feat(dx): Dev Container and CI benchmarks

.devcontainer/Dockerfile

@@ -0,0 +1,14 @@
+FROM ghcr.io/cirruslabs/flutter:3.41.6
+
+# Linux desktop toolchain + libs the app links against.
+RUN apt-get update \
+ && apt-get install -y --no-install-recommends \
+      clang cmake ninja-build pkg-config libgtk-3-dev libsecret-1-dev \
+      libsqlite3-dev libglu1-mesa xvfb git curl ca-certificates \
+ && rm -rf /var/lib/apt/lists/*
+
+# Docker CLI so the dev stack (docker/docker-compose.yml) can be started from inside the container.
+COPY --from=docker:27-cli /usr/local/bin/docker /usr/local/bin/docker
+COPY --from=docker:27-cli /usr/local/libexec/docker/cli-plugins/docker-compose /usr/local/libexec/docker/cli-plugins/docker-compose
+
+ENV DISPLAY=:99

.devcontainer/devcontainer.json

@@ -0,0 +1,23 @@
+{
+  "name": "Querya Desktop",
+  "build": { "dockerfile": "Dockerfile" },
+  "features": {
+    "ghcr.io/devcontainers/features/docker-outside-of-docker:1": {}
+  },
+  "postCreateCommand": "flutter config --enable-linux-desktop && flutter pub get",
+  "postStartCommand": "docker compose -f docker/docker-compose.yml up -d || echo 'Docker stack not started (run it manually: cd docker && docker compose up -d)'",
+  "forwardPorts": [5432, 3306, 6379, 27017, 8123, 9000],
+  "portsAttributes": {
+    "5432": { "label": "PostgreSQL" },
+    "3306": { "label": "MySQL" },
+    "6379": { "label": "Redis" },
+    "27017": { "label": "MongoDB" },
+    "8123": { "label": "ClickHouse HTTP" }
+  },
+  "customizations": {
+    "vscode": {
+      "extensions": ["Dart-Code.dart-code", "Dart-Code.flutter"]
+    }
+  },
+  "hostRequirements": { "cpus": 4, "memory": "8gb" }
+}

.github/workflows/benchmark.yml

@@ -0,0 +1,86 @@
+name: Benchmark
+
+# Informational: compares frame timings of benchmark/grid_perf_bench.dart on the
+# PR against its base branch and comments when the PR is >5% slower. Never
+# blocks a merge.
+
+on:
+  pull_request:
+    branches: [ dev, main ]
+    paths:
+      - 'lib/features/workspace/**'
+      - 'lib/core/database/**'
+      - 'lib/core/csv/**'
+      - 'benchmark/**'
+      - 'scripts/ci/bench_compare.py'
+      - '.github/workflows/benchmark.yml'
+  workflow_dispatch:
+
+permissions:
+  contents: read
+  pull-requests: write
+
+concurrency:
+  group: benchmark-${{ github.ref }}
+  cancel-in-progress: true
+
+jobs:
+  grid-bench:
+    name: Grid performance
+    # Forks get no write token and must never reach any self-hosted runner.
+    if: github.event_name != 'pull_request' || github.event.pull_request.head.repo.full_name == github.repository
+    runs-on: ubuntu-latest
+    timeout-minutes: 45
+    continue-on-error: true
+    steps:
+      - uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Install Linux dependencies
+        timeout-minutes: 8
+        run: |
+          sudo apt-get -o Acquire::Retries=3 update
+          sudo apt-get -o Acquire::Retries=3 install -y clang cmake ninja-build pkg-config libgtk-3-dev libsecret-1-dev libsqlite3-dev xvfb libgl1-mesa-dri
+
+      - uses: subosito/flutter-action@v2
+        with:
+          flutter-version: '3.41.6'
+          channel: 'stable'
+          cache: true
+
+      - name: Run benchmark on PR head
+        timeout-minutes: 15
+        run: |
+          flutter pub get
+          xvfb-run -a -s '-screen 0 1920x1080x24' flutter run --profile -d linux \
+            -t benchmark/grid_perf_bench.dart --dart-define=MODE=scroll \
+            2>&1 | tee "$RUNNER_TEMP/head.log" || true
+
+      - name: Run benchmark on base
+        if: github.event_name == 'pull_request'
+        timeout-minutes: 15
+        run: |
+          git worktree add "$RUNNER_TEMP/base" "origin/${{ github.base_ref }}"
+          rm -rf "$RUNNER_TEMP/base/benchmark"
+          cp -r benchmark "$RUNNER_TEMP/base/benchmark"
+          cd "$RUNNER_TEMP/base"
+          flutter pub get
+          xvfb-run -a -s '-screen 0 1920x1080x24' flutter run --profile -d linux \
+            -t benchmark/grid_perf_bench.dart --dart-define=MODE=scroll \
+            2>&1 | tee "$RUNNER_TEMP/base.log" || true
+
+      - name: Compare
+        id: compare
+        run: |
+          python3 scripts/ci/bench_compare.py "$RUNNER_TEMP/base.log" "$RUNNER_TEMP/head.log" 5 > "$RUNNER_TEMP/report.txt"
+          cat "$RUNNER_TEMP/report.txt"
+          echo "regression=$(head -1 "$RUNNER_TEMP/report.txt" | cut -d= -f2)" >> "$GITHUB_OUTPUT"
+          tail -n +2 "$RUNNER_TEMP/report.txt" > "$RUNNER_TEMP/report.md"
+          cat "$RUNNER_TEMP/report.md" >> "$GITHUB_STEP_SUMMARY"
+
+      - name: Comment on regression
+        if: github.event_name == 'pull_request' && steps.compare.outputs.regression == '1'
+        env:
+          GH_TOKEN: ${{ github.token }}
+        run: gh pr comment "${{ github.event.pull_request.number }}" --body-file "$RUNNER_TEMP/report.md"

docs/ci-runners.md

@@ -67,3 +67,14 @@ pinned in one place (`ci.yml`).
 - Keep the machine updated (`apt upgrade`) and clean old workspaces under
   `/home/runner/actions-runner-*/_work` if the disk fills.
 - The runners hold no secrets of ours; do not add any to the container.
+
+## Dev Container and benchmarks
+
+`.devcontainer/` provides a ready environment (Flutter 3.41.6, Linux desktop
+toolchain, Docker CLI) for GitHub Codespaces or VS Code Dev Containers. On start
+it runs `docker/docker-compose.yml` (PostgreSQL, MySQL, Redis, MongoDB,
+ClickHouse) and forwards their ports.
+
+`.github/workflows/benchmark.yml` runs `benchmark/grid_perf_bench.dart` under
+Xvfb on the PR and on its base branch and comments when frame times are more
+than 5% worse. It is informational and never blocks a merge.

scripts/ci/bench_compare.py

@@ -0,0 +1,57 @@
+#!/usr/bin/env python3
+"""Compare two grid_perf_bench logs and print a Markdown report.
+
+usage: bench_compare.py BASE.log HEAD.log [threshold_percent]
+
+Reads the `BENCH total  p50=.. p90=.. p99=..` line of each log. Always exits 0:
+the report is informational. The first output line is `REGRESSION=1` or
+`REGRESSION=0` so a workflow can decide whether to comment.
+"""
+import re
+import sys
+
+
+def parse(path):
+    metrics = {}
+    try:
+        text = open(path, encoding="utf-8", errors="replace").read()
+    except OSError:
+        return metrics
+    m = re.search(r"BENCH total\s+p50=([\d.]+)\s+p90=([\d.]+)\s+p99=([\d.]+)", text)
+    if m:
+        metrics = {"p50": float(m[1]), "p90": float(m[2]), "p99": float(m[3])}
+    s = re.search(r"stutters\(>12\.5ms\)=(\d+)", text)
+    if s:
+        metrics["stutters"] = float(s[1])
+    return metrics
+
+
+def main():
+    base, head = parse(sys.argv[1]), parse(sys.argv[2])
+    threshold = float(sys.argv[3]) if len(sys.argv) > 3 else 5.0
+    if not base or not head:
+        print("REGRESSION=0")
+        print("Benchmark produced no comparable output (base or head log has no `BENCH total` line).")
+        return
+    rows, regressed = [], False
+    for key in ("p50", "p90", "p99", "stutters"):
+        if key not in base or key not in head:
+            continue
+        b, h = base[key], head[key]
+        delta = (h - b) / b * 100 if b else (0.0 if h == 0 else 100.0)
+        # Tiny absolute differences are noise on a shared runner.
+        worse = delta > threshold and (h - b) > (0.3 if key != "stutters" else 2)
+        regressed = regressed or worse
+        unit = "" if key == "stutters" else " ms"
+        rows.append(f"| {key} | {b:.2f}{unit} | {h:.2f}{unit} | {delta:+.1f}% {'⚠️' if worse else ''} |")
+    print(f"REGRESSION={1 if regressed else 0}")
+    print("### Grid scroll benchmark")
+    print()
+    print("| metric | base | PR | change |")
+    print("|---|---|---|---|")
+    print("\n".join(rows))
+    print()
+    print(f"Informational only (threshold {threshold:g}%). Shared CI runners are noisy; re-run before trusting a single result.")
+
+
+main()