Features
Six core modules covering the entire AI customer service pipeline
Multi-Model AI Engine
Support for Claude, GPT, Gemini, Deepseek, Qwen, GLM and more. Switch providers on demand — no vendor lock-in. Unified abstraction via Vercel AI SDK — new models work out of the box.
- 6+ major LLM providers supported
- Per-agent model and parameter configuration
- Fine-tune temperature and max turns
- Auto-fallback to backup models
- API key encrypted storage (AES-256-GCM)
- Streaming output, first token latency < 500ms
Feature Illustration
Knowledge Base RAG
Upload enterprise documents for auto-vectorization (pgvector). AI answers precisely from your knowledge base. Supports PDF, Word, Markdown formats with intelligent chunking strategies and cosine similarity matching.
- Auto document chunking + vectorization
- pgvector cosine similarity search
- Multi-KB binding per agent
- BullMQ async document processing
- Retrieved context auto-injected into system prompt
- Knowledge gap auto-detection and suggestions
Feature Illustration
Human-AI Collaboration
Real-time AI sentiment analysis + confidence scoring. Complex issues auto-escalate to human agents. Agents see full context with AI-suggested replies — zero disruption to customer experience.
- Sentiment detection triggers auto-escalation
- Low confidence auto-upgrades to human
- Full conversation context preserved on handoff
- Agents can view AI analysis summaries
- Skill-based intelligent agent routing
- WebSocket real-time message sync
Feature Illustration
Embeddable Widget
9.3KB ultra-lightweight Preact widget with Shadow DOM isolation — zero CSS conflicts. Embed with a single script tag on any website. Supports theme customization, multi-language, and image uploads.
- Only 9.3KB (gzip) — blazing fast load
- Shadow DOM complete isolation
- Theme color / position configurable
- Image upload and preview support
- Multi-language (CN/EN/JP/KR/TH)
- Auto-reconnect with exponential backoff
Feature Illustration
Enterprise IM Integration
Real-time push notifications to DingTalk, Feishu, WeChat Work. Agents stay informed without switching tools. Supports bidirectional interaction — reply to customers directly from IM.
- DingTalk Stream protocol real-time push
- Feishu WebSocket persistent connection
- WeChat Work callback integration
- Reply to customers directly in IM
- Custom notifications by event type
- Auto-silent during off-hours
Feature Illustration
Analytics Dashboard
Comprehensive tracking of customer service metrics: conversation volume, AI resolution rate, average response time, satisfaction scores. Filter by time periods with intuitive trend charts for data-driven optimization.
- Real-time core metrics dashboard
- Daily/weekly/monthly trend analysis
- AI vs human resolution rate comparison
- Satisfaction scoring (CSAT)
- Knowledge gap hotspot rankings
- Agent workload distribution
Feature Illustration
Still Hesitating? Try Free for 14 Days — Zero Risk
No credit card required. Sign up and start immediately. See how AI customer service saves your team time and money.