Skip to main content

AI Settings of LiveChatAI

AI Configuration Settings

Configure how your chatbot behaves, which models it uses, and how it handles content. Go to Settings → AI Configuration to access these four sections:

  1. Identity
  2. Behavior & Instructions
  3. LLM Models
  4. Knowledge & Safety

Identity

Define how your chatbot appears in the dashboard and in conversations.

  • Chatbot Display Name
    Enter an internal name (e.g., “SupportBot”) to help you identify this bot in the workspace.
  • Company Name
    Enter the name (e.g., “Acme Corp”) the chatbot uses when referring to your organization or products. This is shown to end users when the bot says “Our company…” or introduces itself.

Tip: Choose a friendly, on-brand company name so responses feel cohesive and professional.

Behavior & Instructions

This section allows you to define how your AI chatbot should think, sound, and act.

Setting this correctly helps the chatbot stay aligned with your brand tone and customer service policies, without needing to repeat the same guidelines in every data source.

  • Base Prompt: This prompt runs at the start of every conversation. Keep it concise (2–3 sentences) and focused on overarching guidelines.
💬
Examples

“You are a helpful and friendly support assistant for [Your Company]. Keep answers clear, concise, and professional.”

“Speak with a casual, warm tone and assist users with product-related questions. Don’t guess—only respond when confident.”
  • Custom Instructions (optional): Add or update any additional rules, policies, or tone notes here (e.g., “Always greet users with a friendly welcome” or “If you’re unsure, refer to our returns policy.”).

LLM Models

Choose which large-language model powers your chatbot. Below is a list of available models and their ideal use cases:

Model

Recommended Use Case

Message Credit Cost

GPT-4o

In-depth technical support (APIs, code troubleshooting); multi-turn dialogs requiring nuanced context (billing disputes, complex workflows)

1

GPT-4o-mini

Standard product/service FAQs where nuance matters; balances accuracy and lower latency

1

O3-mini

High-volume support flows (simple shipping/account questions); prototypes or internal bots where cost and speed are top priorities

2

O1

General customer service (retail, hospitality); educational or knowledge-sharing chatbots with moderate complexity

20

Claude 3 Sonnet (Beta)

Regulated or sensitive industries (healthcare, finance); scenarios requiring extra content filtering and policy compliance

1

Claude 3.5 Sonnet (Beta)

Regulated industries needing faster throughput than Claude 3; medium-length policy/procedural customer engagement

1

Llama 3.1 70B

Internal knowledge bases where all data is owned (intranets, private docs); cost-sensitive deployments needing strong reasoning

1

Llama 3.1 405B

Advanced R&D or research assistance (engineering/scientific queries); technical documentation bots for specialized audiences

1

Llama 4 Scout 17B

Customer support where speed + cost are priorities but quality remains high (e-learning, marketing FAQs); resource-constrained environments (mobile apps)

2

Llama 4 Maverick 17B FP8

High-traffic support sites needing sub-second responses; simple-to-moderate complexity chat where minor reasoning trade-offs are acceptable

1

DeepSeek R1

Data-driven retrieval tasks (live dashboards, real-time metrics Q&A); enterprise knowledge graphs or databases requiring on-demand lookup

1

DeepSeek V3

Large enterprise knowledge management (thousands of documents); quick retrieval of specific data points (legal clause lookup, policy search)

1

Gemini 2.5 Pro

Developer portals needing code snippet recommendations or bug diagnosis; multi-modal scenarios combining text and images (hardware troubleshooting)

2

Grok-3 (Latest)

Real-time, trend-driven chatbots (stock prices, event info); public-facing support needing concise, up-to-date answers

4

Note on “Message Credit Cost”:

  • This column indicates the number of credits consumed per full user‐AI exchange (one user query + one AI response).
  • Always verify your plan’s credit allowances when switching models, especially for high-cost engines like O1 (20 credits/request).

Once you’ve chosen your model, click Save Changes to apply it immediately.

Input/Output Length Settings

  • Max Message Length (Input)
    Choose Short, Medium, or Long to cap the tokens users can send in a single message.
  • Max Response Length (Output)
    Select Short, Medium, or Long to limit how verbose the AI’s replies can be.

Tip: If users tend to send long paragraphs, set Max Message Length to “Long.” To avoid overly wordy answers, keep Max Response Length at “Medium” or “Short.”

After selecting your model and length settings, click Save Changes to apply immediately.

Knowledge & Safety

Control how the chatbot uses your data sources and what it says when it cannot find an answer.

  • Restricted to Data Source

When enabled, the bot will only answer using your uploaded files, website crawls, Q&As, or other connected sources. It will not rely on its general LLM knowledge.

Turn this on if you require 100% brand-specific answers without any external assumptions.

  • Custom Fallback Message

Enter the exact text the bot sends when it can’t find an answer in your sources (e.g., “I’m sorry, I don’t have that information. Please contact support at [email protected].”)

This prevents generic, off-brand “I don’t know” replies.

Warning: If “Restricted to Data Source” is checked, Custom Fallback Message is required. Otherwise, the chatbot will have no fallback response.

Saving & Applying Changes

  • After updating any section—Identity, Behavior & Instructions, LLM Models, or Knowledge & Safety—click Save Changes at the bottom of that panel.
  • Changes take effect immediately for all new conversations.

For further assistance with any AI Configuration setting, email [email protected].