Overview
The Q-AI Bot is an AI assistant layered on top of Texter: when switched on for a chat, an OpenAI-backed assistant reads each incoming message, looks up answers in a knowledge base you control, replies in natural language, and decides, turn by turn, whether to keep going or hand back to your Texter bot or a human agent.
It is not a replacement for your Texter bot but a specialist you route a conversation to: the YAML bot handles structured menus and flows, and when a conversation needs open-ended, knowledge-grounded answers, the bot flips the AI on for that chat until it is done. It is configured per project: each business's Texter environment has its own knowledge base, system prompt, and settings, so two businesses get two completely different assistants.
If you have never worked with a Texter bot YAML file, start with the YAML Overview. The Q-AI Bot plugs into that same bot, so the concepts there (nodes, on_complete, funcs) carry over.
Why it exists
A traditional menu-driven bot is great at predictable paths ("press 1 for sales") but struggles with free-form questions. The Q-AI Bot fills that gap:
- It answers in the contact's own words. No menus to navigate: the contact asks, the assistant replies.
- It is grounded in your content. Answers come from your knowledge base, so the assistant speaks for your business, not the open internet.
- It knows when to step aside. It can resolve a simple chat on its own, or escalate to a human the moment a conversation needs one.
- Its replies are structured, not just text. Every answer comes back in a fixed shape your bot flow can act on, and a project can extend that shape (see Response Schema).
Key terms
These words appear throughout this section; skim them now.
| Term | What it means |
|---|---|
| AI mode | The on/off switch on a chat that decides whether the Q-AI Bot is in control. On = the assistant answers; off = the Texter bot answers. (Other pages may note its internal name, external-bot mode, same thing.) |
| Session | One continuous run of the assistant on a chat, from the moment AI mode turns on until it turns off. A session has its own memory; ending and restarting it starts fresh. |
| Turn | One incoming message answered by one assistant reply. A session is made of turns. |
| Knowledge base | The searchable store of your project's content that the assistant pulls answers from (help articles, price lists, policies, website pages). |
| RAG | Retrieval-Augmented Generation: the technique the assistant uses to answer: it retrieves relevant snippets from the knowledge base and generates an answer grounded in them. |
| Termination reason | A short label recorded when a session ends (for example resolved or handed to a human) that tells your bot and reports why the assistant stepped away. |
| Handoff | What happens when a session ends: AI mode switches off and control returns to your Texter bot or a human agent. |
| Scenario | A small, event-driven rule in your Texter environment that reacts to chat events. The Q-AI scenarios are what wire the Texter side to the assistant. |
Where an AI turn sits in a conversation
A contact's message arrives through a channel (for example WhatsApp) and is shown in Texter, handled either by the Texter bot (your YAML: menus, flows, CRM lookups) or, only while AI mode is on, by the Q-AI Bot. A single AI turn is walked through on How It Works; the full session arc is on Conversation Lifecycle.
The four subsystems
The Q-AI Bot is made of four parts. You don't operate them by hand (background automation workflows run them), but knowing them makes the rest of this section easier to follow.
| Subsystem | What it does |
|---|---|
| The live reply loop | The live path: each incoming message becomes an AI turn and routing decides whether to continue or hand off. The heart of the feature. |
| Re-engagement (Abandoned Bot) system | A per-project ladder of timed follow-ups for contacts who go quiet mid-conversation. |
| Knowledge-base ingestion | How your content gets into the assistant's knowledge base: uploaded files and scraped website pages. |
| Provisioning + reporting | Setting up a new project's assistant, plus the reports that track what it said and what it cost. |
How to read this section
These pages build on each other; if you are new, read them in order:
- How It Works: one conversation walked through end to end, from switching the AI on to the reply being sent.
- Conversation Lifecycle: how a session starts, the ways it ends, what handoff means, and the Texter scenarios that wire it all together.
- Response Schema: the structured reply the assistant returns on every turn, and how the bot YAML reads it.
- Abandoned Bot System: how quiet conversations are re-engaged and eventually closed.
- Knowledge Base overview: how your content becomes answers, including the file pipeline and the website-scraping pipeline.
- RAG & OpenAI Concepts: the under-the-hood mechanics, chunking, embeddings, vector search, and the model settings that shape answers.
To provision the AI for a real project, use the Onboard AI Bot tool.