Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.sxth.ai/llms.txt

Use this file to discover all available pages before exploring further.

​
Nudges

Nudges are proactive suggestions generated based on user state. They help re-engage users and surface timely insights.

​
How Nudges Work

  1. Adapter defines templates — What nudges are possible in your domain
  2. Engine checks triggers — Based on state (inactivity, momentum drop, patterns)
  3. Nudges are generated — Stored with status pending
  4. Your app delivers them — Email, push notification, in-app banner
  5. User acts or dismisses — Status updated to acted or dismissed

​
Defining Nudge Templates

In your adapter: 
def get_nudge_templates(self) -> dict:
    return {
        "stalled_deal": {
            "title": "Deal may be stalling",
            "template": "No activity on {deal_name} for {days} days. Time to re-engage?",
            "priority": 7,  # 1-10, higher = more important
        },
        "momentum_drop": {
            "title": "Momentum slipping",
            "template": "Activity on {deal_name} has dropped. What's blocking progress?",
            "priority": 8,
        },
        "streak_risk": {
            "title": "Don't break your streak!",
            "template": "You're on a {streak_days}-day streak. Keep it going!",
            "priority": 6,
        },
        "milestone": {
            "title": "Nice progress!",
            "template": "You've completed {count} sessions. Celebrate!",
            "priority": 3,
        },
    }

​
Using the Nudge Engine

from sxth_mind.engine import BaselineNudgeEngine

# Create engine with adapter and storage
engine = BaselineNudgeEngine(mind.adapter, mind.storage)

# Check and generate nudges for a user
nudges = await engine.check_and_generate("user_123")

for nudge in nudges:
    print(f"[{nudge.priority}] {nudge.title}: {nudge.message}")

​
Retrieving Pending Nudges

# Via Mind
nudges = await mind.get_pending_nudges("user_123")

# Via HTTP API
# GET /nudges/{user_id}

​
Nudge Lifecycle

class Nudge:
    id: str
    nudge_type: str          # Template key
    title: str
    message: str
    priority: int            # 1-10
    status: str              # pending, delivered, dismissed, acted
    project_mind_id: str
    scheduled_for: datetime | None
    delivered_at: datetime | None
    acted_at: datetime | None
    created_at: datetime
    updated_at: datetime

​
Status Flow

pending → delivered → acted
                   ↘ dismissed

​
Updating Status

Via HTTP API: 
# Mark as dismissed
POST /nudges/{nudge_id}/dismiss

# Mark as acted upon
POST /nudges/{nudge_id}/act

​
Trigger Types

The baseline engine checks for:
TriggerDescription
InactivityUser hasn’t interacted in X days
Momentum dropActivity level has decreased significantly
Pattern matchSpecific patterns detected (adapter-defined)
Stage-basedCertain journey stages trigger nudges

​
Customizing Triggers

Override the engine for custom trigger logic: 
from sxth_mind.engine import BaselineNudgeEngine

class MyNudgeEngine(BaselineNudgeEngine):
    async def check_and_generate(
        self,
        user_id: str,
        project_id: str | None = None,
    ) -> list[Nudge]:
        nudges = []

        # Get user state
        user_mind = await self.storage.get_user_mind(user_id)
        if not user_mind:
            return nudges

        # Custom trigger: user hasn't logged in for 7+ days
        if user_mind.days_since_last_interaction > 7:
            nudges.append(self._create_nudge(
                "comeback",
                user_mind,
                project_mind=None,
                days=user_mind.days_since_last_interaction,
            ))

        # Add more custom triggers...

        return nudges

​
Best Practices

Limit nudge frequency. Users will mute or ignore if you nudge too often.
High-priority nudges should be rare and important. Most should be 3-6.
Each nudge should have a clear action the user can take.
If a user dismisses a nudge type repeatedly, consider reducing frequency.
Monitor acted vs dismissed rates to improve nudge templates.