RAG & training
Hybrid retrieval, semantic cache, web scraper v2 and SSE training on websites, documents and Q&A.
VYBR!S
Organisations and agencies deploying AI chatbots for support, sales and bookings need more than a standalone widget: training on their own content, quality monitoring, live handoff to staff, CRM integrations and insight across multiple bots. VYBR!S was built as a multi-tenant SaaS platform — from RAG training to fleet analytics — and runs in production on AWS EC2.
What VYBR!S delivers for administrators, agencies and end clients.
Hybrid retrieval, semantic cache, web scraper v2 and SSE training on websites, documents and Q&A.
Moderator queue with SSE: the chatbot escalates to a human agent when answers or exceptions need a person — AI with limits, not a black box.
Kanban pipeline, lead qualification, email sequences and rules triggered by intent or sentiment.
WhatsApp, Slack, Telegram plus HubSpot, Salesforce, Zendesk, booking and webhooks.
Organisations and workspaces, white-label branding, RBAC, MFA, SSO and SCIM 2.0.
Fleet overview with sessions, CSAT, confidence and cache hit rate — plus per-bot drill-down.
End-to-end platform — from RAG pipeline to admin portal at vybris.com.
FastAPI backend with PostgreSQL, Redis and Qdrant. ARQ workers for training, crawls and billing. Production deployment on AWS EC2 with Docker Compose and nginx.
BGE-M3 embeddings, AWS Bedrock as LLM, hybrid retrieval (semantic + BM25) with reranking and semantic response cache in Redis for faster, consistent answers.
React, TypeScript and Vite with TanStack Query. Production portal at vybris.com plus admin hub: chatbot management, knowledge hub, test chat, embed widget and billing.
Conversation Studio, golden eval cases with eval gate, guardrails, PII redaction, audit logging and MFA/SSO/SCIM for enterprise environments.
/performance page with five KPI cards, activity chart, CSAT-vs-confidence comparison and per-bot drill-down to detailed analytics and conversations.
Privacy and confidentiality of customer data.
All figures, bot names and charts in the screenshot above and on the live dashboard are synthetic demo data. They come from an idempotent seed script (seed_demo_analytics.py) with 10 demo chatbots and varied performance profiles — so real customer data is never visible in demos or case studies.
Live dashboard shows demo data — not real customer metrics.
Platform designed and built by ultimAItech.