Automations

Set-and-forget workflows — scheduled backups, disk alerts, container health checks, AI-driven maintenance, and webhook-triggered deploys.

Automations are Talome's answer to "I wish this happened automatically." A nightly backup. A disk usage alert. A container health check that restarts crashed services and notifies you on Telegram. An AI-driven improvement session that runs while you sleep.

You describe the workflow in plain English, the AI builds the automation, and it runs on schedule without further intervention. Or you build it yourself in the dashboard UI or via the API.

Watch: Creating a disk usage alert automation in one conversation

Anatomy of an Automation

Every automation has the same structure: a trigger that decides when it runs, and steps that define what happens.

Trigger: "Every hour" (cron: 0 * * * *)
    ↓
Step 1: Check disk usage (tool_action: get_disk_usage)
    ↓
Step 2: Is usage >= 90%? (condition: $.usage_percent gte 90)
    ↓ yes
Step 3: Send alert (notify: "Disk at {{usage_percent}}%")

Data flows between steps. The output of Step 1 becomes the input context for Step 2's condition check. Template variables like {{usage_percent}} resolve from previous step results.

Trigger Types

Trigger	Configuration	Example
Schedule (cron)	`expression`: a cron string	`0 3 * * *` — every day at 3 AM
Container stopped	`container`: container name	Fires when `jellyfin` stops unexpectedly
Disk usage exceeds	`threshold`: percentage	Fires when disk usage crosses 90%
Webhook	`path`: URL path	Fires when POST hits `/api/webhooks/deploy`
App installed	`appId`: optional filter	Fires when any (or a specific) app is installed
Manual	(none)	Run on demand from the dashboard or API

Cron Expressions

Cron expressions use the standard 5-field format: minute hour day-of-month month day-of-week.

Expression	Meaning
`0 * * * *`	Every hour, on the hour
`/15 * * *`	Every 15 minutes
`0 3 * * *`	Daily at 3:00 AM
`0 0 * * 0`	Weekly, Sunday at midnight
`0 2 1 * *`	Monthly, 1st at 2:00 AM
`0 4 * * 1-5`	Weekdays at 4:00 AM
`30 22 * * *`	Daily at 10:30 PM

The AI can validate cron expressions before saving with validate_cron:

Talome Assistant

Does "0 3 * * *" mean every day at 3am?

⚙

validate_cron → valid: fires at 03:00 every day

Yes. That expression fires at 03:00 every day. Next 3 runs: tomorrow at 3:00 AM, day after at 3:00 AM, and so on.

Step Types

tool_action

Execute any automation-safe Talome tool with specific parameters.

{
  "type": "tool_action",
  "tool": "get_disk_usage",
  "params": {}
}

{
  "type": "tool_action",
  "tool": "backup_app",
  "params": { "appId": "jellyfin" }
}

ai_prompt

Run an AI prompt with a curated set of allowed tools. The AI reasons about the situation and decides which tools to call — just like in a chat conversation, but unattended.

{
  "type": "ai_prompt",
  "prompt": "Check all containers for health issues. If any container has restarted more than 3 times in the last hour, investigate the logs and send a notification with the diagnosis.",
  "allowedTools": ["list_containers", "get_container_logs", "inspect_container", "send_notification"]
}

The ai_prompt step is the most powerful step type. It combines the AI's reasoning with specific tool access — you define the goal and boundaries, the AI figures out the execution.

condition

Check a value from a previous step's output. If the condition is false, the automation stops (subsequent steps are skipped).

{
  "type": "condition",
  "field": "$.usage_percent",
  "operator": "gte",
  "value": 90
}

The field uses JSONPath-style dot notation to extract values from previous step results.

notify

Send a notification through a configured channel (Telegram, Discord, email, webhook).

{
  "type": "notify",
  "channel": "telegram",
  "message": "Disk usage is at {{usage_percent}}%. Top directories: {{top_dirs}}"
}

Template variables ({{variable}}) resolve from the accumulated context of all previous steps.

Creating Automations

Via AI Conversation

The fastest way — describe what you want and the AI builds it:

Talome Assistant

Every night at 2am, back up Jellyfin and Sonarr, then notify me on Telegram when it's done

⚙

create_automation → created: Nightly backup, cron 0 2 * * *, 3 steps

Automation created: "Nightly backup -- Jellyfin + Sonarr". Schedule: daily at 2:00 AM. Steps: backup Jellyfin, backup Sonarr, Telegram notification. Next run: tomorrow at 2:00 AM.

Via the Dashboard

Navigate to Automations in the sidebar. The UI provides:

A visual automation builder with drag-and-drop steps
Cron expression helper with human-readable preview
Tool picker that shows only automation-safe tools
Template variable autocomplete in notification messages
Run history with logs for each execution

Via the API

curl -X POST http://localhost:4000/api/automations \
  -H "Content-Type: application/json" \
  -d '{
    "name": "Disk alert",
    "trigger": { "type": "cron", "expression": "0 * * * *" },
    "steps": [
      { "type": "tool_action", "tool": "get_disk_usage", "params": {} },
      { "type": "condition", "field": "$.usage_percent", "operator": "gte", "value": 90 },
      { "type": "notify", "channel": "telegram", "message": "Disk at {{usage_percent}}%" }
    ]
  }'

8 Complete Examples

1. Nightly Backup

Back up critical apps every night.

Talome Assistant

Back up Jellyfin and Sonarr every night at 2am

⚙

create_automation → created: Nightly backup, cron 0 2 * * *, 2 steps

Automation created: "Nightly backup" -- daily at 2:00 AM, backs up Jellyfin then Sonarr.

Resulting automation:

{
  "name": "Nightly backup",
  "trigger": { "type": "cron", "expression": "0 2 * * *" },
  "steps": [
    { "type": "tool_action", "tool": "backup_app", "params": { "appId": "jellyfin" } },
    { "type": "tool_action", "tool": "backup_app", "params": { "appId": "sonarr" } }
  ]
}

2. Disk Usage Alert

Check disk space hourly and alert when it's getting full.

Talome Assistant

Alert me on Telegram when disk usage goes above 90%

⚙

create_automation → created: Disk usage alert, hourly, 3 steps with condition

Automation created: "Disk usage alert" -- checks every hour, alerts on Telegram if disk usage is 90% or above.

{
  "name": "Disk usage alert",
  "trigger": { "type": "cron", "expression": "0 * * * *" },
  "steps": [
    { "type": "tool_action", "tool": "get_disk_usage", "params": {} },
    { "type": "condition", "field": "$.usage_percent", "operator": "gte", "value": 90 },
    { "type": "notify", "channel": "telegram", "message": "Disk usage is at {{usage_percent}}%. Free space: {{free_gb}} GB." }
  ]
}

3. Container Health Check with Recovery

Monitor containers every 5 minutes. If any are unhealthy, restart them and notify.

Talome Assistant

Every 5 minutes, check if any containers are unhealthy. If so, restart them and tell me on Discord.

⚙

create_automation → created: Container health monitor, every 5 min, AI prompt + notify

Automation created: "Container health monitor" -- every 5 minutes, an AI prompt checks all containers, restarts unhealthy ones, and sends a summary to Discord.

{
  "name": "Container health monitor",
  "trigger": { "type": "cron", "expression": "*/5 * * * *" },
  "steps": [
    {
      "type": "ai_prompt",
      "prompt": "Check all containers. For any that are unhealthy or exited unexpectedly, restart them. Report what you found and what you restarted.",
      "allowedTools": ["list_containers", "inspect_container", "restart_container", "get_container_logs"]
    },
    { "type": "notify", "channel": "discord", "message": "Health check complete: {{summary}}" }
  ]
}

4. Download Cleanup

Clean up completed downloads daily to reclaim disk space.

Talome Assistant

Every day at midnight, clean up completed torrents in qBittorrent

⚙

create_automation → created: Download cleanup, daily at midnight, 2 steps

Automation created: "Download cleanup" -- daily at midnight, lists completed torrents then uses AI to remove old ones with good ratios.

{
  "name": "Download cleanup",
  "trigger": { "type": "cron", "expression": "0 0 * * *" },
  "steps": [
    { "type": "tool_action", "tool": "qbt_list_torrents", "params": { "filter": "completed" } },
    {
      "type": "ai_prompt",
      "prompt": "Review completed torrents. Remove any that finished downloading more than 7 days ago and have a ratio above 1.0. Report how many were cleaned up and how much space was freed.",
      "allowedTools": ["qbt_list_torrents"]
    }
  ]
}

5. Weekly Library Scan

Scan Jellyfin libraries every Sunday to pick up manually added files.

Talome Assistant

Scan all Jellyfin libraries every Sunday at 6am

⚙

create_automation → created: Weekly Jellyfin scan, Sundays at 6 AM

Automation created: "Weekly Jellyfin scan" -- every Sunday at 6:00 AM, triggers a full library scan.

{
  "name": "Weekly Jellyfin scan",
  "trigger": { "type": "cron", "expression": "0 6 * * 0" },
  "steps": [
    { "type": "tool_action", "tool": "jellyfin_scan_library", "params": {} }
  ]
}

6. AI-Driven Overnight Improvement

Let the AI improve Talome's own code while you sleep.

Talome Assistant

Every night at 3am, review open issues and apply up to 2 fixes

⚙

create_automation → created: Nightly self-improvement, daily at 3 AM

Automation created: "Nightly self-improvement" -- daily at 3:00 AM, an AI prompt reviews open issues, plans and applies up to 2 fixes with TypeScript validation.

{
  "name": "Nightly self-improvement",
  "trigger": { "type": "cron", "expression": "0 3 * * *" },
  "steps": [
    {
      "type": "ai_prompt",
      "prompt": "Review open issues with list_issues. Pick the highest priority issue. Use plan_change to understand the fix, then apply_change to implement it. Stop after 2 successful changes or if any change fails.",
      "allowedTools": ["list_issues", "plan_change", "apply_change", "read_file", "list_changes", "track_issue"]
    }
  ]
}

7. Webhook-Triggered Deploy

Redeploy an app when a webhook fires (e.g., from a CI pipeline or GitHub Actions).

Talome Assistant

When my deploy webhook fires, pull the latest image for my-custom-app and restart it

⚙

create_automation → created: Deploy my-custom-app, webhook trigger, 3 steps

Automation created: "Deploy my-custom-app" -- triggered by webhook, upgrades the image, restarts the app, and sends a Telegram notification. Webhook URL: POST http://your-server:4000/api/webhooks/deploy/my-custom-app. Call this URL from your CI pipeline to trigger the deploy.

{
  "name": "Deploy my-custom-app",
  "trigger": { "type": "webhook", "path": "/deploy/my-custom-app" },
  "steps": [
    { "type": "tool_action", "tool": "upgrade_app_image", "params": { "appId": "my-custom-app" } },
    { "type": "tool_action", "tool": "restart_app", "params": { "appId": "my-custom-app" } },
    { "type": "notify", "channel": "telegram", "message": "Deployed latest version of my-custom-app." }
  ]
}

8. Conditional Restart on High Memory

Check container memory every 10 minutes and restart heavy containers.

Talome Assistant

If any container uses more than 2GB of memory, restart it and notify me. Check every 10 minutes.

⚙

create_automation → created: Memory watchdog, every 10 min, AI prompt + notify

Automation created: "Memory watchdog" -- every 10 minutes, an AI prompt checks container stats, restarts any using more than 2 GB, and sends a Telegram summary.

{
  "name": "Memory watchdog",
  "trigger": { "type": "cron", "expression": "*/10 * * * *" },
  "steps": [
    {
      "type": "ai_prompt",
      "prompt": "Check container stats. If any container is using more than 2 GB of memory, restart it and report which container was restarted.",
      "allowedTools": ["list_containers", "get_container_stats", "restart_container"]
    },
    { "type": "notify", "channel": "telegram", "message": "Memory watchdog: {{summary}}" }
  ]
}

Automation-Safe Tools

Automations can only use tools explicitly marked as automation-safe. This prevents unattended workflows from accidentally running destructive operations like uninstalling apps, deleting volumes, or pruning Docker resources.

What's allowed:

All read-tier tools (list, get, search, check, inspect)
Curated modify-tier tools (restart, backup, scan, notify, set preferences)
AI prompt steps with explicitly listed allowedTools

What's blocked:

Destructive-tier tools (uninstall, delete, prune, cleanup)
Shell commands (run_shell)
Self-improvement tools (apply_change) unless explicitly listed in an ai_prompt step

Check which tools are available for automations:

Talome Assistant

What tools can automations use?

⚙

list_automation_safe_tools → 87 tools across 6 categories

Automations can use 87 tools across these categories: Docker (list_containers, get_container_logs, restart_container, ...), System (get_system_stats, get_disk_usage, ...), Apps (list_apps, restart_app, backup_app, ...), Media (get_library, get_downloads, ...), Arr (arr_get_status, arr_get_queue_details, ...), and Notifications (send_notification, ...).

Condition Operators Reference

Use these in condition steps to control automation flow:

Operator	Description	Example
`eq`	Equals	`$.status eq "running"`
`neq`	Not equals	`$.status neq "healthy"`
`gt`	Greater than	`$.count gt 100`
`lt`	Less than	`$.free_gb lt 50`
`gte`	Greater than or equal	`$.usage_percent gte 90`
`lte`	Less than or equal	`$.cpu_percent lte 10`
`contains`	String contains	`$.message contains "error"`

The field uses dot notation to traverse the JSON output from previous steps. $ refers to the root of the previous step's result.

Monitoring Automation Runs

Via the AI

Talome Assistant

Show me recent automation runs

⚙

get_automation_runs → last 10 runs retrieved

Last 10 runs: Nightly backup (cron, Success, 2m 34s at 2:00 AM), Disk usage alert (cron, Skipped -- condition false, 1s at 1:00 AM), Container health (cron, Success, 8s at 12:55 AM), Memory watchdog (cron, Success -- restarted jellyfin, 12s at 12:50 AM), Disk usage alert (cron, Skipped, 1s at 12:00 AM). The nightly backup ran successfully in 2m 34s. The disk alert hasn't triggered -- your disk is at 67%. The memory watchdog restarted Jellyfin at 12:50 AM (it was using 2.3 GB).

Via the Dashboard

The Automations page shows:

All automations with their status (active/paused), trigger type, and next run time
Run history for each automation with expandable logs
Success/failure counts and average duration
A "Run Now" button to trigger any automation manually

Editing and Managing Automations

Updating an Automation

Talome Assistant

Change the disk alert threshold from 90% to 85%

⚙

list_automations → found Disk usage alert (id=3)

⚙

update_automation → threshold updated from 90% to 85%

Updated "Disk usage alert" -- threshold changed from 90% to 85%. Next run: top of the next hour.

Deleting an Automation

Talome Assistant

Delete the weekly library scan automation

⚙

list_automations → found Weekly Jellyfin scan (id=5)

This will permanently delete "Weekly Jellyfin scan" and all its run history. Confirm?

Confirmed

⚙

delete_automation → deleted Weekly Jellyfin scan

Deleted "Weekly Jellyfin scan."

Troubleshooting

Automation Didn't Run

Check the trigger. For cron triggers, verify the expression matches your expectation:

Talome Assistant

My 3am automation didn't run last night

⚙

list_automations → found Nightly self-improvement, active

⚙

validate_cron → valid: daily at 3:00 AM

⚙

get_automation_runs → last run was 3 days ago

The automation "Nightly self-improvement" is active with cron "0 3 * * *" (daily at 3:00 AM). But the last run was 3 days ago. Your server was rebooted at 2:47 AM and didn't finish starting until 3:12 AM. The cron trigger fired during the restart window and was missed. I can run it now manually, or shift the schedule to 3:30 AM to give the server boot time.

⚙

update_automation → schedule shifted to 30 3 * * *

Condition Never Triggers

If your alert automation never fires, the condition might be too strict:

Talome Assistant

The disk alert has never triggered -- is it working?

⚙

get_automation_runs → 240 runs, all conditions evaluated false

⚙

get_disk_usage → 67% used

The automation has run 240 times (hourly for 10 days). Every run hit the condition step and evaluated to false -- your disk usage has never exceeded 90%. Current disk usage: 67%. The automation is working correctly; your disk just hasn't hit the threshold. Want me to lower the threshold to 80% for earlier warning?

AI Prompt Step Fails

If an ai_prompt step fails, check the allowed tools list:

Talome Assistant

The health check automation keeps failing

⚙

get_automation_runs → last 3 runs failed: restart_container not in allowedTools

The last 3 runs failed with: "Tool restart_container is not in the allowedTools list for this ai_prompt step." The AI tried to restart an unhealthy container but restart_container wasn't in the allowed tools. Adding it now.

⚙

update_automation → added restart_container to allowedTools

Fixed. The AI prompt step now has permission to restart containers.

Anatomy of an Automation

Trigger Types

Cron Expressions

Step Types

tool_action

ai_prompt

condition

notify

Creating Automations

Via AI Conversation

Via the Dashboard

Via the API

8 Complete Examples

1. Nightly Backup

2. Disk Usage Alert

3. Container Health Check with Recovery

4. Download Cleanup

5. Weekly Library Scan

6. AI-Driven Overnight Improvement

7. Webhook-Triggered Deploy

8. Conditional Restart on High Memory

Automation-Safe Tools

Condition Operators Reference

Monitoring Automation Runs

Via the AI

Via the Dashboard

Editing and Managing Automations

Updating an Automation

Deleting an Automation

Troubleshooting

Automation Didn't Run

Condition Never Triggers

AI Prompt Step Fails

Automation Schema Reference

AI Assistant

Self-Improvement

Backups

On this page

Automations

Source

Source

Automation Schema Reference

AI Assistant

Self-Improvement

Backups

On this page