Dashboard
Real interaction data from dotdotduck's own site, used as a working example of what dddk can capture for your product. The point isn't how large these numbers are — it's what you get on your site once dddk is wired in.
Agent run satisfaction
yes / no / dismiss signal collected at the end of every agent run with onLoopEnd: feedback. One explicit "satisfied" is one clean labelled training row — pipe this straight into a downstream RL loop, no inference from clickstream needed.
No agent_feedback rows yet — runs that opt into onLoopEnd: { kind: 'feedback' } will populate this.
Proactive — explicit yes/no signal
Yes / no / dismiss ratio per prompt. Marked low-confidence when n < 30 — treat these as directional, not decision-grade, until they ripen.
No proactive responses yet — waiting on traffic.
Geography
No data yet
Top palette items
No data yet
Top paths
No data yet
AI interactions
By trigger
No data yet
Agent runs
No data yet
Stop reasons
No data yet
Voice input
No voice attempts yet
Recent events (sanitized)
Exactly the rows this demo page sees. For privacy reasons we deliberately don't capture IP, raw question text, or answer bodies on this public showcase — this is a capability demo, not a hard SDK constraint, and your own host can opt into more fields when integrating for real analytics.
No data yet
Capability surface — what dddk can collect / what this demo shows / what this demo skips for privacy
- +Every user intent — palette opens, voice transcripts, Dwell selections, skill start/finish, agent ask/answer, confirm outcomes for destructive actions
- +Agent run lifecycle — started, completed, stop reasons, mode switches, tool failures, continue/stop on every pause
- +End-of-run agent_feedback — satisfied / not / skipped, or ask_user multi-choice answers
- +Anonymous session + visitor id (UUIDs, no PII)
- +Locale, device class, viewport
- +Country / region from Cloudflare header (IP never stored)
- +Skill / palette item id, via source (gesture / voice / mobile-button)
- +Answer length, latency, duration, size_chars
- +Proactive prompt yes / no / dismiss / shown (labelled RL signal)
- +Any host-defined custom event you choose to track
- ✓Sessions / visitors / events / palette-opens (overview)
- ✓Headline yes-rate + daily trend + per-skill breakdown (agent_feedback)
- ✓Per-prompt yes-rate with confidence flag
- ✓Top countries by visitor count
- ✓Top palette items + top paths
- ✓Skill start / completion rate
- ✓Webagent answered count + avg chars + avg latency + via breakdown
- ✓Agent run stats + stop-reason breakdown + pause decisions
- ✓Voice attempts / success rate / error codes
- ✓Last 10 raw rows (sanitized, audit-visible)
- ✓CSV / JSON export with date / locale / device filters
- −Raw user question text
- −AI answer bodies
- −Selected text / voice transcripts
- −IP address / cookies / third-party trackers
- −PII (email, phone, name, etc.)