• feat: show OAuth channel quotas by @techblack in #2366
• fix(zai): support GLM-5.2+ reasoning_effort and GLM-5.3 always-on thinking by @MrLiuGangQiang in #2314
• feat(codex): support gpt-6-astra by @MrLiuGangQiang in #2399
• fix: render image-generation traces without GraphQL errors by @mvanhorn in #2297
• ci: enable manual unstable Docker image publishing by @llc1123 in #2403
• chore(deps): bump google.golang.org/grpc from 1.82.1 to 1.83.1 in /integration_test/gemini by @dependabot[bot] in #2404
• fix(requests): keep API key filter visible without results by @MrLiuGangQiang in #2405
• chore: sync model developers data by @github-actions[bot] in #2412
• feat: add ZenMux native video generation by @llc1123 in #2410
• fix(frontend): preserve channel form focus and onboarding dismissal by @llc1123 in #2411

New Contributors

• @omlahore made their first contribution in #2348
• @Solaris-star made their first contribution in #2372
• @remote-controlled-man made their first contribution in #2370
• @Azhc made their first contribution in #2342
• @imzrme made their first contribution in #2387
• @itachaaa made their first contribution in #2384
• @kaiwishc made their first contribution in #2392

Full Changelog: v1.0.0-beta9...v1.0.0-beta10

What's Changed

• feat: unify reasoning effort across protocols and add model endpoint routing by @AkaChou in #2330
• fix: dialog double scrollbar introduced in 2bfea19 by @BLumia in #2355
• feat: show codex reset details, close #2301 by @looplj in #2360
• fix(frontend): repair height-constrained scroll layout in mobile dialogs by @djdembeck in #2354
• fix(codex): strip user field from outbound requests by @llc1123 in #2352
• fix: preserve transformed Responses stream events by @MrLiuGangQiang in #2365
• fix: dockerfile compitable, close #2367 by @looplj in #2368
• fix(anthropic): reject invalid tool input on clean EOF by @llc1123 in #2363
• feat: add opencode session header, close #2361 by @looplj in #2369

Full Changelog: v1.0.0-beta8...v1.0.0-beta9

What's Changed

• opt: anthropic signature recognization by @looplj in #2197
• feat: unify auto-refresh controls and stabilize list animations by @AkaChou in #2198
• fix: correct currency code example from RMB to CNY by @hanuxxx in #2202
• feat(quota): add Charm Hyper credit balance checker by @djdembeck in #2199
• fix(quota): color Codex quota bars by usage by @djdembeck in #2166
• feat(quota): track OpenCode Go quota via official usage API by @jveko in #2204
• feat: hide unroutable configured models from public lists by @CberYellowstone in #2215
• fix(trace): only persist explicitly identified traces by @MrLiuGangQiang in #2208
• fix(responses): harden stream terminal, retry, and resource boundaries by @wtj-0527 in #2196
• feat: add xai_responses channel type by @kirafishy in #2212
• fix(prompts): parse project GUIDs in prompt schema by @Likhixang in #2200
• chore(actions): upgrade workflows to node24 actions by @llc1123 in #2219
• fix(responses): preserve image tool_choice without pass-through by @llc1123 in #2218
• fix(codex): skip pass-through for unsupported response limits by @llc1123 in #2217
• feat(requests): add reusable column reordering with drag-and-drop by @djdembeck in #2222
• feat: persit chunks for failed stream request by @looplj in #2224
• feat(quota): allow exempting channels from quota filtering by @djdembeck in #2214
• feat: add xAI subscription SSO channel and Responses support for xai by @llc1123 in #2225
• feat: provider quota estimate by @looplj in #2228
• feat: add codex missing headers by @looplj in #2230
• fix(actions): derive unstable version from latest tag by @llc1123 in #2231
• fix(llm): preserve top-level reasoning tokens by @djdembeck in #2237
• fix(nanogpt): escalate quota status to exhausted at 100% by @djdembeck in #2236
• fix(codex): fix HTTP 400 missing reasoning.context all_turns by @djdembeck in #2235
• feat(frontend): add sans/serif/mono font options to theme menu by @ChancenJ in #2234
• chore: minor ui/error optimization by @looplj in #2238
• fix: image outbound & override by @looplj in #2241
• fix: restore meta muse lineup in model catalog by @djdembeck in #2258
• fix(channel): write UTC updated_at when saving model prices by @ChancenJ in #2257
• fix(frontend): grant system-level route access to owner users by @qinkangdeid in #2254
• fix(copilot): send Grok models through Responses API by @ripenedcat in #2250
• opt: gemini channel compatible by @looplj in #2259
• chore: sync model developers data by @github-actions[bot] in #2242
• feat: introduce unified opencode transformer, close #2240 by @looplj in #2260
• feat: GC 支持剥离已存储的请求/响应载荷 by @CberYellowstone in #2246
• fix(frontend): correct conversation sidebar height by @KremeCN in #2267
• fix: skip postgres monotonic updated_at cleanup by @djdembeck in #2266
• fix: use HTTP URL for WebSocket model discovery by @techblack in #2265
• fix: should reload channels after txn commited, close #2256 by @looplj in #2261
• fix: forward image parameter in /v1/images/generations by @clhey in #2251
• fix: double count reasoning token for throughput by @looplj in #2270
• fix(requests): expose personal key callers to project owners by @brynne8 in #2269
• fix: use ping for Responses WebSocket channel tests by @techblack in #2264
• fix(requests): honor effective API key scopes by @brynne8 in #2277
• fix: stabilize auto-refresh list transitions by @AkaChou in #2275
• fix(codex): accept completed JSON Responses from compatible relays by @MrLiuGangQiang in #2243
• fix(channels): map ollama_anthropic to the ollama provider by @hihanley in #2299
• chore: sync model developers data by @github-actions[bot] in #2295
• fix: i18n conversation viewer and tool_choice stat by @djdembeck in #2298
• docs(i18n): clarify storage policy and payload storage copy by @brynne8 in #2296
• fix: preserve conditional template overrides by @CBOSSX in #2287
• docs(i18n): clarify where store-chunks persists stream chunks by @trim21 in #2306
• feat: support Codex alpha search proxy by @JerryLiu369 in #2274
• fix(providers): accept boolean experimental flag by @djdembeck in #2305
• fix(gemini): preserve concrete MIME types for remote media by @llc1123 in #2205
• feat: hot-reload model catalog from upstream via backend by @CberYellowstone in #2263
• fix(channels): validate bulk ordering weight input by @JasonWenTheFox in #2335
• fix(frontend): support clipboard copy on insecure HTTP origins by @JasonWenTheFox in #2334
• feat(frontend): add 100-row pagination option by @JasonWenTheFox in #2333
• fix: skip spurious default-value changes that trigger SQLite rebuilds by @ChancenJ in #2331
• chore: sync model developers data by @github-actions[bot] in #2325
• fix(ui): add viewport height floor to DialogContent by @ronronner02 in #2328
• feat(channels): support export/import channel model prices as JSON by @ChancenJ in #2322
• fix(frontend): trim price inputs before saving channel model prices by @trim21 in #2321
• feat: auto set prompt cache key from trace id by @looplj in #2339
• fix: request performance metric accurate by @looplj in #2341
• feat: support add image in playground by @BLumia in #2340
• fix: decode snake_case OIDC claims for email linking by @imryao in #2320
• fix: classify upstream stream interruptions and keep latency metrics on failed executions by @zhaozhaozz in #2317
• fix(frontend): paginate API key selectors and filters by @suixinio in #2346
• ci: validate bundled providers.json against its schema by @trim21 in #2315
• feat: return AxonHub cost on usage.cost across API formats by @trim21 in #2300
• feat(api): support downstream Responses WebSocket by @techblack in #2255
• feat: add mine api keys tab, close #2271 by @looplj in #2349
• fix: requests db index name length & rules by @looplj in #2351

New Contributors

• @hanuxxx made their first contribution in #2202
• @jveko made their first contribution in #2204
• @kirafishy made their first contribution in #2212
• @Likhixang made their first contribution in #2200
• @ripenedcat made their first contribution in #2250
• @techblack made their first contribution in #2265
• @clhey made their first contribution in #2251
• @CBOSSX made their first contribution in #2287
• @JerryLiu369 made their first contribution in #2274
• @ronronner02 made their first contribution in #2328
• @imryao made their first contribution in #2320

Full Changelog: v1.0.0-beta7...v1.0.0-beta8

What's Changed

• fix: handle moderation responses in empty response detection by @Silentely in #2099
• feat: add provider quota collection controls by @Silentely in #2103
• fix(gemini): preserve file data MIME type for Vertex by @llc1123 in #2091
• feat: backup & restore systems config, close #2101 by @looplj in #2104
• feat: response with trace headers, close #2100 by @looplj in #2107
• feat: check new beta version, close #2097 by @looplj in #2111
• fix(frontend): improve requests table mobile responsiveness by @djdembeck in #2118
• fix: reuse bundled Windows binary by @Jonesxq in #2119
• fix(analytics): support MySQL for channel stats by @ChancenJ in #2123
• fix: 修正拼写错误“oltphttp”->“otlphttp” by @bookshiyi in #2121
• fix(frontend): load theme fonts by @llc1123 in #2122
• fix(responses): keep images in tool results instead of dropping them by @GhostFlying in #2117
• feat: add configurable max_multipart_memory for backup restore by @vincentyzhj in #2126
• feat: add per-channel API key rule actions by @Lxiny-zy in #2120
• fix(transformer): 合并连续 function_call/custom_tool_call 为单条 assistant 消息 by @Snowy117 in #2114
• fix(responses): scope fallback prompt_cache_key per conversation by @Kizunad in #2031
• feat(permissions): align project roles with UI access by @brynne8 in #2095
• feat: model load balance strategy by @looplj in #2127
• chore: sync model developers data by @github-actions[bot] in #2131
• fix: admin page permission protect by @hen7777777 in #2138
• fix(responses): handle reasoning text stream events by @cesaryuan in #2151
• fix: harden MemorySize parsing and document max_multipart_memory by @vincentyzhj in #2145
• fix(openai): keep message content valid for strict upstreams by @Lxiny-zy in #2137
• fix(frontend): constrain quota popover width on mobile by @KremeCN in #2143
• fix(frontend): prevent system settings double scroll by @KremeCN in #2142
• fix(cline): align quota windows with API behavior by @KremeCN in #2141
• feat(cline): fetch recommended models dynamically by @KremeCN in #2140
• feat(channels): add Groq channel by @rupaut98 in #2144
• fix(transform): add channel switch to downgrade mid-conversation system messages by @AkaChou in #2124
• feat(requests): show final reasoning_effort in request logs by @AkaChou in #2158
• feat(channels): unified API key management dialog with bulk import by @Lxiny-zy in #2156
• fix(cline): normalize empty message content by @KremeCN in #2155
• fix(frontend): align analytics filter UX by @cesaryuan in #2154
• fix(responses): preserve function calls completed in done events by @MrLiuGangQiang in #2057
• feat(requests): redesign request log table by @brynne8 in #2162
• perf(frontend): virtualize model price dialog to fix lag with many models by @suixinio in #2163
• feat: add sse keep alive, close #2146 by @looplj in #2157
• chore: sync model developers data by @github-actions[bot] in #2176
• feat(channels): per-credential auto disable with scheduled recovery by @zhaozhaozz in #2180
• fix(responses): preserve polymorphic reasoning and Codex metadata by @MedAIerHHL in #2178
• feat(apikeys): sync and transfer profile templates by @MedAIerHHL in #2177
• fix(responses): report truncation and failure in streaming conversion by @Lxiny-zy in #2171
• fix(responses): 兼容 Responses API created_at 的浮点整数形式 by @Silentely in #2187
• fix(stream): report incomplete streams to the client by @zhaozhaozz in #2185
• request 请求体支持 对话阅览 模式,便于请求体数据查看 by @qintang in #2182
• feat: add qiniu/fenno channel by @looplj in #2188
• fix(claude-code): automate cache compatibility and OpenAI effort mapping by @AkaChou in #2172
• feat(models): add thinkingmachines as model developer by @djdembeck in #2190
• fix(sqlite-time-format-compat): 兼容 SQLite TEXT 时间戳格式并修复 backup 注册时区退化 by @Silentely in #2189
• chore: add log for chat heartbeat by @looplj in #2191
• feat: show request cache rate, close #2170 by @looplj in #2193
• fix(responses): retry incomplete streams before done by @wtj-0527 in #2192
• fix: model associate condition caused model not found, close #2183 by @looplj in #2194

New Contributors

• @Jonesxq made their first contribution in #2119
• @bookshiyi made their first contribution in #2121
• @vincentyzhj made their first contribution in #2126
• @Lxiny-zy made their first contribution in #2120
• @Snowy117 made their first contribution in #2114
• @Kizunad made their first contribution in #2031
• @rupaut98 made their first contribution in #2144
• @AkaChou made their first contribution in #2124
• @MrLiuGangQiang made their first contribution in #2057
• @qintang made their first contribution in #2182
• @wtj-0527 made their first contribution in #2192

Full Changelog: v1.0.0-beta6...v1.0.0-beta7

What's Changed

• fix(dashboard): fix daily chart X-axis date display using Date.UTC by @ChancenJ in #1995
• fix: dialog ui for test channel when enable channel, close #1969 by @looplj in #1996
• fix(anthropic): synthesize terminal stream events when upstream omits finish_reason by @mvanhorn in #2000
• chore: sync model developers data by @github-actions[bot] in #2004
• feat: display last name before first name in sidebar for Chinese locale by @hihanley in #1998
• fix(apikeys): avoid double-counting cached and reasoning tokens in total by @hihanley in #1999
• feat: add IBM model developer with Granite models by @djdembeck in #2003
• feat: add analytics page with multi-dimensional filtering by @ChancenJ in #2005
• feat(traces): add span statistics to trace detail page and filter pop… by @ChancenJ in #2008
• fix: move analytics helpers out of resolvers file to prevent go generate from breaking them by @ChancenJ in #2019
• feat: support set-if-absent body overrides by @GhostFlying in #2017
• feat: expose prompt cache key to override templates by @GhostFlying in #2016
• fix(provider-quota): use official Cline quota limits by @KremeCN in #2012
• feat(channel): auto-configure model prices by @ousing9 in #1993
• feat: add Kimi Code quota tracking by @u33e in #2010
• fix(codex): preserve Responses Lite request requirements by @shay-wong in #2020
• fix(traces): normalize tool_use arguments JSON to fix span dedup by @ChancenJ in #2013
• feat: config auto reload by @looplj in #1997
• fix: embed IANA timezone database for Windows compatibility by @ChancenJ in #2024
• feat: add trace and thread status management (archive / retain) by @ChancenJ in #2023
• chore: sync model developers data by @github-actions[bot] in #2026
• feat(quota): add Minimax channel quota checker by @ChancenJ in #2029
• fix(openai): write image edit model before files by @llc1123 in #2028
• chore: sync model developers data by @github-actions[bot] in #2037
• feat(quota): add Zhipu channel quota checker by @ChancenJ in #2032
• fix(ent): align prompt project ownership by @brynne8 in #2039
• feat(models): add Tencent developer with Hunyuan models by @KtzeAbyss in #2046
• feat: add archive/retain buttons to trace and thread detail pages by @ChancenJ in #2045
• fix(channels): preserve third-party Codex base URL by @brynne8 in #2040
• feat(models): add poolside as model developer by @djdembeck in #2059
• feat(channels): add Anthropic messages API format support to the Ollama channel by @hihanley in #2058
• feat: 支持配置全局渠道测试 Prompt by @CberYellowstone in #2042
• feat:增加每次请求新连接功能，使得类似resin之类的代理池能够轮换ip by @xuyufengfei in #2060
• fix(permissions): align system settings scope by @brynne8 in #2053
• feat: add time-based pricing schedule for channel model prices by @ChancenJ in #2050
• fix: Codex Remote Compaction V2 的响应在 AxonHub 中被当作 Empty Reponse 错误 by @w568w in #2041
• fix(openai): transform Responses image results by @cesaryuan in #2071
• fix(orchestrator): detect stream completion from JSON data type and finish_reason by @trim21 in #2068
• fix(backup): stream backup JSON to prevent OOM on large datasets by @djdembeck in #2067
• fix(apikeys): make table sorting functional by @MedAIerHHL in #2063
• feat(apikeys): add allowed cidrs for api keys by @amghazanfari in #2072
• fix: pass through canceled request check by @looplj in #2030
• Import mobile UI by @nb5p in #1896
• feat: add cache write for response api by @looplj in #2074
• feat: add Inclusion AI as model developer by @djdembeck in #2078
• refactor: make sesstion sticky configurable by @looplj in #2076
• chore: sync model developers data by @github-actions[bot] in #2081
• fix(openai): normalize empty finish_reason to nil for providers that always emit it by @sleepwalkera in #2073
• fix(responses): expand namespace function tools by @525300887039 in #2079
• feat: support OpenAI-compatible /v1/moderations by @Silentely in #2082
• fix: data storage page crash by @looplj in #2085
• feat(project-users): add invitation registration by @brynne8 in #2055
• fix(frontend): keep detail page header visible when scrolling by @KremeCN in #2087
• fix(anthropic): preserve multiple Responses reasoning signatures by @RinStel in #2054
• fix(orchestrator): retry transport-level errors via channel failover by @Yuna-Celisse in #2090

New Contributors

• @mvanhorn made their first contribution in #2000
• @hihanley made their first contribution in #1998
• @ousing9 made their first contribution in #1993
• @u33e made their first contribution in #2010
• @CberYellowstone made their first contribution in #2042
• @w568w made their first contribution in #2041
• @MedAIerHHL made their first contribution in #2063
• @amghazanfari made their first contribution in #2072
• @sleepwalkera made their first contribution in #2073
• @525300887039 made their first contribution in #2079
• @RinStel made their first contribution in #2054
• @Yuna-Celisse made their first contribution in #2090

Full Changelog: v1.0.0-beta5...v1.0.0-beta6

What's Changed

• opt(transformer): build image data URLs via xurl.BuildDataURL by @suixinio in #1882
• chore: sync model developers data by @github-actions[bot] in #1887
• fix: passThrough column not toggleable in column visibility settings by @Silentely in #1886
• fix:跨渠道重试成功之后将原失败渠道记入渠道面板的监控统计 by @xuyufengfei in #1891
• fix(openai): avoid stream aggregation nil panics by @luoxue03 in #1890
• feat(openapi): support querying api keys and templates by name by @suixinio in #1838
• fix: tracing document config by @looplj in #1897
• opt: admin page permission guard, close #1894 by @looplj in #1898
• fix(frontend): avoid logout on upstream auth errors by @lihaojiee in #1902
• fix(responses): preserve encrypted reasoning item boundaries by @ttttmr in #1901
• fix: test request detail nav back, close #1912 by @looplj in #1915
• feat: add scripts to cleanup sqlite, close #1905 by @looplj in #1916
• opt: auto set default test model, close #1895 by @looplj in #1917
• fix(ollama): extract token usage from final stream chunk for correct status by @myml in #1923
• feat(orchestrator): enforce strict per-channel max concurrency by @suixinio in #1920
• fix: scope-level aware RouteGuard and silent 403 handling by @brynne8 in #1909
• fix(playground): only show thinking in last assistant message by @myml in #1461
• fix: clear the previous state, close #1918 by @looplj in #1925
• chore: sync model developers data by @github-actions[bot] in #1934
• feat: add OpenCode Go provider quota tracking by @zhaozhaozz in #1926
• fix(channels): fix test API keys dialog footer buttons being hidden by @Silentely in #1942
• fix(channels): Update the Base URL of the Ollama service provider to the official address. by @Silentely in #1941
• fix: improve error reporting for /v1/models endpoint by @trim21 in #1931
• opt: reduce S3 Class A operations with native ObjectStore by @suixinio in #1939
• fix(channel): add missing optional() to modelMappings zod schema by @junjiangao in #1946
• fix: retry stream errors before first content by @lihaojiee in #1944
• feat: add round-robin load balancer strategy by @yuanhhs in #1953
• feat: add personal API key type and per-user usage statistics by @brynne8 in #1933
• fix(httpclient): cap upstream error response body at 1 MB to prevent OOM by @djdembeck in #1955
• feat: ip access control by @looplj in #1956
• chore: remove the test source filter for request, close #1913 by @looplj in #1957
• feat: show reasoning token usage by @looplj in #1960
• feat: add codex headers by @looplj in #1963
• chore: sync model developers data by @github-actions[bot] in #1964
• ci: publish Helm chart to GHCR --issue=#1965 by @zouxingyuks in #1966
• feat: openai inbound accept thinking to compatible with ds/zhipu by @looplj in #1970
• feat: add Cline channel support by @KremeCN in #1972
• fix(aggregator): skip non-standard SSE events with empty choices by @Black0416 in #1959
• fix(trace): include instruction content in system_instruction span key by @ChancenJ in #1974
• feat: add Cline provider quota support by @KremeCN in #1975
• fix(frontend): constrain API keys table scrolling by @KremeCN in #1977
• feat(codex): update models and default version by @zouxingyuks in #1985
• feat(frontend): add GPT-5.6 models and pricing by @shay-wong in #1982
• feat(ollama): add tool call support for request and response transfor… by @myml in #1967
• feat(transformer): 渠道级 reasonning effort mapping by @sycghj in #1910
• chore: sync model developers data by @github-actions[bot] in #1988
• feat: model condition with requeset header by @looplj in #1986
• ci: add nightly docker publish workflow by @trim21 in #1981
• feat(channels): add manual disable/enable API key in channel edit panel by @ChancenJ in #1990

New Contributors

• @junjiangao made their first contribution in #1946
• @yuanhhs made their first contribution in #1953
• @zouxingyuks made their first contribution in #1966
• @Black0416 made their first contribution in #1959
• @ChancenJ made their first contribution in #1974
• @shay-wong made their first contribution in #1982
• @sycghj made their first contribution in #1910

Full Changelog: v1.0.0-beta4...v1.0.0-beta5

What's Changed

• fix: preserve filter state when navigating to request detail page by @MoshiCoCo in #1819
• opt: add empty tolerance for claude code read, close #1681 by @looplj in #1820
• 修复：复制渠道时同步复制模型价格 by @luoxue03 in #1811
• opt: gc & retry by @looplj in #1823
• fix(i18n): add missing translation keys for 3 anthropic channel variants by @MoshiCoCo in #1824
• chore: upgrade claude code oauth by @looplj in #1827
• fix(frontend): 允许 Codex 渠道切换协议 by @llc1123 in #1808
• feat:在渠道配置中新增“重试错误信息”规则 by @xuyufengfei in #1828
• chore: sync model developers data by @github-actions[bot] in #1830
• feat: update default models for bailian, xiaomi, volcengine, zhipu channels by @MoshiCoCo in #1829
• feat: codex image gen/edit, close #1816 by @looplj in #1831
• chore: channel action dialog minor ui by @looplj in #1832
• docs: correct dev Go version floor to 1.26 to match go.mod by @Dashsoap in #1833
• opt: openai response schema by @looplj in #1834
• feat: toggle IP ban icon by @nb5p in #1848
• feat: add GLM-5.2 to zai channel default models by @qinkangdeid in #1842
• fix(oidc): avoid privacy-deny error on every sign-in by @qinkangdeid in #1841
• fix: model circuit breakder, close #1821 by @looplj in #1854
• fix(model_fetcher): avoid duplicate /v1 for anthropic-like endpoints by @KremeCN in #1852
• feat: add array_remove override op by @cesaryuan in #1851
• feat: reset codex quota right now, close #1849 by @looplj in #1855
• fix(frontend): correct api key creator filter scope by @KremeCN in #1856
• fix: allow third-party Codex channels to edit Base URL by @brynne8 in #1835
• fix: enforce strict local channel RPM admission by @lihaojiee in #1840
• fix: openai video schema, close #1845 by @looplj in #1857
• fix(frontend): hide api key creator column without system user permission by @KremeCN in #1859
• feat: test all api keys, close #1853 by @looplj in #1858
• opt: init channel performances async, close #1861 by @looplj in #1864
• fix(frontend/dashboard): align channel success rate card height with daily overview by @lihaojiee in #1862
• fix: should drop the invalid signature for unknown openai/anthropic, … by @looplj in #1866
• fix: api key token usage scope, close #1865 by @looplj in #1867
• feat: highlight low cache hit rate, close #1837 by @looplj in #1868
• fix: hide prefixed model, close #1732 by @looplj in #1869
• fix: codex image generation by @looplj in #1870
• chore(deps-dev): bump vite from 7.3.2 to 7.3.5 in /frontend by @dependabot[bot] in #1874
• feat: add transformer for cerebras, close #1777 by @looplj in #1876
• fix: image generation bypass, close #1802 by @looplj in #1877
• feat: record request url & pass-through flag, close #1798 by @looplj in #1878
• opt: add busy timeout for sqlite dsn, close #1768 by @looplj in #1880

New Contributors

• @luoxue03 made their first contribution in #1811
• @Dashsoap made their first contribution in #1833
• @KremeCN made their first contribution in #1852
• @cesaryuan made their first contribution in #1851
• @brynne8 made their first contribution in #1835
• @lihaojiee made their first contribution in #1840

Full Changelog: v1.0.0-beta3...v1.0.0-beta4

What's Changed

• chore: 调整图像输入size大小限制 by @sxueck in #1769
• feat: add modalities field to /v1/models API response by @MoshiCoCo in #1772
• feat: add IP blocklist settings by @hen7777777 in #1771
• fix(frontend): guard execution channel query by permission by @MoeMion in #1770
• opt: backup/restore usages logs seprated, close #1712 by @looplj in #1766
• chore: upgrade claude code headers by @looplj in #1774
• fix: allow users to update language preference by @evanhon0110 in #1782
• fix: cache did not refresh sometime by @looplj in #1783
• fix(datastorage): remove unbounded in-memory cache of request/response bodies (OOM) by @suixinio in #1781
• fix(llm): claude code quota check header by @looplj in #1785
• feat(openapi): add apiKeyQuotaUsages query for service-account keys by @suixinio in #1779
• fix(openai): aggregate stream choices by index to avoid nil panic by @SAY-5 in #1775
• chore: upgrade quic by @looplj in #1786
• feat: add OpenAI audio APIs (TTS / STT) with SSE streaming by @EmccK in #1784
• chore: sync model developers data by @github-actions[bot] in #1795
• fix(audio): 修复 TTS 流式 Accept、外部存储与持久化内存占用 by @EmccK in #1792
• feat(channels): adding Anthropic messages API format support to the OpenCodeGo channel (#1751) by @MoshiCoCo in #1791
• feat: support redis multi mode by @wangYX657211334 in #1788
• refactor(neuralwatt): replace current_period_start/end with kwh_reset_date by @djdembeck in #1794
• fix: preserve Responses function call namespace by @DevilSi0722 in #1797
• feat: 添加响应超时重试配置 by @xuyufengfei in #1804
• feat: add evolink provider by @EvoLinkAI in #1773
• feat(quota): add Apertis quota monitoring and refactor openai_responses handling by @djdembeck in #1793
• feat: channel retryable statu codes, close #1803 by @looplj in #1807
• feat: route with media condition, close #1729 by @looplj in #1809

New Contributors

• @MoeMion made their first contribution in #1770
• @evanhon0110 made their first contribution in #1782
• @SAY-5 made their first contribution in #1775
• @EmccK made their first contribution in #1784
• @DevilSi0722 made their first contribution in #1797
• @xuyufengfei made their first contribution in #1804
• @EvoLinkAI made their first contribution in #1773

Full Changelog: v1.0.0-beta2...v1.0.0-beta3

What's Changed

• chore: sync model developers data by @github-actions[bot] in #1706
• fix: scroll for the channel rate limit dialog, close #1701 by @looplj in #1708
• feat: render structured request log content, close #1698 by @looplj in #1711
• fix: 修复了多图输入情况下转发失效的问题 by @sxueck in #1714
• feat: add advanced model route conditions, close #1670 by @looplj in #1717
• 修复 Responses 非透传模式下工具原语保真 by @deqiying in #1703
• feat: support non-stream fallback to require-stream channels by @ttttmr in #1661
• fix(llm): accept structured response item arguments by @llc1123 in #1728
• fix(channels): show real i18n label for single-variant types like opencode_go by @qinkangdeid in #1734
• chore: add claude-opus-4-7 to claudecode DefaultModels by @qinkangdeid in #1733
• feat(llm): support responses websocket sessions by @llc1123 in #1730
• Preserve request log filters on detail navigation by @zhaozhaozz in #1723
• chore: 添加了登陆后页面标题自定义配置项 by @sxueck in #1720
• fix: model settings dialog, close #1716 by @looplj in #1718
• Support case-insensitive model filter patterns by @zhaozhaozz in #1725
• chore: add claude-opus-4-8 to claudecode DefaultModels by @qinkangdeid in #1739
• feat: opencode trace, close #1740 by @looplj in #1744
• feat: reactivate api key, close #1735 by @looplj in #1745
• fix: model association page throws Invalid input for channels without endpoints by @djdembeck in #1746
• chore: sync model developers data by @github-actions[bot] in #1754
• fix(anthropic): handle server-side tool blocks without panicking by @xdqi in #1750
• opt: request detail sheet annimation, close #1749 by @looplj in #1757
• fix: should set default model type if type not present, close #1752 by @looplj in #1760
• feat: support disabling db auto migration by @wangYX657211334 in #1727
• fix(frontend): avoid leaking "{{resource}}" placeholder in NOT_FOUND errors by @qinkangdeid in #1762
• fix: auto reasoning effort for qwen max, close #1751 by @looplj in #1765

New Contributors

• @sxueck made their first contribution in #1714
• @xdqi made their first contribution in #1750
• @wangYX657211334 made their first contribution in #1727

Full Changelog: v1.0.0-beta1...v1.0.0-beta2

What's Changed

• feat: 请求日志记录新增'推理强度'字段 by @deqiying in #1506
• feat: per-channel concurrency queue + admission control by @suixinio in #1503
• feat: auto reasoning effort with suffix by @looplj in #1515
• feat: add optional read-write separation for database by @lazhenyi in #1516
• feat(quota): add wafer, synthetic, and neuralwatt quota checkers by @djdembeck in #1509
• fix(dashboard): align token stats by API key/channel/model with totals by @raikyou in #1524
• feat: add Anthropic endpoints for Volcengine, AiHubMix, and Xiaomi MiMo by @banlanzs in #1521
• display token usage in channel cards via tokenStatsByChannel API by @banlanzs in #1528
• feat(quota): add provider quota enforcement with per-limit status tracking by @djdembeck in #1518
• feat: customized channel endpoints by @looplj in #1529
• fix(pipeline): prevent channels from getting stuck in processing status by @djdembeck in #1530
• chore: reformat llm package by @looplj in #1531
• fix(models): clicking Archive on a disabled model had no effect by @djdembeck in #1535
• fix(ui): show used/total in NanoGPT quota display instead of used/remaining by @djdembeck in #1543
• fix(models): resolve vision capability from modalities when explicit field is absent by @djdembeck in #1542
• opt: disable spell check for API key textarea by @BLumia in #1544
• opt: custom endpoint api format by @looplj in #1545
• feat: align endpoint i18n and prompt protection preview by @looplj in #1546
• feat(oidc): Initial OpenID Connect Support by @LazuliKao in #1091
• scheduler: centralized cron job management by @looplj in #1550
• chore: update issue template by @looplj in #1551
• fix(llm): missing anthropic thinking disabled transform by @looplj in #1552
• feat: openai/deepseek completion api, close #1458 by @looplj in #1554
• chore: change prompt content column, close #1553 by @looplj in #1580
• fix: max model association priority, close #1507 by @looplj in #1581
• opt: responses stream failed handle & request executions index, close #1473 by @looplj in #1585
• opt: openai outbound reasoning field config, close #1584 by @looplj in #1586
• chore: sync model developers data by @github-actions[bot] in #1588
• feat: global pass-through setting by @raikyou in #1591
• feat: add lowercaseModelName channel setting for cross-channel load b… by @nb5p in #1589
• feat: expose inbound request headers to override templates by @ttttmr in #1520
• feat(apikeys): add API key profile templates by @djdembeck in #1590
• feat: prefer outbound format equals to inbound format by @looplj in #1592
• 修改前端，"测试场"新增测试模型网关模型 by @zhai23 in #1597
• fix(gc): drop nil bind args so VACUUM actually runs on Postgres by @suixinio in #1602
• chore: add gpt-5.5 for codex plans by @qinkangdeid in #1605
• refactor: remove the signature prefix marker to improve pass through by @looplj in #1593
• Support image parts in Ollama chat transformer by @zhzy0077 in #1610
• feat: add Qiniu provider support by @JackChiang233 in #1612
• fix(channels): import Qiniu icon in config_channels by @suixinio in #1613
• feat(openapi): add updateAPIKeyProfiles and loadApiKeyProfileTemplate by @suixinio in #1617
• chore: sync model developers data by @github-actions[bot] in #1627
• fix: detect synthetic weekly quota exhausted state by @djdembeck in #1619
• feat: auto enable wal for sqlite by @looplj in #1633
• feat: customized channel endpoints with base url by @looplj in #1635
• fix: prevent panic when client canceled for pass-through, close #1634 by @looplj in #1637
• fix: should omit empty name for function call, close #1587 by @looplj in #1639
• fix: fetch models for codex with auth.json, close #1636 by @looplj in #1645
• feat: hide upstream error message by @LRan1028 in #1640
• feat: gc now with options, close #1618 by @looplj in #1647
• feat: backup usage stats by @henryz78 in #1632
• feat(transformer): add cross-provider search citation and annotation support by @ttttmr in #1643
• feat: add volume price mode, close #1582 by @looplj in #1651
• doc: update add channel doc by @looplj in #1653
• chore(llm): change openai default reasoning_field to content by @looplj in #1656
• fix(llm): clear ReasoningEffort when DeepSeek thinking is disabled by @Lucent-Snow in #1663
• fix: request detail preview in dark mode by @looplj in #1666
• feat: add modelBlacklistRegex system setting to filter channel-derive… by @nb5p in #1673
• chore: sync model developers data by @github-actions[bot] in #1675
• fix(llm): close text block before starting thinking block in inbound stream by @hen7777777 in #1677
• feat: opencode go channel, close #1614 by @looplj in #1682
• Add developer-level model association rules by @zhaozhaozz in #1672
• feat: auto decompress request body, close #1680 by @looplj in #1688
• fix(pipeline): expose RawRequest before llm middlewares by @ttttmr in #1687
• feat: introduce Atlas Cloud as sponsor by @looplj in #1693
• feat: add atlascloud provider by @lucaszhu-hue in #1695
• feat: rotate api key, close #1692 by @looplj in #1696
• fix: should not error if can access to api keys, close #1674 by @looplj in #1697
• fix: cache not refresh in some cache by @looplj in #1699
• fix(gc): avoid SELECT * when querying old requests/executions for cle… by @FlyLoongZ in #1700

New Contributors

• @suixinio made their first contribution in #1503
• @BLumia made their first contribution in #1544
• @ttttmr made their first contribution in #1520
• @zhai23 made their first contribution in #1597
• @zhzy0077 made their first contribution in #1610
• @JackChiang233 made their first contribution in #1612
• @LRan1028 made their first contribution in #1640
• @henryz78 made their first contribution in #1632
• @Lucent-Snow made their first contribution in #1663
• @zhaozhaozz made their first contribution in #1672
• @lucaszhu-hue made their first contribution in #1695
• @FlyLoongZ made their first contribution in #1700

Full Changelog: v0.9.38...v1.0.0-beta1