feat(docker-git): add resource limits for MCP Playwright sidecar

[konard]

Summary

Adds configurable CPU and RAM limits for the MCP Playwright browser sidecar container, separate from the main service container's limits. This addresses #259, which reported that the Playwright sidecar could consume too many host resources.

• New CLI flags: --playwright-cpu (alias --playwright-cpus) and --playwright-ram (alias --playwright-memory).
• Defaults to 30% of the host CPU/RAM, falling back to the main service limits when not set.
• Limits are wired through:
  • CLI parser (parser-options, parser-apply)
  • Template defaults (template-defaults)
  • Apply overrides (apply-overrides)
  • Resource normalization & resolution (resource-limits)
  • Docker Compose template (templates/docker-compose)
• Accepts the same value formats as the existing --cpu/--ram flags: percent (30%), absolute CPU (1.5), and absolute RAM (512m, 4g, etc.).

Reproduction

Before this change, the MCP Playwright sidecar inherited only the main --cpu/--ram limits (or none), so Chromium could consume a large share of host resources during long browsing sessions.

After this change:

docker-git create my-project \
  --cpu 50% --ram 50% \
  --playwright-cpu 20% --playwright-ram 1500m

produces a docker-compose.yml where:

• service has cpus: <50% of host> and mem_limit/memswap_limit: <50% of host>
• playwright has cpus: <20% of host> and mem_limit/memswap_limit: 1500m

When --playwright-* flags are omitted, the sidecar falls back to the main service limits, and when those are also omitted, it uses the 30% default.

Implementation notes

• The codebase keeps three synchronized copies (canonical packages/lib, mirror at packages/app/src/lib, and frontend subset at packages/app/src/docker-git/frontend-lib); all three were updated together.
• Resource normalization moved into resource-limits.ts (new resolveResourceLimitsIntent helper) to keep command-builders.ts under the 300-line ESLint cap.
• apply-overrides.ts was split into applyTokenOverrides + applyResourceOverrides to satisfy the 50-line-per-function cap.

Test plan

• bun run typecheck passes (exit 0)
• bun run lint passes (exit 0)
• bun run test passes — 273 app tests + 173 lib tests
• New unit tests:
  • packages/lib/tests/core/resource-limits.test.ts — defaults & explicit overrides for playwright limits
  • packages/lib/tests/core/templates.test.ts — docker-compose template emits separate sidecar limits
  • packages/app/tests/docker-git/parser-playwright-resource.test.ts — CLI parser accepts new flags and aliases
• Existing CPU/RAM behavior preserved (no regressions in the resolved-limits or template tests)

Changeset

• .changeset/playwright-resource-limits.md (minor bump for @prover-coder-ai/docker-git and @effect-template/lib)

Closes #259

[konard]

Working session summary

All 15 CI checks pass (Build, Types, Lint, Test, Lint Effect-TS, Snapshot, dist-deps-prune × 2, plus 7 E2E suites).

PR #264 is finalized: #264

• Title: feat(docker-git): add resource limits for MCP Playwright sidecar
• Status: Ready for review (no longer draft), all 15 checks green
• Adds --playwright-cpu/--playwright-cpus and --playwright-ram/--playwright-memory flags with a 30% default, falling back to the main service limits when unset
• Synchronized across all three lib copies (canonical, app/lib mirror, frontend-lib)
• Ships with a changeset entry, new parser tests, plus updated resource-limits and templates tests; 273 app + 173 lib tests pass locally and on CI

---

This summary was automatically extracted from the AI working session output.

[konard]

🤖 Solution Draft Log

This log file contains the complete execution trace of the AI solution draft process.

💰 Cost: $17.724580

📊 Context and tokens usage:

Claude Opus 4.7: (5 sub-sessions)

1. 116.4K / 1M (12%) input tokens, 26.2K / 128K (21%) output tokens
2. 116K / 1M (12%) input tokens, 17.6K / 128K (14%) output tokens
3. 117.2K / 1M (12%) input tokens, 25.6K / 128K (20%) output tokens
4. 116.7K / 1M (12%) input tokens, 31.6K / 128K (25%) output tokens
5. 59.2K / 1M (6%) input tokens, 6.2K / 128K (5%) output tokens

Total: (13.3K new + 457.8K cache writes + 23.0M cache reads) input tokens, 132.2K output tokens, $17.724580 cost

🤖 Models used:

• Tool: Anthropic Claude Code
• Requested: opus
• Model: Claude Opus 4.7 (claude-opus-4-7)

📎 Log file uploaded as Gist (7737KB)

• View complete solution draft log

---

Now working session is ended, feel free to review and add any feedback on the solution draft.

[konard]

✅ Ready to merge

This pull request is now ready to be merged:

• All CI checks have passed
• No merge conflicts
• No pending changes

---

Monitored by hive-mind with --auto-restart-until-mergeable flag