Every feature, fix, and improvement — documented as it ships. Subscribe to get notified.
May 2026
3 days ago
v0.9.0NewImproved
AI Agent Orchestration & Groq Flash
This release ships the long-awaited agent routing engine — route conversations to FAQ, Sales, Support, or Handoff agents based on confidence scoring. Plus: Groq Flash is now available at sub-100ms latency.
agent-routing-diagram.png
⚡
Agent orchestration engine agents
Route messages to specialised agents (FAQ, Support, Sales, Handoff) using confidence scoring. Configure thresholds per-bot in the Agents settings screen.
🚀
Groq Flash provider providers
Added Groq's llama-3.3-70b-versatile and mixtral-8x7b models. Average latency under 80ms in testing — ideal for high-volume widgets.
📊
Analytics heatmap analytics
New 7-day hourly heatmap shows peak conversation times. Drill into any hour to see top topics and resolution rate.
🎨
Widget customisation expanded widget
Widget now supports custom fonts, border radius, and a pre-chat form to collect visitor name and email before the conversation starts.
Apr 2026
4 weeks ago
v0.8.0NewFixed
Webhooks, n8n & Zapier integrations
Fire webhooks on any conversation event and connect to your favourite automation tools. Plus a major overhaul of the conversation logs viewer.
🔗
Webhook engine integrations
Configure webhooks for message.received, conversation.resolved, handoff.triggered and more. Includes retry logic and delivery logs.
⚙️
n8n & Zapier native connectors integrations
Pre-built trigger and action nodes for n8n and Zapier. Route conversations to HubSpot, Slack, Notion, or any of 5,000+ apps.
📋
Conversation log viewer v2 logs
Full transcript view, source citations per message, token/cost breakdown, and one-click export to CSV or JSON.
🐛
RAG context truncation bug rag
Fixed an issue where large PDFs caused context chunks to be silently truncated, leading to incomplete answers.
Mar 2026
2 months ago
v0.7.0New
Team roles & multi-workspace support
Invite your team, assign Admin / Editor / Viewer roles, and manage multiple workspaces from a single account. Built for agencies and growing teams.
👥
Role-based access control team
Invite teammates via email. Admins can manage billing and API keys; Editors build bots; Viewers read analytics only.
🏢
Multiple workspaces platform
Create and switch between workspaces from the sidebar. Perfect for agencies managing multiple client accounts.
🔑
API key management UX api
Redesigned API keys screen with scoped permissions, last-used tracking, and one-click revoke.
Feb 2026
3 months ago
v0.6.0ImprovedBreaking
Real-time analytics dashboard
Live metrics dashboard with cost tracking, sentiment analysis, and resolution rates. Breaking: analytics API response shape changed — see migration notes.
📈
Live analytics dashboard analytics
Real-time charts for message volume, cost per model, resolution rate, and sentiment. Data updates every 30 seconds via SSE.
😊
Sentiment analysis analytics
Automatic sentiment scoring on every conversation. Surface frustrated users before they churn.
⚠️
Analytics API v2 breaking
GET /analytics/overview now returns a data wrapper. Update any direct API clients. v1 endpoint deprecated March 2026.
Jan 2026
4 months ago
v0.5.0New
Embeddable widget ships
The vanilla-JS embeddable chat widget is live. Copy one script tag, paste it anywhere, and your AI bot is live on any website in seconds.
⊞
Embeddable widget v1 widget
Zero-dependency vanilla JS widget. Under 8KB gzipped. Customise accent colour, position, greeting, and avatar from the dashboard.
🌐
Public widget API api
New unauthenticated POST /widget/:botId/chat endpoint with per-IP rate limiting and CORS open for any origin.
Dec 2025
5 months ago
v0.4.0New
RAG pipeline & knowledge base
Upload PDFs, scrape URLs, write FAQs. Qdrant-powered semantic retrieval injects relevant context into every AI response automatically.
📚
Knowledge base with RAG knowledge
Upload PDFs and DOCX files. Scrape any URL. Write FAQs inline. All content is chunked, embedded with text-embedding-3-small, and indexed in Qdrant.
🔍
Semantic search retrieval knowledge
Top-k retrieval injects the most relevant chunks into every LLM system prompt. Source citations appear in conversation logs.
Stay ahead of the curve
New features ship every week. Subscribe to the changelog.