I’ve been building an embeddable AI agent that turns any website into an interactive, context-aware experience.
Unlike typical FAQ bots or generic chat assistants, this one doesn’t just read your PDFs or docs, it also learns from how visitors interact with your site and adapts its responses, recommendations, or offers accordingly.
Technically:
• Runs fully client-side (no external API calls or backend dependencies)
• Uses RAG pipelines to reference your own PDFs, documentation, or content base
• Employs local embeddings to personalize context and answers per visitor session
• Supports custom styling and can be embedded anywhere on the page
Early results (from SaaS testers):
• +25–30% increase in trial signups
• ~20% drop in support tickets
• Self-serve setup in under 2 minutes
I’d love technical feedback, especially around how it handles embeddings, on-site personalization logic, or UX flow. Happy to share implementation details or discuss architecture trade-offs.
Comments URL: https://news.ycombinator.com/item?id=45822367
Points: 1
# Comments: 0
Source: orchis.app
