Troubleshooting

Two commands answer most questions:

olace status    # what state is everything in?
olace logs -f   # what is the daemon doing right now?

The app says the background service is not running

olace service status, then olace service restart.
Not installed? olace service install, or sign in again; olace signin offers to set it up.
Still down? Run it in the foreground and read the output directly: olace run.

`olace update` rolls back under `olace run`

On a computer without systemd (most containers) Olace runs in the foreground with olace run. From 1.1.12 on, olace update restarts it onto the new version in place. Updating FROM 1.1.11 or earlier still rolls back, because the old version does the restart. Stop olace run, run olace update, then start olace run again. Later updates work in place.

No local provider running

Olace starts your runtimes on demand, but one must be installed. olace setup checks and installs.
Installed but undetected? Confirm the runtime answers where Olace expects it: Ollama at localhost:11434, llmster at localhost:1234, llama.cpp at 127.0.0.1:8143. Custom locations go in ollama_url / lmstudio_url / llamacpp_url in daemon.json.
LM Studio models vanished after you quit the LM Studio app? Quitting it stops the llmster inside it. Olace starts llmster again on its own, immediately when a message needs it and within a couple of minutes when you are only browsing, so give it a moment before digging further. If it never comes back, check that the lms command is still installed.

A paired device shows offline

Is the host awake and online? A sleeping computer is unreachable, exactly like a closed laptop.
On the host, olace status: the tunnel should show connected. If not, give it a moment; it reconnects on its own with backoff.
Pairing marked "Paused. Upgrade to reactivate" is over your plan's pairing cap, not broken. See Plans & credits.
Version mismatch is shown as an update-needed state, never a deleted pairing. Update both sides.

A model will not load, or runs out of GPU memory

Free memory first: olace unload, then retry.
Pick a smaller variant of the model; the Model hub labels what fits your hardware.
Turn on auto-optimize to fit more context in the same memory: olace settings auto-optimize on.

The Bridge refuses connections

Is it on? olace bridge shows configured state and whether it is actually listening.
The Bridge is loopback-only by design. Calling it from another machine will not work; run the Bridge on that machine instead, or front it with a reverse proxy plus olace bridge key --require-key.
Browser tool getting a 403 origin_not_allowed? Desktop apps and pages served from your own computer are allowed by default, so this means a hosted web app. The daemon log names the exact origin; allow it with olace bridge cors add <origin>.
401 means a key is set and your tool did not send it: Authorization: Bearer <key>.

Something works in the desktop app but not on the web

That may be by design: local AI and hosting pairings need a computer, and Direct Mode chats start in the desktop and mobile apps, so the web app does not offer them. See Privacy & data.

Check spam, and give it a minute on a slow connection.
The code screen offers a resend after a short wait.

Lost access to your email

Still signed in on any device? Sign in on the new one by approving it from that device, then change it in Settings › Account › Email.
Signed up with Google and still have that Google account? Sign in with Google.
Neither? Write to support@olace.app with your order ID or receipt if you bought a plan. Once we verify the account is yours, we change its email after a waiting period, and the old address is notified. Your plan, credits and history stay on the same account, and your PIN or Recovery Key unlocks your synced conversations as usual.

When you report a problem

Include your platform, olace version, olace status --json, and the relevant lines of olace logs. Reach us at support@olace.app. Logs stay on your machine and are never sent automatically.