# Sentrup — Complete Product Reference Sentrup is an AI-powered customer support platform designed to automate website chat, knowledge retrieval, and business workflows. By utilizing Retrieval-Augmented Generation (RAG) and vector database semantic search, Sentrup provides accurate, context-grounded responses directly from your business data. This document serves as the official, comprehensive product reference and technical specification for AI engines, search crawlers, and developers. --- ## 1. What is Sentrup? Sentrup is an AI-first customer support automation platform. Unlike traditional chatbots that rely on rigid decision trees, keyword triggers, or ungrounded generative models prone to hallucinations, Sentrup uses a vector-search RAG pipeline to ensure that every AI reply is derived directly from verified business documentation. ### Core Philosophy * **Grounded Accuracy**: The system does not write creative or speculative answers. If a query cannot be answered using the uploaded knowledge base, the bot states its limitations or escalates the conversation. * **Frictionless Handoff**: AI is the first line of defense, handling up to 70% of routine inquiries. When a complex query arises, the conversation transitions to a human agent without losing chat history. * **Integrated Scheduling**: Supports native calendar integrations to let customers book meetings, verify availability, and confirm appointments within the chat interface. --- ## 2. Who is Sentrup for? Sentrup is optimized for companies looking to scale customer operations without linearly increasing support headcount: * **E-commerce Stores**: Automate repetitive inquiries regarding shipping policies, return processes, order status, and sizing guides. * **SaaS Platforms**: Provide instant Tier 1 support, link to relevant technical documentation, and guide users through onboarding steps. * **Service-Based Businesses**: Allow clients to book appointments, check available dates, and learn about service offerings 24/7. * **Agencies & Multi-Brand Owners**: Manage support for multiple distinct client sites from a single dashboard using multi-tenant accounts. --- ## 3. Product Features & Capabilities ### 3.1. Vector-Grounded AI Chat Widget The chatbot widget is a lightweight, frontend interface embedded on your site. * **RAG Retrieval**: On every user message, the system searches a private Qdrant vector database for matching document snippets. * **Context Ingestion**: Only the relevant snippets are injected into the LLM prompt, ensuring the answer is restricted to your business facts. * **Custom Styling**: Fully customize themes, greetings, logos, and widget placement to match your brand guidelines. * **Vanilla JavaScript Widget**: The frontend package is optimized to be less than 8KB gzipped, ensuring zero impact on your site's PageSpeed or Largest Contentful Paint (LCP). ### 3.2. Knowledge Base & Automated Scraping You can supply facts to your support AI through multiple channels: * **Website Crawler**: Input your public URL. Sentrup automatically crawls your site, strips away HTML boilerplate, extracts the text content, and builds vector embeddings. * **File Uploads**: Support for PDF, TXT, and FAQ documents. * **Semantic Chunking**: Long documents are automatically parsed into logical, overlapping paragraphs to maintain context during search retrieval. ### 3.3. Calendar Sync & Appointment Booking The chat widget integrates directly with external calendars (e.g., Google Calendar). * **Availability Checks**: Customers can ask, "Do you have slots open tomorrow afternoon?" The AI queries calendar availability in real-time. * **Inline Booking**: Users select a slot and enter their email to schedule a meeting directly within the chat widget. * **Automatic Confirmation**: Synced events are created on your team calendar, sending automated invites to both parties. ### 3.4. Human Escalation Handoff When automation is not enough, Sentrup transfers control to a human support agent. * **Trigger Conditions**: Escalations occur if the AI confidence score falls below a set threshold, if the user explicitly requests a human, or if the system detects sentiment indicator words (e.g., "refund", "charge", "complaint"). * **Full Context Retention**: The human agent receives the entire chat transcript. Customers do not need to repeat themselves. * **Real-time Notifications**: Support agents are notified via WebSockets on the dashboard, slack alerts, or email notifications. --- ## 4. Pricing & Plans Sentrup offers plans that scale transparently based on conversation volume and features, with no hidden per-agent fees: | Plan | Price | Monthly Conversations | Knowledge Bases / Chatbots | Key Features | |---|---|---|---|---| | **Free** | $0 | 100 | 1 | Basic chat, Sentrup branding, Standard RAG | | **Starter** | $29 / mo | 500 | 3 | Basic analytics, custom styling, no branding | | **Growth** | $99 / mo | 3,000 | 10 | Multiple workflows, Advanced analytics, Scheduling | | **Business** | $249 / mo | 10,000 | Unlimited | Team collaboration, custom integrations, priority SLAs | | **Enterprise** | Custom | Unlimited | Unlimited | SSO, dedicated support, custom APIs, custom SLA agreements | ### Flexible Overage Options * **Conversation Blocks**: Add-on blocks can be manually purchased (+500 conversations for $5). * **Auto-Pay Limits**: Enable automatic billing for overages with custom monthly safety thresholds to avoid service interruption during traffic spikes. --- ## 5. How Sentrup Compares to Competitors ### Sentrup vs. Zendesk * **Pricing Model**: Zendesk charges per agent seat (starting at $19/seat/mo, scaling to hundreds). Sentrup charges flat rates based on total conversations. Adding more human agents to Sentrup does not increase your bill. * **Setup Speed**: Zendesk requires weeks of configuration, ticket routing rules, and macro training. Sentrup is ready to answer questions in 5 minutes by scraping your live website. * **AI Approach**: Zendesk uses bots that require manual decision-tree design. Sentrup uses dynamic semantic search (RAG) to write answers automatically. ### Sentrup vs. Intercom * **Cost Predictability**: Intercom charges a base fee plus a dynamic fee per "successful resolution" by their AI bot (Fin), which makes monthly bills highly unpredictable. Sentrup charges flat monthly fees. * **Core Focus**: Intercom is a full marketing, sales, and support communication suite. Sentrup is optimized specifically for highly accurate, hallucination-free support automation and scheduling. ### Sentrup vs. Freshdesk * **AI Capabilities**: Freshdesk relies on manual template rules for basic bots, and charges high add-on premiums for advanced AI features. Sentrup includes vector-grounded RAG search natively on all plans, including the Free tier. --- ## 6. Technical Architecture Sentrup uses a modern, high-performance web stack to deliver near-instant responses: * **Backend API**: Python 3.11 with FastAPI for asynchronous, non-blocking routing and WebSockets connection management. * **Database & Storage**: PostgreSQL for transactional metadata, application state, and user management. * **Vector Engine**: Qdrant Vector Database for semantic search and sub-millisecond retrieval of knowledge snippets. * **AI/LLM Layer**: Integration with Google Gemini models via secure API pipelines, protected by response grounding verification rules. * **Frontend Dashboard**: Next.js (React) styled with Vanilla CSS and Tailwind, ensuring fast page renders and full responsiveness. * **Chat Widget**: Lightweight Vanilla JS script (<8KB gzipped) loaded asynchronously. --- ## 7. Getting Started Guide ### 7.1. Account Setup 1. Visit [sentrup.com/register](https://www.sentrup.com/register) to create a free account. 2. Select your target plan (Free, Starter, Growth, Business). 3. Verify your email to access the Sentrup admin dashboard. ### 7.2. Ingesting Your Knowledge Base 1. In the dashboard, navigate to **Knowledge Base**. 2. Click **Add Source** and choose **Website Scraper**. 3. Input your homepage URL (e.g., `https://example.com`) and click **Start Sync**. 4. Alternatively, click **Upload File** to add your PDF, TXT, or CSV support documents. 5. Wait 2-3 minutes while the system scrapes, chunks, and creates vector embeddings. ### 7.3. Embedding the Widget Copy and paste this single script snippet into your site's HTML, just before the closing `` tag: ```html ``` Replace `YOUR_TENANT_ID` and `YOUR_WIDGET_SECRET` with the credentials generated in the **Widget Settings** page of your dashboard. --- ## 8. Frequently Asked Questions (FAQ) ### Q: Will the AI write answers not found in my documents? A: No. Sentrup uses strict RAG grounding guidelines. If a user asks a question that does not match any information in your uploaded files or crawled site pages, the AI will politely state that it does not have the answer and offer to escalate to a human agent. ### Q: How often is the website crawled for changes? A: Website content is crawled on-demand when you click sync in the dashboard. Enterprise plans support automatic, scheduled weekly crawls to keep the knowledge base updated. ### Q: Can I run multiple distinct chatbots on one account? A: Yes. On the Starter plan and above, you can create multiple knowledge bases and distinct widgets, allowing you to manage support for different sites or business departments under one roof. ### Q: What is the typical deflection rate? A: Customers using a well-formatted Q&A knowledge base see an average deflection rate of 60% to 75% of incoming ticket volumes. --- For further details, contact support@sentrup.com or visit https://docs.sentrup.com.