fix(docker): persistent container + exec pattern

[KailasMahavarkar]

Summary

• Switches install.md Option A from docker run --rm per session → persistent hyperstack-mcp container + docker exec per session
• Adds stdin close / SIGTERM handling to src/index.ts so bun exits cleanly when the client disconnects
• Bumps version to 1.0.1 to trigger image rebuild with the fix
• Adds migration troubleshooting: how to clean up legacy containers

Why

Previous pattern spawned a new container for every claude invocation. Combined with missing stdin lifecycle handling, orphaned containers piled up over multiple sessions.

New pattern

# One-time setup
docker run -d --name hyperstack-mcp --restart unless-stopped \
  --memory=512m --cpus=1 \
  --entrypoint sleep \
  ghcr.io/orkait/hyperstack:main infinity

{
  "mcpServers": {
    "hyperstack": {
      "command": "docker",
      "args": ["exec", "-i", "hyperstack-mcp", "bun", "/app/src/index.ts"]
    }
  }
}

One container. All sessions exec into it. Zero per-invocation container cost.

Validated

• Tested docker exec -i hyperstack-mcp bun /app/src/index.ts locally → MCP init response returned correctly
• stdin close handler prevents orphan bun processes inside the persistent container

🤖 Generated with Claude Code