TencentCloud/CubeSandbox

Who is using Cube Sandbox First of all, thank you for using and supporting Cube Sandbox. Knowing how the community actually uses it helps us make it better — and helps others discover what they can build with it. The purpose of this issue We'd like to hear real-world scenarios to shape our roadmap. We want to better support teams building on Cube Sandbox. We're happy to see Cube Sandbox grow together with its users. What we expect from you If you are using (or evaluating) Cube Sandbox, please leave a comment below with the following information: Your company, school or organization (individual users are welcome too). Your city and country. Your contact info: email, WeChat or X/Twitter (at least one, optional). What scenario you use Cube Sandbox for (e.g., coding agents, RL rollout, browser automation, data analysis...). You can refer to the following sample answer for the format: Organization: Tencent Location: Shenzhen, China Contact: [email protected] Purpose/Scenario: Sandbox ru

Problem / Motivation MySQL 表中的数据使用软删除，且无清理机制，数据量会一直增大 Proposed Solution 增加定时清理机制 Alternatives Considered NA Additional Context NA

Problem / Motivation Background All template building, distribution, snapshot management, and compatibility checking logic is currently embedded within a single package (~13,400 lines of Go) in CubeMaster, running as part of the CubeMaster process. At startup, CubeMaster initializes the template module in-process, registers HTTP routes, and launches background goroutines (Snapshot Reconciler, Artifact GC, compatibility scanning, etc.). Pain Points 1. Tight Coupling — Cannot Scale or Upgrade Independently Template building involves CPU/IO-intensive operations (OCI image pulling, ext4 filesystem creation) that compete for resources with CubeMaster's latency-sensitive paths (scheduling, sandbox management). Background tasks run as in-process goroutines with no coordination mechanism across multiple CubeMaster replicas. Any changes to template building logic require a full CubeMaster release — no independent iteration. 2. Concurrency Safety in Multi-Replica Deployments (PR #194) PR #

What this issue is about Cube Sandbox doesn't have an Integrations section yet. We'd love your help building one — so developers around the world can plug Cube into the AI agent frameworks they already use. 💡 Cube is compatible with the E2B SDK — in most cases, switching from E2B to Cube is just a single env var change, while gaining MicroVM-level isolation. That makes each integration guide pretty light to write. What you can contribute An integration guide for one AI agent framework. Candidate frameworks (pick one — first come, first served; you're also welcome to propose frameworks not listed here): Framework Type LangChain Power your agent with Cube LlamaIndex Power your agent with Cube Dify Power your agent with Cube AutoGen Power your agent with Cube CrewAI Power your agent with Cube Google ADK Power your agent with Cube Claude Code Run your agent inside Cube OpenClaw Run your agent inside Cube OpenHands Run your agent inside Cube Others Ea

What this issue is about Cube Sandbox doesn't have a Use Cases gallery yet. We'd love your help building one — tell us what you've built with Cube so other developers can draw inspiration from your real-world experience. What you can contribute Any solution or case study that uses Cube in a real business scenario. Examples: a sandboxed runtime for AI coding agents, browser automation, code execution sandboxes, SWE-bench evaluation, data analysis notebooks, multi-agent collaboration environments — anything that shows Cube doing real work. Each case study just needs to cover four things: The problem you're solving (business context, in 1–2 sentences) Your approach (an architecture diagram and the key components) Key code or config (just enough to convey the design; doesn't need to be fully runnable) Lessons learned (why you chose Cube, what trade-offs you made) Where to submit a PR Create a new standalone markdown file for your case study — one file per case. The site will auto