Most Sero OSS alpha setup problems fall into one of these buckets:
pnpm install finished, but terminals or memory features still failThe repo already runs repair hooks during pnpm install for the two main native
module risk areas:
node-ptybetter-sqlite3If install was interrupted, scripts were skipped, or Electron ABI drift still shows up at runtime, rerun the relevant repair from the repo root.
node-pty issuesIf terminal sessions still fail, use the manual rebuild flow from the dedicated node-pty guide.
better-sqlite3 issuesThis is the right next step if Electron logs mention ERR_DLOPEN_FAILED,
NODE_MODULE_VERSION, or better-sqlite3.
pnpm dev or bash scripts/dev.sh does not start cleanlyFrom the repo root, the canonical first-run command is:
If you are working directly inside the desktop app package, the equivalent is:
If startup looks stuck or stale, clean up old processes and retry:
Useful runtime logs:
/tmp/sero-vite.log/tmp/sero-electron.log/tmp/sero-web-remote-watch.log/tmp/sero-remote-<plugin>.log for any plugin dev remotes you enabledSero works best with Apple's container CLI available at:
Quick checks:
If the system is installed but not running:
If containers still are not available, Sero can continue in host mode. That is supported, but it is intentionally a reduced experience.
Host mode is a supported fallback for:
Host mode is not the supported path for:
If you hit one of those gaps, check whether the workspace should be running in container-backed mode instead.
See Support Scope for the canonical matrix, Containers and Host Mode for runtime-specific guidance, and Containers and Dev Servers for the dev-server quick path.
Check the server binding first. Many frameworks default to localhost, which can make the server reachable from inside the container shell but not from Sero's preview URL. Prefer binding to all interfaces:
Then register or re-register the server:
Use the URL reported by sero devserver list. In container-backed workspaces, that URL may use the container IP rather than localhost.
Container-backed dev servers reduce host port conflicts because the server listens inside the workspace container. Two workspaces can often use the same container port without both claiming the same host port.
This is not a guarantee that every network issue is solved. If the preview still fails, confirm container health, server binding, proxy/DNS behavior, and the URL reported by the registry.
The dev-server registry is in memory and URLs can change after container restart/recreation. Run:
If the listed URL differs from the one in your browser or preview tab, open the fresh URL or register the server again. If no servers are listed, start and register the dev server again.
sero browser ... commands operate on visible in-app browser tabs. sero app ... screenshot, interaction, preview, and recording commands use the UI-backed app-control bridge. Both depend on the relevant panel being loaded and visible.
Quick recovery:
If screenshots say the app panel is not found or not visible, switch to the target app and retry after it renders. See Browser and Capture.
Checkpoint and undo operations can restore files, and turn undo can also rewind the chat/session tree. If recovery does not behave as expected:
See Checkpoints and Undo for the recovery matrix.
From the repo root, gather the smallest helpful signal first:
Use pnpm test:ci when you need the current alpha PR-gate shape.
When reporting the problem, include:
Before sharing logs, redact tokens, auth files, and private local paths.