test(eval): exercise GitHub Actions reporter

.github/workflows/github-actions-reporter-example.yml

@@ -0,0 +1,52 @@
+name: GitHub Actions reporter example
+
+on:
+  pull_request:
+  workflow_dispatch:
+
+permissions:
+  contents: read
+
+jobs:
+  reporter-example:
+    runs-on: ubuntu-latest
+    env:
+      CARGO_NET_GIT_FETCH_WITH_CLI: true
+    steps:
+      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
+      - name: Configure git auth for private dependencies
+        run: |
+          token="${BT_GITHUB_TOKEN:-$GITHUB_TOKEN}"
+          git config --global url."https://x-access-token:${token}@github.com/".insteadOf "https://github.com/"
+        env:
+          BT_GITHUB_TOKEN: ${{ secrets.BT_GITHUB_TOKEN }}
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+      - name: Install Rust
+        uses: dtolnay/rust-toolchain@efa25f7f19611383d5b0ccf2d1c8914531636bf9 # master
+        with:
+          toolchain: 1.92.0
+      - uses: actions/setup-node@49933ea5288caeca8642d1e84afbd3f7d6820020 # v4.4.0
+        with:
+          node-version: "22"
+      - name: Enable pnpm
+        run: |
+          corepack enable
+          corepack prepare pnpm@10.28.2 --activate
+      - name: Install example dependencies
+        run: pnpm install --dir examples/eval-reporters --frozen-lockfile
+      - name: Build bt
+        run: cargo build --bin bt
+      - name: Exercise the GitHub Actions reporter
+        shell: bash
+        run: |
+          set +e
+          output=$(cd examples/eval-reporters && ../../target/debug/bt eval \
+            --no-send-logs \
+            --reporter=github-actions \
+            github-actions.eval.ts 2>&1)
+          status=$?
+          set -e
+
+          printf '%s\n' "$output"
+          grep -Eq '^::error title=case [0-9]+ errored,file=.*examples/eval-reporters/github-actions\.eval\.ts,line=22,col=13::intentional error from the GitHub Actions reporter example$' <<<"$output"
+          test "$status" -ne 0

examples/eval-reporters/README.md

@@ -0,0 +1,16 @@
+# Eval reporter examples
+
+## GitHub Actions
+
+Install this example's dependencies, build `bt`, and run:
+
+```bash
+pnpm install --dir examples/eval-reporters
+cargo build --bin bt
+cd examples/eval-reporters
+../../target/debug/bt eval --no-send-logs \
+  --reporter=github-actions \
+  github-actions.eval.ts
+```
+
+The second case intentionally throws. In GitHub Actions, the reporter emits an `::error` workflow command so the case appears as an annotation. The command exits non-zero because the eval contains an errored case; this is expected for this example.

examples/eval-reporters/github-actions.eval.ts

@@ -0,0 +1,32 @@
+import { Eval } from "braintrust";
+
+type Input = {
+  message: string;
+  shouldError: boolean;
+};
+
+Eval("reporter-example", {
+  evalName: "github-actions-reporter-example",
+  data: (): Array<{ input: Input; expected: string }> => [
+    {
+      input: { message: "successful case", shouldError: false },
+      expected: "successful case",
+    },
+    {
+      input: { message: "annotated case", shouldError: true },
+      expected: "annotated case",
+    },
+  ],
+  task: ({ message, shouldError }: Input): string => {
+    if (shouldError) {
+      throw new Error(
+        "intentional error from the GitHub Actions reporter example",
+      );
+    }
+    return message;
+  },
+  scores: [
+    ({ output, expected }: { output: string; expected: string }): number =>
+      output === expected ? 1 : 0,
+  ],
+});

examples/eval-reporters/package.json

@@ -0,0 +1,11 @@
+{
+  "name": "bt-eval-reporter-examples",
+  "private": true,
+  "type": "module",
+  "dependencies": {
+    "braintrust": "2.2.2"
+  },
+  "devDependencies": {
+    "tsx": "4.23.0"
+  }
+}