Skip to content

Core Concepts

Understanding these building blocks will make every other feature in FRENZY.BOT easier to use. This page explains how the system works in plain English.

How FRENZY.BOT works

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│  Your Data  │ ──→ │   Indexing   │ ──→ │   Qdrant    │
│ Files, URLs │     │  (Embeddings)│     │(Vector DB)  │
│    FAQs     │     └──────────────┘     └──────┬──────┘
└─────────────┘                                 │
                                                │ retrieve
┌─────────────┐     ┌──────────────┐     ┌──────┴──────┐
│   User      │ ──→ │  FRENZY.BOT  │ ──→ │  AI Model   │
│  Question   │     │  (RAG Engine)│ ←── │(via OpenRouter)│
└─────────────┘     └──────┬───────┘     └─────────────┘
                           │
                    ┌──────┴──────┐
                    │   Response  │
                    │ + Lead Data │
                    │ + Analytics │
                    └─────────────┘
  1. You upload your data — files, website URLs, and FAQ pairs go into the Knowledge Base.
  2. The system indexes it — content is converted into numerical embeddings and stored in Qdrant (vector database) so it can be searched by meaning, not just keywords.
  3. A user asks a question — through the widget, WhatsApp, Telegram, or any connected channel.
  4. FRENZY.BOT retrieves relevant content — the system finds the most relevant pieces of your knowledge base using semantic search.
  5. The AI model generates a response — the retrieved content + the user's question are sent to the AI model (Claude, GPT, Gemini, etc.) via OpenRouter.
  6. Everything is tracked — the conversation, response time, token usage, and any captured lead data are stored for reporting.

Bot

Your bot is the AI assistant you're building. Each bot has its own:

  • Knowledge base — The documents, URLs, and FAQs it learns from.
  • Settings — AI model, system prompt, temperature, branding, widget configuration.
  • Conversations — Chat history with users.
  • Leads — Contacts captured during conversations.
  • Analytics — Usage metrics and funnel data.
  • Widget — Its own embed code with unique data-bot-id.

In a multi-bot setup, you can run multiple bots from a single installation — each completely isolated from the others.

Where to manage: Settings, Knowledge Base, Embed, and the workspace dropdown in the header.

Knowledge Base

The knowledge base is the library your bot reads before answering any question. It's what makes your bot specific to your business instead of a generic AI.

Three data sources:

Source Best for Indexing speed
Files (PDF, DOCX, TXT, CSV, etc.) Internal documents, policies, product specs Minutes
URLs (any public webpage) Website content, help articles, blog posts Minutes
FAQs (manual Q&A pairs) Critical answers you want to control exactly Instant

Key principle: If a topic isn't in the knowledge base, the bot may answer generically or decline to answer. The more relevant content you add, the better the bot performs.

For the full guide, see Knowledge Base.

Channels

Channels are the places where your users interact with the bot.

Channel Type Setup
Website widget Primary Embed script on your site
WhatsApp Messaging WAHA Docker + QR code pairing
Telegram Messaging Bot token from @BotFather
Discord Community Bot token from Developer Portal
Instagram Social Page Access Token
Messenger Social Facebook Page integration
Slack Workplace App token
Webhook Custom Any HTTP endpoint

Each channel can have its own reply style — the AI adapts its tone and response length to match the platform (e.g., shorter on WhatsApp, more detailed on the website widget).

For details, see Channels.

Sessions & conversations

A session is one complete conversation between a user and your bot. Each session tracks:

  • Messages — Every question and response in the conversation.
  • Tokens — How many AI tokens were consumed (affects cost).
  • Duration — How long the conversation lasted.
  • Source — Which channel the conversation came from.
  • Lead data — Any contact information captured during the chat.

Sessions are the foundation of analytics, quality review, and lead capture. Review them in Conversations.

Leads & integrations

A lead is a contact captured during a chat — name, email, phone, company, and any custom fields you've configured.

Leads flow through a pipeline:

New → Contacted → Qualified → Converted (or Lost)

Integrations automatically route leads to your external tools:

  • Email — SMTP notifications to your team.
  • Webhook — JSON payloads to your own endpoints.
  • Google Sheets — Append rows for lightweight CRM.
  • MailChimp — Add contacts to email marketing lists.

For details, see Lead Management and Integrations.

Modules

Modules are optional capabilities you can enable or disable without affecting the core system:

Module What it does
Live Web Access (Tavily) Lets the bot search the live web when internal data isn't enough
Human Handoff Escalates conversations to human agents
Text-to-Speech Converts AI responses to audio playback
Appointment Booking In-chat scheduling via Calendly
Lead Form Captures contact information mid-conversation
WhatsApp Automated messaging via WAHA

Browse and manage modules in Modules.

Quick self-check

Use this checklist to verify your bot is fully operational:

  • [ ] Knowledge base populated — At least one file, URL, or FAQ added and showing Live status.
  • [ ] OpenRouter connected — AI Engine shows a connected provider and selected model.
  • [ ] Widget deployed — Embed script is on your website and the chat bubble appears.
  • [ ] Integrations enabled — Leads are being routed to your CRM, email, or webhook.

If all four are true, your system is fully functional and ready for production traffic.