release(main): release onwards 2.0.0

.release-please-manifest.json

@@ -2,5 +2,5 @@
   ".": "15.4.0",
   "fusillade-arsenal": "4.1.0",
   "fusillade-core": "6.0.0",
-  "onwards": "1.2.0"
+  "onwards": "2.0.0"
 }

onwards/CHANGELOG.md

@@ -5,6 +5,76 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [2.0.0](https://github.com/doublewordai/control-layer/compare/onwards-v1.2.0...onwards-v2.0.0) (2026-10-23)
+
+
+### ⚠ BREAKING CHANGES
+
+* retire the serving-class presets, grants and default classes ([#2068](https://github.com/doublewordai/control-layer/issues/2068))
+* **limits:** replace gateway rate limits with per-account realtime in-flight limits ([#1915](https://github.com/doublewordai/control-layer/issues/1915))
+* **onwards:** frame SSE before applying buffer limit
+
+### Features
+
+* account files carry priority, latency targets and prices ([#2075](https://github.com/doublewordai/control-layer/issues/2075)) ([a2f91a6](https://github.com/doublewordai/control-layer/commit/a2f91a6e2bfad9f67af1a47116aa9e613c1d29dc))
+* count the inference requests dwctl's own layers refuse (DW-851) ([#1978](https://github.com/doublewordai/control-layer/issues/1978)) ([b822bd0](https://github.com/doublewordai/control-layer/commit/b822bd0bae2731a1db94679479ae86e5b304e6e2))
+* **dwctl:** let slow first tokens stream until a 20-second failover deadline ([#1808](https://github.com/doublewordai/control-layer/issues/1808)) ([444ad7c](https://github.com/doublewordai/control-layer/commit/444ad7cfd5d04d5077ef62b61a2fb9cd72b5f03e))
+* **dwctl:** record where a translated body fails its schema, without its content (DW-951) ([#1991](https://github.com/doublewordai/control-layer/issues/1991)) ([53b07dc](https://github.com/doublewordai/control-layer/commit/53b07dc2bf5932cc3c037f803095eb46e90b161f))
+* **limits:** cap batch requests in flight per model in onwards ([#1924](https://github.com/doublewordai/control-layer/issues/1924)) ([834827d](https://github.com/doublewordai/control-layer/commit/834827dbc2b09604e9759f8b715eb85b3e2c1e22))
+* **limits:** count in-flight limit checks and refusals per account ([#1957](https://github.com/doublewordai/control-layer/issues/1957)) ([f6a7486](https://github.com/doublewordai/control-layer/commit/f6a74865a43a71bb9c3d9228958dfde66c364477))
+* **limits:** name the account and report its limit on in-flight metrics ([#1958](https://github.com/doublewordai/control-layer/issues/1958)) ([72c6e97](https://github.com/doublewordai/control-layer/commit/72c6e974d435a20ed5cd36f977016434b3b7be4b))
+* **limits:** replace gateway rate limits with per-account realtime in-flight limits ([#1915](https://github.com/doublewordai/control-layer/issues/1915)) ([be53043](https://github.com/doublewordai/control-layer/commit/be530434b6d2fec1c5c2cfa0c366c9ced6954a5e))
+* **onwards:** check legacy completions for unsupported parameters (DW-855) ([#1996](https://github.com/doublewordai/control-layer/issues/1996)) ([500873b](https://github.com/doublewordai/control-layer/commit/500873b76a678ff2d2983d04c5e797121a285111))
+* **onwards:** conversation affinity for priority pools ([#1892](https://github.com/doublewordai/control-layer/issues/1892)) ([b2ff624](https://github.com/doublewordai/control-layer/commit/b2ff624480cade3f985b1a02045304add2339705))
+* **onwards:** count and log the requests onwards rejects itself (DW-851) ([#1937](https://github.com/doublewordai/control-layer/issues/1937)) ([98c234e](https://github.com/doublewordai/control-layer/commit/98c234e6c97fd154496c6c2554683c124270b8bb))
+* **onwards:** count sanitized third-party errors ([#2054](https://github.com/doublewordai/control-layer/issues/2054)) ([dc82a0b](https://github.com/doublewordai/control-layer/commit/dc82a0be4ba644fa3563d9ea8d306a5fe072f1b0))
+* **onwards:** flag chat parameters not every backend supports ([#1925](https://github.com/doublewordai/control-layer/issues/1925)) ([b2271fb](https://github.com/doublewordai/control-layer/commit/b2271fb2b626d760937cfaf7ed77b0b0a6fe5ecb))
+* **onwards:** make AIMD share control decide under overload and recover ([#1803](https://github.com/doublewordai/control-layer/issues/1803)) ([8bb4a6f](https://github.com/doublewordai/control-layer/commit/8bb4a6fc3a865993d810bcd671165e13fe7c7ce7))
+* **onwards:** record where a body fails the strict schema, without its content (DW-951) ([#1990](https://github.com/doublewordai/control-layer/issues/1990)) ([e1cb85a](https://github.com/doublewordai/control-layer/commit/e1cb85a5a525bb360edaf0ab2005fed0db6b680e))
+* **onwards:** send a scheduling priority on every dynamo request ([#1848](https://github.com/doublewordai/control-layer/issues/1848)) ([094fe22](https://github.com/doublewordai/control-layer/commit/094fe227e2bcd3c86f2b090a6b7a4fa44eb2d17d))
+* **onwards:** stop logging rejections and unsupported parameters ([#1963](https://github.com/doublewordai/control-layer/issues/1963)) ([68315f4](https://github.com/doublewordai/control-layer/commit/68315f4941a15217cdbd31f0adba3a0c22aa9478))
+* **onwards:** trace rejections and stop logging caller-limit refusals ([#1952](https://github.com/doublewordai/control-layer/issues/1952)) ([f051e4c](https://github.com/doublewordai/control-layer/commit/f051e4c8324f413422fc237abcf3ddac504b9650))
+* overlay targets are scheduling hints, not a serving class ([#2087](https://github.com/doublewordai/control-layer/issues/2087)) ([4723cfa](https://github.com/doublewordai/control-layer/commit/4723cfa5453762d1f40e4afe3e6bdf8ffd68e31e))
+* **responses:** accept Codex tool namespace groups ([#1935](https://github.com/doublewordai/control-layer/issues/1935)) ([9d46bf7](https://github.com/doublewordai/control-layer/commit/9d46bf782d49dec56fa0c262c76e503c3d98e422))
+* serving classes v1 ([#1774](https://github.com/doublewordai/control-layer/issues/1774)) ([e1c9275](https://github.com/doublewordai/control-layer/commit/e1c92756cbe16953360aa4055cf1011992c55767))
+* **tracing:** customer attributes on request and batch spans ([#1972](https://github.com/doublewordai/control-layer/issues/1972)) ([60db3ec](https://github.com/doublewordai/control-layer/commit/60db3ec246429061cc153b9b5bfea07abc218ca2))
+
+
+### Bug Fixes
+
+* **analytics:** checkpoint records before enrichment ([#1789](https://github.com/doublewordai/control-layer/issues/1789)) ([7ed4000](https://github.com/doublewordai/control-layer/commit/7ed4000159b0ea8ef496e84437517c7511a0bce6))
+* answer unreadable request formats with 415, and count content types (DW-985) ([#2083](https://github.com/doublewordai/control-layer/issues/2083)) ([58e3a52](https://github.com/doublewordai/control-layer/commit/58e3a527a892c9a3be0c6b2f20c7b8671babac76))
+* **build:** refresh the image CA bundle after eMudhra intermediate compromise ([#2077](https://github.com/doublewordai/control-layer/issues/2077)) ([ae3442c](https://github.com/doublewordai/control-layer/commit/ae3442c03c4a2ea7c42771499039febc20932a57))
+* Correctly handle otel tracing through onwards ([#1827](https://github.com/doublewordai/control-layer/issues/1827)) ([107d619](https://github.com/doublewordai/control-layer/commit/107d619195a55c5b0cac65801aaffad06256668a))
+* end interrupted streams with an error event (DW-851) ([#1979](https://github.com/doublewordai/control-layer/issues/1979)) ([843f421](https://github.com/doublewordai/control-layer/commit/843f421ef04ec8d4ca29c762f01aae9be1dc7206))
+* give every response the gateway id and send streamed usage only on request ([#1973](https://github.com/doublewordai/control-layer/issues/1973)) ([ccb3919](https://github.com/doublewordai/control-layer/commit/ccb3919682884ea2fb963eb48d55b690d6bd6877))
+* keep API keys out of rate-limit and proxy debug logs ([#1960](https://github.com/doublewordai/control-layer/issues/1960)) ([43bf8e8](https://github.com/doublewordai/control-layer/commit/43bf8e84cf8e08a0d863351b86b12c913f57ec92))
+* **onwards:** answer every capacity refusal with 529 and Retry-After ([#1913](https://github.com/doublewordai/control-layer/issues/1913)) ([8ca5e6d](https://github.com/doublewordai/control-layer/commit/8ca5e6d4802ebc0e3fe85802d35c7dffd06680d4))
+* **onwards:** frame SSE before applying buffer limit ([5ebb4f1](https://github.com/doublewordai/control-layer/commit/5ebb4f122e5cba74fc971afd0e94b4d050ae96a4))
+* **onwards:** give trusted providers' errors the OpenAI envelope ([#1977](https://github.com/doublewordai/control-layer/issues/1977)) ([8d2960e](https://github.com/doublewordai/control-layer/commit/8d2960e1734b6a9efc38f321757811ea64b5171e))
+* **onwards:** return realtime-only failover statuses as sent when no provider is left ([#1811](https://github.com/doublewordai/control-layer/issues/1811)) ([2b954b9](https://github.com/doublewordai/control-layer/commit/2b954b9b22620e4764fffc5793c9e65303170e37))
+* **onwards:** stop re-asking overloaded providers and answer 429 when every provider is full ([#1911](https://github.com/doublewordai/control-layer/issues/1911)) ([ad19d05](https://github.com/doublewordai/control-layer/commit/ad19d05bcda37f50cbf258a79882b75df3efed39))
+* preserve JSON key order in forwarded request bodies ([#1930](https://github.com/doublewordai/control-layer/issues/1930)) ([a8721a8](https://github.com/doublewordai/control-layer/commit/a8721a8ca8886b2437648131488a833e9461be0f))
+* **proxy:** preserve upstream rate-limit responses ([#1800](https://github.com/doublewordai/control-layer/issues/1800)) ([8a2052b](https://github.com/doublewordai/control-layer/commit/8a2052b3627d7c120b42bf436891529ba3de3f89))
+* set TCP_NODELAY on gateway sockets ([#1931](https://github.com/doublewordai/control-layer/issues/1931)) ([909a27e](https://github.com/doublewordai/control-layer/commit/909a27e95f7ace63385c8b8d3872b5a2290dd0d3))
+
+
+### Performance Improvements
+
+* **onwards:** share a pool's key set between requests ([#1901](https://github.com/doublewordai/control-layer/issues/1901)) ([18cd38c](https://github.com/doublewordai/control-layer/commit/18cd38c64a5f6af25a85d29120a3ca186e382226))
+* **tracing:** filter and slim the spans sent to the trace backend (DW-931) ([#1984](https://github.com/doublewordai/control-layer/issues/1984)) ([4a07bb1](https://github.com/doublewordai/control-layer/commit/4a07bb1199d8aaf25cec7a6c19f3815d63b5a3b7))
+
+
+### Code Refactoring
+
+* retire the serving-class presets, grants and default classes ([#2068](https://github.com/doublewordai/control-layer/issues/2068)) ([a28e3b3](https://github.com/doublewordai/control-layer/commit/a28e3b345dad73a30820d3b8498357e9a7d97a94))
+
+
+### Miscellaneous
+
+* **deps:** bump cmov from 0.5.3 to 0.5.4 ([#1228](https://github.com/doublewordai/control-layer/issues/1228)) ([4f328f6](https://github.com/doublewordai/control-layer/commit/4f328f616c9171216a63da9d308a1aaf28a69bed))
+* **deps:** update rust docker tag to v1.99.1 ([#2072](https://github.com/doublewordai/control-layer/issues/2072)) ([a07b9de](https://github.com/doublewordai/control-layer/commit/a07b9de12b2dfff14835b96b4f1bd27a472446fa))
+
 ## [1.2.0](https://github.com/doublewordai/control-layer/compare/onwards-v1.1.0...onwards-v1.2.0) (2026-09-16)
 
 

onwards/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "onwards"
-version = "1.2.0"
+version = "2.0.0"
 edition = "2024"
 rust-version = "1.88.0"
 description = "A flexible LLM proxy library"