AI Customer Support
24/7 AI responses with seamless handoff to human agents—free your team to focus on what matters.
No waiting—ready to use
Trusted by teams
From models and knowledge bases to prompts, human collaboration, analytics, and logs—one cohesive system.
Configure multiple LLM and multimodal providers; visitors and admins share one place to manage models and usage—easy to swap vendors and control cost.
Ingest documents and retrieve with vectors so answers stay on-brand; replies can show whether KB, model, or web search was used for review and tuning.
Edit the prompt templates that power RAG, optional web search, and other flows across your scenarios.
Online presence, live sessions over WebSocket, seamless handoff and teamwork; embed the visitor widget on any site.
Daily or custom ranges for widget opens, sessions, messages, AI success/failure, KB hit rate, and more—see how operations are trending.
Persisted by category and event with trace_id and keyword search—trace critical paths and incidents for troubleshooting and audit.
A polished UI that makes day-to-day admin work easier.

加载中...
Three steps from the repo to the visitor widget.
Copy the .env template and fill required database and admin settings.
Bring up backend, frontend, and dependencies with Docker Compose (see README).
Mount the chat widget on your site; configure models and KB in the console, then go live.
Start from the open-source repo, or explore the online demo first.