Chatbot Configuration

The chatbot detail page holds every setting that shapes how your bot answers and presents itself. Open Dashboard → your chatbot → Configuration tab. A typical first setup takes five minutes: name the bot, pick a provider, set your service name, add your domain, and (on paid plans) write a short system prompt. Each field is explained below.

AI Provider Selection

The provider is the company whose AI runs your chatbot. Each chatbot can use its own provider and model, so you can run a fast, cheap bot on one site and a smarter one elsewhere. Pick the dropdown, then choose a model from the second dropdown.

ProviderModelsNotes
Gemini (default)gemini-2.5-flash, gemini-3.1-flash-lite, gemini-3-flash, gemini-3.5-flashGoogle's AI models. Used by default for new chatbots.
OpenAIgpt-5.4-nano, gpt-4.1-miniOpenAI's models
xAI Grokgrok-4-1-fast-reasoningxAI's fast reasoning model
Mistralmistral-medium-latest, mistral-small-2506EU-based AI provider (France)
Claude(coming soon)Anthropic's Claude models
The exact model lineup changes as providers release new versions. The list in the dashboard dropdown is always the authoritative one — treat the table above as a snapshot.

Model Selection

The model is the specific AI version within a provider. The dropdown shows only models for your selected provider, each tagged with a speed hint (⚡ to ⚡⚡⚡). When you switch provider, a new bot starts on that provider's first model (Gemini defaults to gemini-2.5-flash).

  • Smarter models (e.g. gemini-3.5-flash, gpt-4.1-mini) write more nuanced answers but can be a touch slower
  • Faster models (e.g. gemini-2.5-flash, gpt-5.4-nano, grok-4-1-fast-reasoning) reply quicker and cost less per message — a good fit for FAQ-style bots

If you're not sure, start with the default Gemini model and only switch if answers feel too slow or not sharp enough.

Switching the provider (not just the model within the same provider) clears the indexed content of all your data sources, because each provider uses a different, incompatible embedding format. Every source is flagged "Re-indexing required" (orange badge) and you must re-index them yourself — until you do, the chatbot can't answer from your content. See Re-indexing.

Service Name

Enter your business or service name (e.g. "ACME Corp" or "TechSupport Pro"). The chatbot uses this to:

  • Identify itself when asked who it represents
  • Provide contextually relevant answers about your organization
  • Reference your business name naturally in conversations

Example: If your service name is "CloudStore", the chatbot might say: "At CloudStore, we offer free shipping on all orders over $50."

Chatbot Name

An internal name used to identify your chatbot in the dashboard. This name is not shown to visitors — it's only for your own reference when managing multiple chatbots.

Examples: "Website Support", "Sales Bot EN", "FAQ Chatbot"

Allowed Domains

Select which domains are permitted to load the chat widget. This is a security-critical setting.

  • Only domains you control should be listed
  • The widget will not load on unlisted domains
  • This prevents unauthorized sites from embedding your chatbot
  • Add all domains where you want the widget to appear (e.g. example.com, shop.example.com)
If the widget doesn't appear on your website, check that the domain is listed in Allowed Domains. The domain must match exactly (including subdomains).

Timezone

Sets the chatbot's time awareness for accurate time-based responses. This affects:

  • Business hours references ("We're open from 9 AM to 5 PM")
  • Appointment scheduling (when using the Appointments tool)
  • Time-sensitive greetings ("Good morning!")

Select the timezone that matches your business location.

Custom Role (System Prompt)

The custom role defines your chatbot's personality, behavior, and boundaries through a system prompt — the standing instructions the AI follows in every conversation. This is a paid feature (available on a Basic plan or higher). The prompt can be up to 5,000 characters.

What to Include

  • Role definition: What the chatbot is and what it does
  • Tone and style: Formal, casual, friendly, technical
  • Boundaries: What topics to avoid, when to escalate
  • Domain knowledge: Specific terminology, rules, or processes
  • Response format: Preferred answer length, use of lists, etc.

Example System Prompts

Customer Support Bot:

You are a friendly customer support agent for CloudStore, an online electronics retailer. 
Always be helpful, patient, and professional. Answer questions about our products, 
shipping policies, and return process. If a customer seems frustrated, acknowledge 
their concern before providing a solution. For billing disputes or account issues, 
suggest contacting our support team at support@cloudstore.com. Never discuss competitor 
products or provide personal opinions.

Technical Documentation Bot:

You are a technical documentation assistant for DevTools API. Provide accurate, 
concise answers about our API endpoints, authentication, and SDKs. Always include 
code examples when relevant. Use technical language appropriate for developers. 
If you're unsure about a specific implementation detail, say so and recommend 
checking the official docs. Never generate fake API endpoints or example responses.

Sales Assistant:

You are a sales assistant for FitGear, a sports equipment brand. Help visitors 
find the right products based on their needs. Ask clarifying questions about their 
sport, skill level, and budget. Highlight product benefits and current promotions. 
If they're ready to buy, guide them to the product page. Always mention our 
30-day return policy to reduce purchase anxiety.

Tips for Effective Prompts

  • Be specific — Vague instructions lead to inconsistent behavior
  • Set boundaries — Clearly define what the chatbot should and shouldn't discuss
  • Include examples — Show the chatbot how you want it to respond in specific scenarios
  • Test and iterate — Try different prompts and refine based on actual conversations
  • Keep it focused — Overly long prompts can confuse the model; aim for clarity over completeness