v0.6.0

What's Changed

• fix(flutter): stop shipping LTO bitcode in the iOS static lib (pub.dev 256MiB) by @eric8810 in #196
• docs(rfc): RFC-0036 positioning and layered architecture by @eric8810 in #198
• docs: 0.5.0 → 1.0 roadmap (per RFC-0036) by @eric8810 in #197
• test(elevenlabs): fix flaky stream_abort_mid_session by @eric8810 in #199
• docs: full-chain AI SDK alignment design and impact map by @cunninghamcard-bit in #200
• feat: add stable decision API and official TypeSafe Jev provider by @eric8810 in #221
• feat(stream)!: decode SSE with the sse-stream crate, aligned with eventsource-parser by @cunninghamcard-bit in #213
• feat: support official decision providers, candidate scoring, and native media by @eric8810 in #226
• chore: remove stale audit artifacts, provider inventory and unused Flutter desktop runners by @cunninghamcard-bit in #203
• RFC-0039: ops、FFI 与八种绑定契约 by @eric8810 in #209
• RFC-0040: provider 构造、鉴权与 L0 边界 by @eric8810 in #210
• RFC-0043: 产物体积预算与性能门禁 by @eric8810 in #208
• docs: 校准 roadmap 状态与近期依赖 by @eric8810 in #205
• RFC-0033: registry 维护与安全探测 by @eric8810 in #206
• RFC-0036: 补齐 roadmap 全节点设计与验收映射 by @eric8810 in #207
• RFC-0041: 传输回放与治理证据 by @eric8810 in #211
• ci: skip build and test jobs for documentation-only changes by @cunninghamcard-bit in #263
• RFC-0042: V4 类型边界与统一生成 by @eric8810 in #212
• fix: Bedrock SigV4 signing and secret redaction by @cunninghamcard-bit in #248
• docs: a blank key on a custom base URL is sent as an empty key by @cunninghamcard-bit in #266
• ci: speed up test builds, deduplicate caches, and use prebuilt cargo-ndk by @cunninghamcard-bit in #262
• refactor!: remove the generated ProviderName types by @cunninghamcard-bit in #246
• refactor(web): generate the web wire types with ts-rs by @cunninghamcard-bit in #247
• [1/8] chore: CI on every PR and test fixture cleanup by @cunninghamcard-bit in #249
• [2/8] refactor!: tool calls follow the AI SDK by @cunninghamcard-bit in #250
• [3/8] refactor!: messages and content follow the AI SDK by @cunninghamcard-bit in #251
• [4/8] refactor!: results and streams follow the AI SDK by @cunninghamcard-bit in #252
• [5/8] refactor!: providers are AI SDK factories by @cunninghamcard-bit in #253
• [6/8] fix: vendors follow the upstream packages by @cunninghamcard-bit in #254
• [7/8] feat: the remaining upstream model kinds by @cunninghamcard-bit in #255
• [8/8] refactor!: public JSON equals the AI SDK JSON by @cunninghamcard-bit in #256
• [9/13] refactor!: rewrite json_repair in idiomatic Rust by @cunninghamcard-bit in #257
• [10/13] refactor: tool-call repair is a trait by @cunninghamcard-bit in #258
• [11/13] refactor: provider config is plain data by @cunninghamcard-bit in #259
• [12/13] refactor!: vendor streams are Transformer structs by @cunninghamcard-bit in #260
• [13/13] docs: add the native-Rust porting skills by @cunninghamcard-bit in #261
• [14/14] feat!: ops layer, C ABI and bindings by @cunninghamcard-bit in #265
• [15/15] feat!: generate binding DTOs from the Rust types by @cunninghamcard-bit in #268
• RFC-0045: 上游测试迁移策略 by @cunninghamcard-bit in #281
• fix: Vertex location must be one DNS label by @cunninghamcard-bit in #282
• fix: OpenAI output stream keeps the response id and model by @cunninghamcard-bit in #287
• fix: Mistral injects the JSON instruction like the AI SDK by @cunninghamcard-bit in #283
• fix: presets follow the AI SDK structured outputs setting by @cunninghamcard-bit in #284
• fix: Open Responses and Codex send raw chunks by @cunninghamcard-bit in #285
• fix: Vertex Gemini uses the Google model by @cunninghamcard-bit in #291
• fix: catalogue resolves package names from the descriptor table by @cunninghamcard-bit in #288
• ci: CI result requires the C ABI sanitizer and Linux ARM jobs by @cunninghamcard-bit in #333
• [1/3] Error cause in core and network by @cunninghamcard-bit in #289
• RFC-0047: 错误体系与错误原因（cause） by @cunninghamcard-bit in #286
• RFC-0046: provider 协议的小重构与抽象设计 by @cunninghamcard-bit in #294
• RFC-0044: crate 拆分与依赖反转 by @cunninghamcard-bit in #279
• [2/3] Error cause in string variants by @cunninghamcard-bit in #290
• [3/3] Error cause in ops and bindings by @cunninghamcard-bit in #293
• ci: pull requests build the Android example for arm64 only by @cunninghamcard-bit in #366
• docs(rfc-0045): one fix PR per vendor, one commit per difference by @cunninghamcard-bit in #348
• refactor!: drop retryAfterMs from the error envelope by @cunninghamcard-bit in #349
• fix: OpenAI-compatible unparsable stream chunks are AI SDK parse errors by @cunninghamcard-bit in #354
• fix: Vertex image and Live transcription match the AI SDK by @cunninghamcard-bit in #351
• test: port the upstream Mistral tests by @cunninghamcard-bit in #295
• fix: dropping an xAI streaming transcription closes the socket by @cunninghamcard-bit in #311
• fix: OpenAI Responses, chat, transcription and image behaviour match the AI SDK by @cunninghamcard-bit in #368
• test: port the upstream Google Vertex tests by @cunninghamcard-bit in #352
• fix: Amazon Bedrock errors, warnings, usage and transport match the AI SDK by @cunninghamcard-bit in #357
• test: port the upstream OpenAI-compatible tests by @cunninghamcard-bit in #355
• RFC-0048: 录制 schema 3 与传输回放 by @cunninghamcard-bit in #325
• fix: Groq usage.raw keeps null fields by @cunninghamcard-bit in #308
• test: port the upstream Groq tests by @cunninghamcard-bit in #301
• refactor: move the Groq files to the upstream layout by @cunninghamcard-bit in #302
• fix: Anthropic prompts, betas, safeguards and streams match the AI SDK by @cunninghamcard-bit in #360
• fix: Hugging Face stream errors end with an error finish by @cunninghamcard-bit in #307
• test: port the upstream Cohere tests by @cunninghamcard-bit in #296
• fix: Open Responses missing output is an API call error by @cunninghamcard-bit in #346
• refactor: move the Mistral files to the upstream layout by @cunninghamcard-bit in #297
• fix: DeepSeek usage counts must be numbers by @cunninghamcard-bit in #309
• test: port the upstream DeepSeek tests by @cunninghamcard-bit in #299
• refactor: move the DeepSeek files to the upstream layout by @cunninghamcard-bit in #300
• refactor: move the OpenAI-compatible files to the upstream layout by @cunninghamcard-bit in #356
• fix: Google image inputs and live transcription match the AI SDK by @cunninghamcard-bit in #362
• test: port the upstream Google tests by @cunninghamcard-bit in #363
• refactor: move the Google files to the upstream layout by @cunninghamcard-bit in #367
• fix: Anthropic AWS resolves the default fetch per request by @cunninghamcard-bit in #318
• test: port the upstream Anthropic AWS tests by @cunninghamcard-bit in #319
• refactor: move the Anthropic AWS files to the upstream layout by @cunninghamcard-bit in #320
• refactor: move the Cohere files to the upstream layout by @cunninghamcard-bit in #298
• test: port the upstream OpenAI tests by @cunninghamcard-bit in #369
• refactor: move the OpenAI files to the upstream layout by @cunninghamcard-bit in #370
• build: size profile for the Node and Python bindings by @cunninghamcard-bit in #376
• build: staticlib-release profile for the Go and iOS archives by @cunninghamcard-bit in #375
• test: port the upstream Amazon Bedrock tests by @cunninghamcard-bit in #358
• refactor: move the Amazon Bedrock files to the upstream layout by @cunninghamcard-bit in #359
• test: port the upstream Hugging Face tests by @cunninghamcard-bit in #303
• refactor: move the Hugging Face files to the upstream layout by @cunninghamcard-bit in #304
• build: strip sysroot bitcode from the released static archives by @cunninghamcard-bit in #377
• refactor: move the Google Vertex files to the upstream layout by @cunninghamcard-bit in #353
• build: merge and strip the released static archives by @cunninghamcard-bit in #378
• test: port the upstream xAI tests by @cunninghamcard-bit in #312
• refactor: move the xAI files to the upstream layout by @cunninghamcard-bit in #313
• docs(rfc-0043): size measurements and the reduction plan order by @cunninghamcard-bit in #374
• test: port the upstream Azure tests by @cunninghamcard-bit in #324
• refactor: move the Azure files to the upstream layout by @cunninghamcard-bit in #326
• test: port the upstream Open Responses tests by @cunninghamcard-bit in #347
• refactor: move the Open Responses files to the upstream layout by @cunninghamcard-bit in #350
• test: port the upstream Anthropic tests by @cunninghamcard-bit in #361
• refactor: move the Anthropic files to the upstream layout by @cunninghamcard-bit in #364
• refactor: remove two dead provider-utils helpers by @cunninghamcard-bit in #379
• refactor: share response metadata and provider reference helpers by @cunninghamcard-bit in #381
• refactor: share base64 and file data helpers in provider-utils by @cunninghamcard-bit in #380
• refactor: move azure url rules into openai config by @cunninghamcard-bit in #389
• refactor: resolve anthropic host endpoints outside anthropic by @cunninghamcard-bit in #390
• refactor: give openai its own completion model by @cunninghamcard-bit in #391
• refactor: share tool name mapping and strip file extension by @cunninghamcard-bit in #382
• refactor: share tool result to content by @cunninghamcard-bit in #386
• refactor: resolve preset base urls in preset by @cunninghamcard-bit in #392
• refactor: give decision models a small endpoint interface by @cunninghamcard-bit in #393
• refactor: give anthropic hosts an anthropic config by @cunninghamcard-bit in #394
• refactor: share reasoning effort and budget mapping by @cunninghamcard-bit in #383
• refactor: share provider stream error by @cunninghamcard-bit in #384
• refactor: move vertex partner body rewrites into the compatible config by @cunninghamcard-bit in #395
• refactor: share parse provider options by @cunninghamcard-bit in #385
• refactor: share hostname part check and base URL setting by @cunninghamcard-bit in #387
• refactor: share provider credentials and user agent by @cunninghamcard-bit in #388
• perf(providers,ops): cap Bedrock event-stream frames, cache manifest digest, and move catalogue raw values by @cunninghamcard-bit in #410
• feat: carry vendor-declared limits on listed models by @cunninghamcard-bit in #416
• perf(core): bound stream_text pump channel and gate content accumulation by @cunninghamcard-bit in #406
• fix: end SSE streams at data: [DONE] by @cunninghamcard-bit in #411
• refactor(providers): add vendor internal modules and boundary rule 11 (RFC-0046 R2) by @cunninghamcard-bit in #404
• perf(recording): redact JSON in place and bound pending/inconsistent tracking in recorders by @cunninghamcard-bit in #409
• fix: treat a null error member as no error in stream chunks by @cunninghamcard-bit in #412
• fix: keep the caller's content-type on byte request bodies by @cunninghamcard-bit in #413
• fix(luma): align the image model with @ai-sdk/luma 3.0.54 by @cunninghamcard-bit in #431
• test(core): keep the status retry of a video test from racing its timeout by @cunninghamcard-bit in #452
• feat: record where an ApiCallError was built by @cunninghamcard-bit in #415
• feat: add an opt-in strict stream mode by @cunninghamcard-bit in #417
• fix(deepgram): align transcription with @ai-sdk/deepgram 3.1.24 by @cunninghamcard-bit in #434
• fix(elevenlabs): align speech and batch transcription with @ai-sdk/elevenlabs 3.0.54 by @cunninghamcard-bit in #437
• ci(release): stage the release assets by an explicit list by @cunninghamcard-bit in #451
• fix: send the package user-agent from fal, Deepgram, ElevenLabs and Luma by @cunninghamcard-bit in #444
• chore: pin @ai-sdk/fal, deepgram, elevenlabs and luma by @cunninghamcard-bit in #430
• fix(codegen): re-pin the manifest digest after #415 and #417 by @cunninghamcard-bit in #454
• fix(openai): ignore a null error on a response.failed event by @cunninghamcard-bit in #450
• fix(fal): align with @ai-sdk/fal 3.0.54 by @cunninghamcard-bit in #441
• fix: send max_completion_tokens for the Moonshot and ZenMux presets by @cunninghamcard-bit in #414
• chore(release): prepare 0.6.0 by @eric8810 in #455

Full Changelog: v0.5.0...v0.6.0