Skip to content
bsnwgitPublic

About

One dashboard for your media pipeline — Seerr, Sonarr, Radarr, NZBGet/SABnzbd: health, queues, missing, and why things failed

Topics

Resources

Contributing

Stars

1 star

Watchers

0 watching

Forks

Repository files navigation

Cortexarr

License: PolyForm Noncommercial 1.0.0 GitHub Discussions

One dashboard for a self-hosted media acquisition pipeline — Seerr, Sonarr, Radarr, and your Usenet download client (NZBGet or SABnzbd) — instead of tabbing between five admin UIs to find out what's stuck.

Cortexarr shows each service's own health, what's downloading and what's missing right now, and why things failed, and lets you act on it (approve a request, retry a failed download, search for a missing episode) without leaving the page.

Dashboard

What it does

  • Seerr — requests with who asked for what, and Approve / Decline for anything still pending. An Issues list pairs user-reported problems with requests that failed to reach Sonarr/Radarr, including why they failed (pulled from Seerr's log), with Retry.
  • Sonarr — your series library, each series with its seasons and episodes (monitor, search, delete), the download queue — with Import, Remove & search, and Remove right on a stuck item — missing episodes grouped by series, and a month-grid calendar. The library filters by monitored state, continuing/ended, or missing episodes — filter to missing, select the ones you want, and search all of them at once. Deleting a series offers removing its matching Seerr request too, so Seerr doesn't keep thinking it's wanted and quietly try to fill it again. Can also accept its own webhook (Settings → Services → Webhook) so a health change reaches Cortexarr the moment it happens rather than waiting for the next poll — polling itself never stops, since there's no webhook for "an item's been stuck in the queue a while."
  • Radarr — the same for movies: library, movie pages with release dates and the file on disk, queue, missing, and a release calendar. Same coordinated delete, same optional webhook.
  • NZBGet / SABnzbd — live speed, time left, free disk, the queue with progress and post-processing stage, and history with failure reasons and Retry. Pause/resume a download or everything at once.
  • Live TV (Dispatcharr) — its own Live TV tab, kept apart from the media pipeline. See who is watching and stop a stream; list, search and filter channels, enable or disable them per profile (select a whole page, shift-click a range, or every channel matching a search), and edit a channel's name, number or group; refresh M3U accounts and EPG sources, and add, edit or delete them (admins). Signs in with a username and password or an API key, your choice per service.
  • Duplicate files — each Sonarr and Radarr service can look through every folder in its own library, using nothing but its connection (no SSH, no mount): Radarr flags a movie folder with more than one video file, Sonarr an episode code like S03E02 held by more than one file. A count pill on the service's Dashboard card, a Duplicates tab that appears only while there are any (with the service that found each, a link to the series or movie, and for admins a button to delete the extra copy), and a Duplicate file alert event.
  • Request tracking — every Seerr request followed through approval, Sonarr/Radarr, searching, downloading and import to available, with how long it's been in its stage (and episode counts for TV); alert rules flag a request that stalls in a stage longer than you set. Flags it too when Seerr's own status has gone stale — it says available, but Sonarr/Radarr is still searching — with a one-click Search again. Also flags a request orphaned — nothing in Sonarr/Radarr matches it at all any more, the title having been deleted after Seerr sent or marked it — with one-click Clear. Available requests get a Verify still on disk action that asks Sonarr/Radarr to re-check the file, in case it's been moved or deleted outside the pipeline.
  • Dashboard — requests run full width across the top; below, one card per service with nested cards for whatever needs attention. Each card can be refreshed on its own or opened in the service's own web UI.
  • History — a top-nav tab with reporting over time: per-service uptime (from health checks already being recorded), alert frequency with the most active rules and services, and requests completed with average time to available. A 7/30/90-day range on each. Request trends fill in going forward from when this shipped — nothing before that is backfilled.
  • Statistics — a top-nav tab after History, one tab per service, read live (nothing stored): Sonarr and Radarr report version, queue counts, library size, media library and disk space, and grabbed/imported/failed counts; NZBGet and SABnzbd report per news server data (today, 7/30/365 days, total), article success and failure, and speed/data charts; Seerr reports request counts.
  • Notifications — a top-nav tab, next to Tracking, with four views: Alerts and Rules (below), an activity feed across every service (filterable by pipeline), and an audit log of every change made in Cortexarr.
  • MCP server — AI tools (any MCP client) can do everything the web UI does, through personal API tokens that are read-only by default and never delete anything unless you allow it. Upload your own certificate under Settings → Server to serve everything — the web UI, the API, and /mcp — over HTTPS, for clients that refuse a plain-HTTP endpoint.
  • Per-service health polling with retry/backoff, a connection test that also checks the URL really is the service you picked, role-based access (admin / analyst / viewer) with a Users page to manage them, and notifications over email, Slack/Discord webhooks, ntfy push, and Twilio SMS.
  • Alerts — your own rules for what notifies: a service unreachable or reporting an error, a queue item stuck, a download failed, a request failed or an issue reported — on which services, after how long, through which channels, with an optional "resolved" message when it clears and reminders while it lasts. Snooze or acknowledge a known problem without muting the whole rule, and set a digest window so a bad night is one message every so often, not dozens. Suggested starter rules (a service unreachable, a queue item stuck, a request stalled) are offered as one-click templates while the rule list is empty.
  • Backup — export the service list and notification configuration (not the database) to move a setup or back it up, credentials excluded unless you choose to include them — which requires a password and encrypts the file, rather than writing secrets out in plain text. Re-importing skips anything already there by name, so it's safe to run the same file twice.
  • Self-update — checks GitHub releases against the running version; manual mode (the default) shows what's available, with an Update now button for admins; auto mode downloads and applies it inside a maintenance window you set. Either way it then restarts itself.

Requested features

Planned but not built yet — kept here so none of it gets lost:

  • Force sync with reality, the download client — a queue entry with nothing matching in Sonarr/Radarr's own queue wastes bandwidth for nothing, but correlating them reliably needs a field-by-field check against real download-client data (NZBGet's queue doesn't obviously expose the same id Sonarr/Radarr records) that hasn't been verified against a real instance yet.
  • AI provider integration — pluggable AI providers inside Cortexarr itself (the MCP server for outside AI tools is built).

Screenshots

Series library Series page
Sonarr — series library Sonarr — a series, its seasons and episodes
Movie page Calendar
Radarr — a movie, with its release dates Calendar — a month at a time
Seerr requests Seerr issues
Seerr — requests, approve/decline Seerr — issues, with why a request failed
Download queue Download history
NZBGet / SABnzbd — queue NZBGet / SABnzbd — history and failure reasons
Request tracking Force sync with reality
Tracking — every request's pipeline stage Tracking — a Seerr/reality mismatch and an orphaned request, each with its fix
History and trends Self-update
History — uptime, alert frequency, and request-completion trends Self-update — checks GitHub releases against the running version

Sonarr/Radarr queue with Import/Remove/Redownload

Alerts — current problems with one-click fixes

Activity log

Screenshots use a made-up demo library.

Requirements

Native install:

  • Ubuntu 22.04/24.04 LTS (or any Linux with Python 3.11+ and systemd)
  • Python 3.11+
  • Node.js (for building the frontend — any recent LTS)

Docker: just Docker and Compose — the image builds the frontend itself.

Either way: any of Sonarr, Radarr, Seerr, NZBGet, SABnzbd, or Dispatcharr — add only what you run.

Install

Get the code first, either way:

git clone https://github.com/bsnwgit/cortexarr.git
cd cortexarr

Then, native (Ubuntu, systemd):

bash install.sh

Prompts for an install directory (default /opt/cortexarr) and port (default 8770), then sets up a virtualenv, database, systemd service, and builds the frontend. Prints the generated admin password once at the end — save it.

To remove: bash uninstall.sh (keeps your data by default; --purge to remove that too).

Or Docker:

docker compose up -d --build
docker compose logs -f

Watch the logs for the "Cortexarr admin account created" block — the generated username and password are printed there once, on first boot only, and won't be shown again. Ctrl+C once you've got it, then open http://<this-host>:8770 and sign in.

Everything persistent — config.yaml, the database, logs — lives in the cortexarr-data volume; the image itself holds only code. Settings → General's Update now works in Docker too: it stores the new release in that volume and restarts the container (see the Admin Guide). To move the image itself, run docker compose up -d --build after pulling the new source, or docker compose pull && docker compose up -d for a published image.

To remove: docker compose down (keeps the cortexarr-data volume, so your config and database survive; add -v to delete that too).

Getting started

  1. Open Cortexarr in a browser and sign in as admin with the password the installer printed. Change it from the user menu (your username, top right).
  2. From the same menu, open Settings → Services and choose New service. Pick the type, enter its URL and API key (NZBGet asks for its username and password instead), and Test connection before saving.
  3. Repeat for each service you run. The Dashboard fills in as they're checked. Settings → Services lists them, with test, maintenance mode, and delete.

The User Guide walks through each page; the Admin Guide covers adding services, maintenance mode, notifications, and users.

Locked out?

An admin can reset anyone else's password from Settings → Users, but if you're the only admin and you're locked out yourself, there's no self-service reset yet — go straight to the database. This updates the password in place; nothing else needs to stop or restart.

Native install:

cd /opt/cortexarr   # or wherever you installed it
venv/bin/python3 -c "
import bcrypt, sqlite3, getpass
new_password = getpass.getpass('New password for admin: ')
h = bcrypt.hashpw(new_password.encode(), bcrypt.gensalt()).decode()
conn = sqlite3.connect('cortexarr.db')
conn.execute('UPDATE users SET hashed_password = ? WHERE username = ?', (h, 'admin'))
conn.commit()
conn.close()
print('Password updated.')
"

Docker:

docker exec -it cortexarr python3 -c "
import bcrypt, sqlite3, getpass
new_password = getpass.getpass('New password for admin: ')
h = bcrypt.hashpw(new_password.encode(), bcrypt.gensalt()).decode()
conn = sqlite3.connect('/data/cortexarr.db')
conn.execute('UPDATE users SET hashed_password = ? WHERE username = ?', (h, 'admin'))
conn.commit()
conn.close()
print('Password updated.')
"

The password is typed interactively — it's never in your shell history or visible on screen. If the locked-out account isn't literally named admin, change the username in the WHERE clause first.

Configuration

Startup/infrastructure settings live in config.yaml (copy from config.example.yaml if not using install.sh). Everything else — monitored services, notification channels, retention, polling intervals — is managed through the web UI and stored in SQLite. Service API keys and passwords are encrypted at rest.

License

PolyForm Noncommercial 1.0.0 — see LICENSE.

About

One dashboard for your media pipeline — Seerr, Sonarr, Radarr, NZBGet/SABnzbd: health, queues, missing, and why things failed

Topics

Resources

Contributing

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages