@@ -26,13 +26,19 @@ export default defineConfig((ctx) => ({
         path: "apps/web/src/**/*.{tsx,css}",
         instructions: "Hold changed code to the rules in docs/internals/web-ui.md.",
       },
+      {
+        path: "**/*.test.{ts,tsx}",
+        instructions:
+          "Hold changed tests to the rules in docs/internals/testing.md. Request tests only as those rules allow.",
+      },
     ],
   },
   knowledge_base: {
     code_guidelines: {
       filePatterns: [
         { files: "docs/internals/effect-services.md", applyTo: "{apps,packages,infra}/**/*.ts" },
         { files: "docs/internals/web-ui.md", applyTo: "apps/web/src/**/*.{tsx,css}" },
+        { files: "docs/internals/testing.md", applyTo: "{apps,packages,infra}/**/*.{ts,tsx}" },
       ],
     },
   },
 
@@ -0,0 +1,55 @@
+name: Agent Context Budget
+
+# Fails when a PR raises a value in agentContextBudget.json above main's. AgentContextBudget.test.ts
+# keeps that file equal to the measured context, so the context cannot grow by editing the file.
+# To accept growth on purpose, a maintainer adds the agent-context-increase label.
+on:
+  pull_request:
+    types: [opened, synchronize, reopened, labeled, unlabeled]
+
+permissions:
+  contents: read
+
+concurrency:
+  group: agent-context-budget-${{ github.event.pull_request.number }}
+  cancel-in-progress: true
+
+jobs:
+  budget:
+    name: Agent context budget
+    runs-on: ubuntu-24.04
+    timeout-minutes: 5
+    env:
+      BUDGET: apps/server/src/provider/agentContextBudget.json
+      ALLOW_INCREASE: ${{ contains(github.event.pull_request.labels.*.name, 'agent-context-increase') }}
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v6
+        with:
+          sparse-checkout: apps/server/src/provider/agentContextBudget.json
+          sparse-checkout-cone-mode: false
+
+      - name: Compare with main
+        run: |
+          git fetch --quiet --depth=1 origin main
+          if ! git cat-file -e "FETCH_HEAD:$BUDGET" 2>/dev/null; then
+            echo "main has no agent context budget yet."
+            exit 0
+          fi
+          git show "FETCH_HEAD:$BUDGET" > "$RUNNER_TEMP/main.json"
+          grown=$(jq -r --slurpfile main "$RUNNER_TEMP/main.json" '
+            to_entries[]
+            | select($main[0][.key] != null and .value > $main[0][.key])
+            | "\(.key): \($main[0][.key]) -> \(.value) (+\(.value - $main[0][.key]))"
+          ' "$BUDGET")
+          if [ -z "$grown" ]; then
+            echo "Agent context is not larger than main."
+            exit 0
+          fi
+          printf 'Agent context grew compared to main:\n%s\n' "$grown"
+          if [ "$ALLOW_INCREASE" = "true" ]; then
+            echo "Accepted by the agent-context-increase label."
+            exit 0
+          fi
+          echo "::error file=$BUDGET::Agent context grew compared to main. Cut the added text, or ask a maintainer to add the agent-context-increase label."
+          exit 1
@@ -344,6 +344,12 @@ jobs:
       - name: Test
         run: vp run --filter @t3tools/web test
 
+      # Browser tests run source modules; this boots the production chunk graph.
+      - name: Boot the built web client
+        run: |
+          vp run --filter @t3tools/web build
+          node apps/web/scripts/verify-built-boot.ts
+
   # apps/server sets `fileParallelism: false`, so its test files run strictly
   # one at a time. Sharding spreads them over separate runners instead of
   # separate workers, so no two server test files ever share a machine and the
 
@@ -5,12 +5,11 @@ on:
     branches:
       - main
   # Redeploys main after a variable or secret change. Alchemy redeploys the
-  # Worker when a value it reads changes, so force is rarely needed, and it
-  # replaces the Postgres runtime role (alchemy-run/alchemy#1832).
+  # Worker when a value it reads changes, so force is rarely needed.
   workflow_dispatch:
     inputs:
       force:
-        description: Redeploy every resource, including ones with no detected changes. Replaces the Postgres runtime role.
+        description: Redeploy every resource, including ones with no detected changes.
         type: boolean
         default: false
 
 
@@ -275,6 +275,7 @@ jobs:
             --environment production \
             --platform ${{ inputs.platform }} \
             --message "$OTA_MESSAGE" \
+            --emit-metadata \
             --non-interactive
 
       # Reconcile each version separately because the major is in the fingerprint.
@@ -313,8 +314,11 @@ jobs:
                 --environment production \
                 --platform "$platform" \
                 --message "$message" \
+                --emit-metadata \
                 --non-interactive; then
                 echo ":white_check_mark: $platform/$profile: OTA published to production (fingerprint \`$hash\`)" >> "$GITHUB_STEP_SUMMARY"
+                # The next target's export replaces dist; keep this one for the map upload.
+                rm -rf "dist-$platform-$profile" && mv dist "dist-$platform-$profile"
               else
                 failed=1
                 echo ":x: $platform/$profile: OTA publish failed (fingerprint \`$hash\`)" >> "$GITHUB_STEP_SUMMARY"
@@ -354,3 +358,18 @@ jobs:
             fi
           done
           exit "$failed"
+
+      # Lets the relay symbolicate crash stacks from the updates published
+      # above. A step of its own so the R2 credentials reach only this script,
+      # not `eas update` and the dependencies it runs. Never fails the job.
+      - name: Upload OTA source maps
+        if: ${{ !cancelled() && steps.expo-token.outputs.present == 'true' }}
+        working-directory: apps/mobile
+        env:
+          MOBILE_SOURCEMAPS_R2: ${{ secrets.MOBILE_SOURCEMAPS_R2 }}
+        run: |
+          for dir in dist dist-*; do
+            if [ -f "$dir/eas-update-metadata.json" ]; then
+              node scripts/upload-sourcemaps.mts ota "$dir"
+            fi
+          done
@@ -64,6 +64,12 @@ on:
         required: false
         default: false
         type: boolean
+      # Whether the job opens the packaged app against a seeded database and
+      # opens a thread before uploading it. Linux AppImage or macOS DMG.
+      launch_check:
+        required: false
+        default: false
+        type: boolean
       # A runner of the target's own platform and architecture, for a `runner`
       # that cannot execute the target. The build job skips everything that
       # runs the packaged binaries, and a second job runs it there.
@@ -517,6 +523,38 @@ jobs:
 
           vp run dist:desktop:artifact "${args[@]}"
 
+      # A build can package fine and still not open (a missing module in the
+      # backend or the bundle). This launches the app from the shipped
+      # AppImage or DMG.
+      - name: Launch check
+        if: inputs.launch_check
+        shell: bash
+        run: |
+          set -euo pipefail
+          check=(node scripts/smoke-desktop-app.ts --artifacts-dir "$RUNNER_TEMP/desktop-launch-check")
+          if [[ "${{ inputs.platform }}" == "mac" ]]; then
+            mount="$RUNNER_TEMP/desktop-launch-dmg"
+            hdiutil attach -nobrowse -readonly -mountpoint "$mount" "$(ls "$GITHUB_WORKSPACE"/release/*.dmg)"
+            trap 'hdiutil detach "$mount" -force >/dev/null || true' EXIT
+            app=$(ls -d "$mount"/*.app)
+            executable=$(/usr/libexec/PlistBuddy -c "Print :CFBundleExecutable" "$app/Contents/Info.plist")
+            "${check[@]}" --executable "$app/Contents/MacOS/$executable"
+          else
+            sudo apt-get install -y xvfb
+            appimage=$(ls "$GITHUB_WORKSPACE"/release/*.AppImage)
+            chmod +x "$appimage"
+            (cd "$RUNNER_TEMP" && "$appimage" --appimage-extract >/dev/null)
+            xvfb-run -a "${check[@]}" --executable "$RUNNER_TEMP/squashfs-root/t3code"
+          fi
+
+      - name: Upload launch check output
+        if: failure() && inputs.launch_check
+        uses: actions/upload-artifact@043fb46d1a93c77aae656e7c1c64a875d1fc6a0a # v7
+        with:
+          name: desktop-launch-check-${{ inputs.platform }}-${{ inputs.arch }}
+          path: ${{ runner.temp }}/desktop-launch-check
+          if-no-files-found: ignore
+
       # The verify job runs what this host could not: the packaged natives and
       # the CLI. Its scripts are bundled so that job needs no install.
       - name: Stage verification inputs
 
@@ -344,6 +344,12 @@ jobs:
       - name: Test
         run: vp run --filter @t3tools/web test
 
+      # Browser tests run source modules; this boots the production chunk graph.
+      - name: Boot the built web client
+        run: |
+          vp run --filter @t3tools/web build
+          node apps/web/scripts/verify-built-boot.ts
+
   # Server test files run one at a time, so more small runners beat fewer big ones.
   test_server:
     name: Release server tests ${{ matrix.shard }}
@@ -613,6 +619,7 @@ jobs:
       rust_target: aarch64-apple-darwin
       resource_key: darwin-arm64
       cli_archive: true
+      launch_check: true
 
   desktop_mac_x64:
     name: Desktop macOS x64
@@ -662,6 +669,7 @@ jobs:
       rust_target: x86_64-unknown-linux-gnu
       resource_key: linux-x64
       cli_archive: true
+      launch_check: true
 
   # node-pty has no Linux prebuild and compiles from source, so the arm64 app
   # and archive are built on arm64 hardware rather than cross-built.
 
@@ -103,6 +103,7 @@ An empty database is a bad test. Seed your worktree's `.t3` with a copy of real
 - Use `vp run typecheck`, not `npx tsc`; it caches per package. Both run tsc-rs (`typescript` is the `@tsc-rs/typescript` drop-in).
 - Run `node -e` snippets from `apps/server` or the owning package. `effect` does not resolve from the repo root.
 - Test meaningful logic or observable behavior. Do not render components to static markup to assert props or attributes, or add tests that merely assert callback wiring or mirror the implementation.
+- [Testing](docs/internals/testing.md) carries these same test rules, plus review guidance, for review bots. Change both together.
 - **Do not run repo-wide checks.** No `vp check`, no `vp run -r test`, no `vp run -r typecheck` unless I ask. CI owns the full suite.
 - Backend behavior changes ship with focused tests for that behavior.
 - Proof for a bug fix depends on the bug. Logic, data, state, and protocol bugs get a failing test first (the fewest that fail on `main` for that reason, in the module's existing test file), then the fix. Purely visual bugs (spacing, colour, alignment, a layout shift) get before/after captures as their proof and no test: do not add UI tests for nitpicks, and do not ask for one in review when a capture already shows it. A server-only fix with no visible surface ships the test plus a short runtime transcript.
 
@@ -1,14 +1,13 @@
 {
   "name": "@t3tools/desktop",
-  "version": "0.1.2",
+  "version": "0.1.3",
   "private": true,
   "type": "module",
   "main": "dist-electron/boot.cjs",
   "scripts": {
     "ensure:electron": "node scripts/ensure-electron-runtime.mjs",
     "start": "node scripts/start-electron.mjs",
-    "test": "vp test run --passWithNoTests",
-    "smoke-test": "node scripts/smoke-test.mjs"
+    "test": "vp test run --passWithNoTests"
   },
   "dependencies": {
     "@clerk/electron": "catalog:",