[1/8] chore: CI on every PR and test fixture cleanup

.github/workflows/ci.yml

@@ -50,8 +50,8 @@ jobs:
   #   inside ffi-build, Java and Kotlin share one job.
   # - Anything else (the Rust crates, tools, scripts, contract fixtures,
   #   Cargo manifests, workflows): every job.
-  # Contract tests run for any code change: they check generated files that
-  # live in every binding.
+  # Generator drift checks run for any code change: they check generated
+  # files that live in every binding.
   #
   # Where it runs, on a pull request: the Rust code is platform-neutral
   # (rustls, no platform cfg outside tools/), so jobs run on Linux, and the
@@ -60,7 +60,7 @@ jobs:
   # manifest (dependencies), a build script, line-ending rules
   # (.gitattributes), the toolchain or the workflows.
   # Shared Rust changes still get one Windows leg (Node x64 tests); Swift
-  # contract tests are macOS-only and run whenever they are needed: they
+  # tests are macOS-only and run whenever they are needed: they
   # drive the C ABI from async Swift, which caught a stack overflow no Linux
   # job hit.
   #
@@ -158,7 +158,7 @@ jobs:
       # except the Windows x64 Node test leg, which also runs whenever the
       # shared Rust code changed: it is the only job that runs that code on
       # Windows, where I/O errors and refused connections behave differently
-      # (Swift contract tests cover macOS).
+      # (Swift tests cover macOS).
       - id: legs
         env:
           PLATFORMS: ${{ steps.diff.outputs.platforms }}
@@ -632,9 +632,9 @@ jobs:
           path: ${{ steps.flutter-key.outputs.CACHE-PATH }}
           key: ${{ steps.flutter-key.outputs.CACHE-KEY }}
 
-  # ── Contract tests: shared JSON fixtures (catches cross-language type drift) ──
+  # Generated files must match their generators.
   contract-tests:
-    name: Contract tests
+    name: Generator drift checks
     runs-on: ubuntu-latest
     needs: changes
     if: fromJSON(needs.changes.outputs.run)['contract-tests']
@@ -646,29 +646,20 @@ jobs:
       - uses: Swatinem/rust-cache@v2
         with:
           save-if: ${{ github.ref == 'refs/heads/master' }}
-      - uses: actions/setup-node@v4
-        with:
-          node-version: 22
-      - name: Run Rust contract tests
-        run: cargo test --test contract_test -p aimux-core
       - name: ts-rs export drift (checked-in types must match generator output)
         run: python3 scripts/gen_ts_types.py --check
-      - name: Build Node binding
-        working-directory: bindings/node
-        run: |
-          npm ci
-          npx napi build --platform
-      - name: Run Node contract tests
-        working-directory: contract-tests
-        run: |
-          node --experimental-strip-types run-node.ts
+      # The AI SDK pin tooling: VERSIONS.json format and integrity rules, and
+      # fetch-aisdk.sh against a registry served on 127.0.0.1 (no network).
+      - uses: actions/setup-node@v4
+        with:
+          node-version: 24
+      - name: AI SDK pin scripts (fetch-aisdk.sh, VERSIONS.json helpers)
+        run: node --test scripts/fetch-aisdk.test.mjs scripts/aisdk-fixtures/versions.test.mjs
 
-  # ── Swift contract tests: the eighth binding joins the shared fixtures ─────
-  # Needs macOS (XCTest) and the FFI cdylib, which the Swift package links via
-  # the CAimuxFFI system module. The Dart side needs no job here: flutter-binding
-  # already runs `flutter test`, which picks up bindings/flutter/test/contract_test.dart.
+  # Swift tests need macOS (XCTest) and the FFI cdylib, which the Swift package
+  # links via the CAimuxFFI system module.
   swift-contract-tests:
-    name: Swift contract tests
+    name: Swift tests
     runs-on: macos-latest
     needs: changes
     if: fromJSON(needs.changes.outputs.run)['swift-contract-tests']
@@ -680,7 +671,7 @@ jobs:
           save-if: ${{ github.ref == 'refs/heads/master' }}
       - name: Build the FFI library the Swift package links against
         run: cargo build -p aimux-ffi --release
-      - name: Run Swift contract tests
+      - name: Run Swift tests
         working-directory: bindings/swift
         env:
           DYLD_LIBRARY_PATH: ${{ github.workspace }}/target/release

CHANGELOG.md

@@ -48,6 +48,32 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
   no longer rewrites tracked TypeScript types. `scripts/gen_ts_types.py`
   enables exports and rejects ungated export attributes with `--check`.
 
+### Removed
+
+- The cross-language contract fixtures are retired (#249):
+  `contract-tests/fixtures/wire-format.json` (a record of Rust's own serde
+  output, not of the AI SDK), `tool-call-repair.json`, the Node runner
+  `contract-tests/run-node.ts`, `aimux-core/tests/contract_test.rs` and the
+  bindings' contract and repair tests are deleted. Wire behaviour is pinned
+  instead by fixtures recorded from the pinned AI SDK packages under
+  `fixtures/aisdk/` (versions and sha512 integrity in
+  `fixtures/aisdk/VERSIONS.json`, fetched by `scripts/fetch-aisdk.sh`) and by
+  tests ported from upstream. The two decision fixtures stay in
+  `contract-tests/fixtures/`.
+
+### RFCs
+
+- `docs/aisdk-architecture-alignment.md`: the V4 migration now lands on
+  master layer by layer, each layer CI-green with all its consumers updated,
+  instead of one atomic cutover through an integration branch (§0.4 Q3, D31,
+  §9.1). The new C ABI, codegen, the full recording schema 3 and the binding
+  migration come last; until then the existing C constructors keep their
+  signatures and call the new factories, and ts-rs stays as the interim
+  generator. No release is cut from master until every layer is merged;
+  fixes for released versions ship from a maintenance branch. `ROADMAP.md`
+  defers the ops protocol schema and L0 passthrough and no longer fixes C1
+  at nine exports.
+
 ## [0.5.0] - 2026-09-27
 
 **Breaking release.** 13 PRs since 0.3.0: the cross-language error model

CONTRIBUTING.md

@@ -13,7 +13,6 @@ aimux/
 ├── aimux-provider-utils/  # HTTP utilities: retry, backoff, error parsing, API-key loading
 ├── aimux-ffi/             # C ABI (opaque handle + JSON + push callback) for non-native bindings
 ├── bindings/              # Node, Python, Swift, Kotlin, Flutter, Go, C — share one Rust core
-├── contract-tests/        # Shared JSON fixtures exercised across languages
 ├── rfc/                   # Design docs (RFCs)
 ├── docs/                  # Public product docs (see docs/README.md)
 └── scripts/               # Code-generation and maintenance scripts
@@ -50,9 +49,6 @@ cargo test --workspace
 
 # A single crate
 cargo test -p aimux-providers --tests
-
-# Contract tests (cross-language shared fixtures)
-cargo test --test contract_test -p aimux-core
 ```
 
 If you add or change a provider's request/response shapes, record a new
@@ -108,8 +104,7 @@ Keep the subject line imperative and ≤ 72 characters.
 2. Make your change with focused commits.
 3. Ensure `cargo fmt`, `cargo clippy -D warnings`, and `cargo test --workspace`
    all pass locally. CI runs the same.
-4. If you add a provider or binding, include cassette tests or contract
-   fixtures as appropriate.
+4. If you add a provider or binding, include cassette tests.
 5. Update the relevant doc (README, `docs/`, or `rfc/`) if your change affects
    the public surface.
 6. Open a PR against `master` and fill in the template.

ROADMAP.md

@@ -62,8 +62,8 @@ aimux 是 **provider 接入与治理的运行时**：
 | #185 | `ToolInput{Raw,Parsed}`；tracker 不删，改为对齐 `@ai-sdk/provider-utils` 并移入 aimux-provider-utils（#204）；调整：见 [docs/aisdk-architecture-alignment.md §0.7](docs/aisdk-architecture-alignment.md#07-与-roadmap--rfc-0036-既有承诺的关系) |
 | #170 / #171 前半 | sync/probe 脚本 + scheduled Action；27 个 base_url 分歧 triage |
 | **B1 + C2** | 解锁项：protocol 列 + `from_resolved`；40 个 FFI 构造器转发 shim（B1 形式由 descriptor 数据源 + `create_xxx` 工厂替代，C2 shim 已取消；调整：见 [docs/aisdk-architecture-alignment.md §0.7](docs/aisdk-architecture-alignment.md#07-与-roadmap--rfc-0036-既有承诺的关系)） |
-| **ops 协议 schema** | op 表、消息与错误信封、二进制帧、版本协商——先以文档 + 测试落地（RFC-0036 §4） |
-| **L0 passthrough** | 原样调用任意 provider 端点，享受 auth / 重试 / 录制 |
+| **ops 协议 schema** | op 表、消息与错误信封、二进制帧、版本协商——先以文档 + 测试落地（RFC-0036 §4）；后置到全链路对齐切换之后，设计由 RFC-0039 承接（调整：见 [docs/aisdk-architecture-alignment.md §0.7](docs/aisdk-architecture-alignment.md#07-与-roadmap--rfc-0036-既有承诺的关系)） |
+| **L0 passthrough** | 原样调用任意 provider 端点，享受 auth / 重试 / 录制；后置到全链路对齐切换之后，作为新增能力单独评审（调整：见 [docs/aisdk-architecture-alignment.md §0.7](docs/aisdk-architecture-alignment.md#07-与-roadmap--rfc-0036-既有承诺的关系)） |
 
 版本语义：
 - **Rust 层 breaking**：A2 删枚举、#185 换类型。
@@ -86,7 +86,7 @@ aimux 是 **provider 接入与治理的运行时**：
 
 | 项 | 内容 |
 |---|---|
-| C1 | 9 个导出 + `dispatch`，**同一 dispatch 同时服务 FFI 与 stdio CLI**；旧导出保留共存 |
+| C1 | ops 协议入口 + `dispatch`，**同一 dispatch 同时服务 FFI 与 stdio CLI**；导出数量不再固定为 9 个，按对齐文档 §6.2–§6.3 的函数组设计；旧符号按切换门一次替换，不共存（调整：见 [docs/aisdk-architecture-alignment.md §0.7](docs/aisdk-architecture-alignment.md#07-与-roadmap--rfc-0036-既有承诺的关系)） |
 | C3 | `aimux_error_*` 访问器 → 错误 JSON 信封（含派生 `retry_after_ms`） |
 | D1–D7 | 8 种绑定迁到 ops 薄封装。D1 Kotlin（只依赖 Java artifact）与 D5 Flutter ffigen 过渡版不依赖 C1，可在 0.6/0.7 提前做 |
 | D8 | 类型镜像生成：serde → JSON Schema → 各语言，`--check` 门禁扩到全部输出 |
@@ -100,7 +100,7 @@ aimux 是 **provider 接入与治理的运行时**：
 |---|---|
 | #167 | transport-level replay：mock 挂 HTTP 层跑真协议代码，覆盖全协议 / 全模态；`ProviderRecord` = registry 行 + protocol |
 | 漂移检测 | #170 扩展：定期重录，与 cassette 字节级 diff |
-| 能力矩阵 | provider × 能力，每格由探测或录制支撑 |
+| 能力矩阵 | provider × 能力，每格由探测或录制支撑；后置到全链路对齐迁移之后（见 §6.1） |
 | CLI | `aimux probe / replay / diff`（#179 replay 子命令并入）；#181 debug CLI 并入 |
 
 ### 1.0 ——「冻结」（~1-2 周）
@@ -149,6 +149,8 @@ aimux 是 **provider 接入与治理的运行时**：
 
 ### 6.1 产物体积与性能（硬门禁）
 
+> 2026-10-13 调整：全链路对齐迁移按 [docs/aisdk-architecture-alignment.md §9.1](docs/aisdk-architecture-alignment.md#91-集成原则) 逐层合入主线，每个迁移层的 PR 描述列出产物改前改后的体积；0.9 能力矩阵排在迁移之后。2026-10-15 修订：体积与性能门禁的计量、预算与判定按 [RFC-0043](rfc/0043-size-performance-budgets.md)（#208），不以迁移完成为前提。
+
 | 产物 | v0.5.0 | 目标 | 手段 |
 |---|---|---|---|
 | 桌面 `.a` ×5 | 117–139 MB | ≤ 70 MB | S1（只改 staticlib job） |