Skip to content
Chatvia Docs

Admin panel

Radar

Get in touch: Ahmad Mayahi, creator of Chatvia, Spectra and Shakomako

Radar lists the issues you need to fix in your install: an AI provider that refuses your API key, an account that is out of credit, a model that a provider removed, emails that cannot be sent, a backup that failed. Without it, people just get no answer or no video, and you never learn why. Open it with Radar (the antenna icon) in the admin panel's sidebar, above Settings. A short line at the top of the page says what Radar is for.

You need it when people say something does not work, or to check now and then that all is well. A red number next to Radar in the sidebar counts the open issues. If you never open it, nothing changes: the issues still happen, you just do not see them.

Only admins can open this page.

Issues are not the same as the app's log. The log, on the Logs page, is the app's diary for developers: every warning and error in the code. Radar only shows things you can fix in a settings page or at a provider.

What becomes an issue

Issue What it means
Key refused The service turned the API key away: it is wrong, deleted, or may not use the model.
No key An agent or an app task uses a model of a provider that has no API key saved.
Turned off An agent or an app task uses a model of a provider you turned off.
Out of credit The account at the service has no credit or quota left.
Model gone The provider no longer has a model you use.
Too many requests The service limits how often the account may call it.
Service failing The service had trouble on its side.
Out of reach The server could not connect to the service.
Cannot send An email could not be sent.
Backup failed A backup did not finish.
License off chatvia.ai says your license key is no longer active, so it was removed.
Job failed A background job failed after all its tries, such as making a video or reading a file. One issue per kind of job.
Queue falling behind Jobs wait longer than they should, so replies, training or exports come late.
Task failed A scheduled task failed, such as a nightly clean-up or the license check.
Down A part of the install is down: the queue, the scheduler, the database, Redis, live updates, file storage or email.
Message limit reached A workspace used the monthly message limit you set, so its agents stop answering.
Generation limit reached A workspace used the monthly limit you set on pictures, audio files or videos.
Free limit reached Someone tried to make a new workspace, but without a license key the install holds one.

The services are the AI providers in Settings → AI providers, Firecrawl (the web scraper), IP2Location, email, backups, your license, the queue and the scheduler, the database, Redis, live updates, file storage, and your workspaces' limits. Each shows its logo or a coloured icon in the list.

Some parts of the install stop without an error, such as the scheduler or the queue. Radar checks them every 5 minutes, and checks the scheduler each time an admin opens the admin panel, since a stopped scheduler cannot report itself. Services that each workspace connects on its own, such as its channels or its Perplexity key, are not here: the workspace fixes those itself.

Testing a key with a Check button in Settings or on the Status page never makes an issue.

The list

Each row is one issue with one service, such as "fal.ai refused the API key". When it happens again, the same row counts up.

Column What it shows
Issue The service's logo, what is wrong, and below it what the service said last.
Times How often it happened.
Last seen When it last happened. Hover to see the exact date.
Status Open, Resolved or Ignored. See Statuses.

The list shows open issues, newest first, and refreshes every 30 seconds. Use the filters to show other statuses, one service or one kind of issue.

Tick rows to resolve, ignore or delete several issues at once. Delete resolved removes every resolved issue.

One issue

Click a row to open the issue. It shows:

  • What to do: what to change, and a button to the right settings page. For a refused key there is also a button to the page where the provider makes keys.
  • Summary: the service, the kind of issue, the model, the status, how often it happened, when it was first and last seen, and what the service said last.
  • Last 14 days: a bar for each day, so you see whether it is new, steady or over.
  • Occurrences: a tab with every time it happened, newest first. Filter it by who or by workspace. Click a row to see all its details.

Each occurrence shows when it happened, who it happened to, the workspace (click it to open the workspace), the agent, what the app was doing (such as making a video), the model, the HTTP status and what the service said. The HTTP status is the number a service answers with, such as 403 for "not allowed".

Who is one of:

  • a staff member (an owner, teammate or admin), by name,
  • a chatter, by email,
  • an API key, by name,
  • Visitor: someone who was not signed in,
  • System: the app itself, such as a scheduled task.

A person who was deleted shows as Deleted staff or Deleted chatter. Deleting a workspace deletes its occurrences too.

Copy details at the top copies the issue as plain text, ready to send to someone who can help.

Statuses

  • Open: not fixed yet. Open issues count in the sidebar.
  • Resolved: fixed. Most issues close by themselves: the next time the service answers again, for example right after you add credit, the issue is resolved and shows Closed by itself. A failed job closes when that job runs fine again, a failed task when it next succeeds, a part of the install when it runs again, and a limit when you raise it or the month restarts. A gone model does not close by itself; resolve it once you picked another model.
  • Ignored: you know about it and do not want to hear about it. It keeps counting, but stays out of the sidebar count.

Use Resolve, Reopen, Ignore and Delete at the top of an issue. A resolved issue that happens again opens again.

How long issues stay

Issues not seen for 30 days are deleted once a day, with their occurrences. Change the number of days, or turn the page off, in Settings → Observability → Radar.

When one issue happens more than 100 times in a minute, the extra times are only counted, so the database does not fill up.

Search guides, settings and commands.