NetSendo v2.8.4: Per-Task Token Limits Keep Brain Running

Imagine you start the day with a dozen lists, several scheduled cron tasks and a few campaigns to prepare. By noon, NetSendo's AI assistant Brain stops responding. The message tells you to "increase the limit in settings", but the panel has no setting for it. Until now, this could happen to busy accounts on a normal working day.

NetSendo v2.8.4 fixes this. The token limit now applies to each task separately instead of the whole day. You can change it in the panel or through the API, and one expensive task can no longer block all the others.

🚀 What's New in v2.8.4

The Problem With a Daily Cap

Before this release, Brain stopped every task for the rest of the day once the account used 100,000 tokens. On a large model, that was roughly four campaigns. Accounts with many lists and tasks often ran out by midday. The limit also couldn't be changed anywhere in the panel.

Think of it like a shared office printer with one daily page quota. When one colleague prints a large report, nobody else can print until tomorrow. The new model gives each print job its own page limit. A runaway job gets stopped, and everyone else carries on.

What Counts as a "Task"

Each of these units now has its own token budget:

• 💬 One chat message sent to Brain
• 📋 One plan run
• 🎯 One goal step
• ⏰ One cron task

ℹ️ Note: A plan that runs inside a chat message counts towards that message's budget. It doesn't get a separate budget of its own.

What Happens When a Task Exceeds Its Budget

When a task goes over its limit, Brain stops it cleanly and predictably:

• The task makes no further AI calls.
• The chat message or plan report shows how many tokens it used and where to change the limit.
• The remaining steps of the plan are marked as skipped, and nothing is sent after them.
• A goal step stopped this way is not retried, so it can't keep using tokens in a loop.
• Other tasks keep working as usual.

⚠️ Important: When a plan is stopped by the limit, Brain doesn't send anything after that point. This protects your subscribers from half-finished campaigns. Check the plan report to see which steps were skipped.

Configuration in the Panel and API

If you manage settings in code, use the new task_token_limit field:

curl -X PUT https://your-netsendo.example.com/brain/api/settings \
  -H "Authorization: Bearer YOUR_API_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"task_token_limit": 150000}'

⚠️ Migration note: daily_token_limit is no longer used. Update any scripts, n8n workflows or infrastructure-as-code that set it to use task_token_limit instead.

💼 Benefits & Use Cases

🛠️ Step-by-Step: Configure Your Per-Task Limit

1. Update to v2.8.4

   Pull the latest image and restart your stack:

   docker compose pull && docker compose up -d

2. Open Brain Settings

   Go to Brain → Settings and find Token limit per task. Today's usage is shown next to it.

3. Choose a Value

   Keep the 100,000 default or adjust it to match your model and typical tasks. Setting it to 0 removes the limit.

4. Update API Integrations

   Replace daily_token_limit with task_token_limit in any calls to PUT /brain/api/settings.

5. Verify

   Send /status on Telegram or open the Monitor to confirm the limit and today's usage.

💡 Common pitfall: If a plan run inside chat stops unexpectedly, remember that it shares the chat message's budget. Run large plans as separate tasks or raise the limit.

🐛 Fixes in This Release

• Cleaner approval summaries: The approval summary in the Monitor and on Telegram used to show ****Steps to execute:**** with doubled bold markers. It now displays correctly.
• Natural dry-run wording: A dry run of an immediate send used to say the campaign would go out "at now" ("o teraz" in Polish, "um jetzt" in German, "a las ahora" in Spanish). It now says the campaign would go out now. A scheduled campaign shows the time it's scheduled for.

✅ Conclusion

NetSendo v2.8.4 makes Brain more reliable for busy accounts:

• Token budgets now apply to each task, so one heavy task can't block the rest of your day.
• The limit is configurable in the panel and through task_token_limit in the API.
• Tasks that hit the limit stop safely: no further AI calls, no partial sends and no retries.
• Approval summaries and dry-run messages are clearer.

Because NetSendo is self-hosted, you control your data, your AI settings and your sending volume, with no vendor lock-in.