Scheduled Tasks

A real personal assistant doesn't wait for you to ask. Kraken can run tasks on a cron schedule — no external scheduler needed. Tell it "summarize my GitHub notifications every morning" or "check this RSS feed weekly" and it just does it. Scheduled tasks execute as full conversations, with access to the knowledge graph, skills, and all tools.

Creating a Schedule

from kraken import KrakenClient

client = KrakenClient(api_url="http://localhost:8080")

client.sessions.create(session_key="schedules")  # optional: context session

schedule = client.chat(
    "Create a schedule called 'daily-standup' that runs every weekday at 9am. "
    "It should summarize what I worked on yesterday.",
    session_key="schedules",
)

# Direct API — no chat needed
import httpx

httpx.post(
    "http://localhost:8080/v1/schedules",
    headers={"Authorization": "Bearer sk-..."},
    json={
        "name": "daily-standup",
        "cron_expression": "0 9 * * 1-5",
        "task_prompt": "Summarize what I worked on yesterday and list priorities for today.",
        "metadata": {"channel": "slack-standup"},
    },
)

curl -X POST http://localhost:8080/v1/schedules \
  -H "Authorization: Bearer $KRAKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "name": "daily-standup",
    "cron_expression": "0 9 * * 1-5",
    "task_prompt": "Summarize what I worked on yesterday and list priorities for today.",
    "max_runs": 260
  }'

Cron Expressions

Standard 5-field cron format:

┌───────────── minute (0–59)
│ ┌───────────── hour (0–23)
│ │ ┌───────────── day of month (1–31)
│ │ │ ┌───────────── month (1–12)
│ │ │ │ ┌───────────── day of week (0–7, 0 and 7 = Sunday)
│ │ │ │ │
* * * * *

Common patterns

How Schedules Execute

When a schedule fires:

1. The worker creates a new session with the schedule's task_prompt as the message
2. If origin_session_id was set, the new session inherits context from that session
3. The agent processes the task with full access to:
   • Knowledge graph (all entities, relationships, communities)
   • Skills (relevant procedures loaded by similarity)
   • Tools (browser, code execution, etc.)
   • User model
4. The response is stored in the new session
5. The schedule's run_count increments
6. If max_runs is reached, the schedule is disabled

Practical Examples

Weekly Project Summary

{
  "name": "weekly-summary",
  "cron_expression": "0 17 * * 5",
  "task_prompt": "Generate a weekly summary: what entities were added to my knowledge graph this week, what skills were created or updated, and what are the key themes across my conversations?"
}

Daily Learning Digest

{
  "name": "learning-digest",
  "cron_expression": "0 8 * * *",
  "task_prompt": "Review my recent conversations and identify: (1) new technical concepts I discussed, (2) questions I asked that I might want to revisit, (3) any action items I mentioned but might not have completed."
}

Automated Research

{
  "name": "tech-radar",
  "cron_expression": "0 10 * * 1",
  "task_prompt": "Use the browser to check the latest releases of the tools and frameworks in my knowledge graph. Summarize any notable updates or breaking changes."
}

Knowledge Graph Maintenance

{
  "name": "graph-cleanup",
  "cron_expression": "0 3 * * 0",
  "task_prompt": "Review the knowledge graph for inconsistencies: duplicate entities with different names, stale relationships, or entities that should be merged. List suggested cleanups."
}

Managing Schedules

List all schedules

curl http://localhost:8080/v1/schedules \
  -H "Authorization: Bearer $KRAKEN_API_KEY"

Update a schedule

curl -X PATCH http://localhost:8080/v1/schedules/{id} \
  -H "Authorization: Bearer $KRAKEN_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "enabled": false
  }'

Delete a schedule

curl -X DELETE http://localhost:8080/v1/schedules/{id} \
  -H "Authorization: Bearer $KRAKEN_API_KEY"

Configuration

The schedule executor runs as part of the worker process. It checks for due schedules every minute.

Delivery Platforms

Schedule results are stored in sessions. To deliver results to external platforms (Discord, Slack, email), you have two options:

Option 1: Webhook integration

Create a tool that posts to your platform, then reference it in the task prompt:

{
  "name": "slack-standup",
  "cron_expression": "0 9 * * 1-5",
  "task_prompt": "Summarize yesterday's work and post it using the send-to-slack tool in the #standup channel."
}

Option 2: Poll from your integration

Your Discord bot or Slack app can poll Kraken for schedule results:

# In your Discord bot, check for new schedule outputs
sessions = kraken.sessions.list(limit=5)
for session in sessions:
    if session.metadata.get("schedule_name") == "daily-standup":
        # Forward to Discord channel
        ...