fix(function): isolated-vm worker pool to prevent single-worker bottleneck

[waleedlatif1]

Summary

• Replaced single isolated-vm worker process with a configurable pool (default 4 workers)
• Executions are distributed across workers using least-loaded selection
• Workers spawn lazily and clean up after idle timeout
• Added env vars for pool tuning (IVM_POOL_SIZE, IVM_MAX_CONCURRENT, IVM_MAX_PER_WORKER, IVM_WORKER_IDLE_TIMEOUT_MS, IVM_QUEUE_TIMEOUT_MS)
• Defaults are permissive (10k concurrent, 2500/worker, 5min queue timeout) — no behavior change for existing users

Type of Change

• Bug fix

Testing

Tested manually. Existing vitest suite (30 tests) passes unchanged.

Checklist

• Code follows project style guidelines
• Self-reviewed my changes
• Tests added/updated and passing
• No new warnings introduced
• I confirm that I have read and agree to the terms outlined in the Contributor License Agreement (CLA)

[vercel]

The latest updates on your projects. Learn more about Vercel for GitHub.

Project	Deployment	Actions	Updated (UTC)
docs	Ready	Preview, Comment	Feb 6, 2026 7:42pm

[waleedlatif1]

@cursor review

[greptile-apps]

Greptile Overview

Greptile Summary

• Introduces a lazy-spawned isolated-vm worker pool (default 4) and distributes executions by least-loaded worker instead of a single subprocess.
• Adds global/per-worker concurrency caps plus queueing with a configurable queue timeout and worker idle cleanup.
• Updates env config to expose pool tuning knobs (IVM_POOL_SIZE, IVM_MAX_CONCURRENT, IVM_MAX_PER_WORKER, IVM_WORKER_IDLE_TIMEOUT_MS, IVM_QUEUE_TIMEOUT_MS).
• Main risk area is correctness of pool accounting (startup/exit paths and active execution counters) which impacts saturation decisions and queue draining.

Confidence Score: 3/5

• Mergeable after fixing pool accounting bugs that can break worker capacity tracking.
• The worker-pool approach is straightforward and mostly self-contained, but there are concrete counter/accounting issues (startup failure double-decrement and per-worker activeExecution drift during cleanup) that can miscompute pool capacity and queue draining behavior under failures.
• apps/sim/lib/execution/isolated-vm.ts

Important Files Changed

Filename	Overview
apps/sim/lib/core/config/env.ts	Adds env vars for isolated-vm worker pool sizing and queue/idle timeouts; no functional logic changes here.
apps/sim/lib/execution/isolated-vm.ts	Replaces single isolated-vm worker with a lazy worker pool plus global/per-worker concurrency limits and queueing; found counter/accounting bugs in worker startup failure handling and active execution tracking.

Sequence Diagram

sequenceDiagram
  participant Caller as Caller
  participant Pool as isolated-vm.ts
  participant Worker as Child worker
  participant VM as isolated-vm-worker.cjs

  Caller->>Pool: "executeInIsolatedVM(req)"
  alt "totalActiveExecutions >= MAX_CONCURRENT"
    Pool->>Pool: "enqueueExecution(req)"
  else "capacity available"
    Pool->>Pool: "acquireWorker()"
    alt "existing worker found"
      Pool->>Worker: "dispatchToWorker(req)"
    else "spawn new worker"
      Pool->>Worker: "spawnWorker()"
      Worker->>VM: "spawn node isolated-vm-worker.cjs"
      VM-->>Worker: "IPC ready"
      Pool->>Worker: "dispatchToWorker(req)"
    else "all saturated"
      Pool->>Pool: "enqueueExecution(req)"
    end
  end

  Worker->>VM: "IPC execute(executionId, request)"
  opt "VM requests fetch"
    VM-->>Pool: "IPC fetch(fetchId, url, optionsJson)"
    Pool->>Pool: "secureFetch(validateProxyUrl + fetch)"
    Pool-->>VM: "IPC fetchResponse(fetchId, response)"
  end

  VM-->>Pool: "IPC result(executionId, result)"
  Pool->>Pool: "decrement counts + reset idle timeout"
  Pool->>Pool: "drainQueue()"
  opt "worker idle"
    Pool->>Worker: "cleanupWorker() after idle timeout"
  end

Loading

[greptile-apps]

_{2 files reviewed, 2 comments}

_{Edit Code Review Agent Settings | Greptile}

[cursor]

Cursor Bugbot has reviewed your changes and found 2 potential issues.

^{Bugbot Autofix is OFF. To automatically fix reported issues with Cloud Agents, enable Autofix in the Cursor dashboard.}