Skip to main content
The system prompt is the most important configuration for your chatbot. It tells the AI who it is, how to behave, and what rules to follow. Every response your chatbot generates is influenced by this prompt.

​
Where to Edit

Go to your chatbot → Settings tab → Model & Instructions sub-tab. The system prompt textarea supports up to 3,000 characters.

​
Default Prompt

When you create a chatbot, FIFE.BOT auto-generates a generic system prompt. You should customize it to match your specific use case.

​
Prompt Templates

FIFE.BOT includes 12 pre-built prompt templates in the dropdown above the system prompt field. Each template is available with preset wording in English and Czech (the chatbot UI and widget support 14 interface languages separately — see Widget appearance). Templates cover common use cases:
  • Customer support agent
  • Sales assistant
  • Technical documentation helper
  • HR / onboarding bot
  • Product guide
  • FAQ bot
  • And more
Select a template to populate the system prompt field. You can then edit it further.

​
Writing an Effective Prompt

​
Structure

A good system prompt has these sections: 
1. Identity — Who is the bot?
2. Tone & style — How should it communicate?
3. Scope — What topics can it cover?
4. Rules — What should it never do?
5. Fallback — What to do when it doesn't know?

​
Example: Customer Support Bot

You are a friendly customer support assistant for Acme Corp.

Tone: Professional but warm. Use simple language. Avoid jargon.

You help customers with:
- Order status and tracking
- Returns and refunds
- Product information
- Account issues

Rules:
- Never make up information. If you don't know, say so.
- Never share internal policies or pricing logic.
- Always suggest contacting human support for billing disputes.
- Keep responses concise — 2-3 paragraphs max.

When you don't have an answer, say:
"I don't have that information right now. Please contact our support team at support@acme.com."

​
Example: Technical Documentation Bot

You are a technical documentation assistant for the XYZ API.

Answer questions based strictly on the provided documentation.
Include code examples when relevant. Use markdown formatting.

Rules:
- Only answer questions about XYZ API.
- If a feature isn't documented, say "This isn't covered in the current documentation."
- Never generate code that uses deprecated endpoints.
- Always specify the API version in examples.

​
Best Practices

Be specific

“You are a support agent for Acme Corp’s e-commerce platform” is better than “You are a helpful assistant.”

Set boundaries

Explicitly define what the bot should NOT discuss. “Never discuss competitor products” or “Don’t provide medical advice.”

Define the fallback

Tell the bot what to do when it doesn’t know — redirect to email, phone, or a specific URL.

Control length

“Keep responses under 3 sentences” or “Provide detailed step-by-step explanations” — be explicit about response length.

​
Common Mistakes

MistakeProblemFix
Too vague”Be helpful” gives no directionSpecify the domain, tone, and rules
Too long3,000 chars of rules confuses the modelFocus on the top 5-10 most important rules
No fallbackBot makes up answers when unsureAdd explicit “when you don’t know” instructions
No scope limitsBot answers off-topic questionsDefine what topics are in/out of scope
Contradictory rules”Be concise” + “Explain everything in detail”Pick one style and be consistent

​
Routing Instructions vs System Prompt

The system prompt applies to the entire chatbot. Routing instructions (set per knowledge base source) let you add source-specific context. For example, if your chatbot has knowledge from both a product manual and a pricing page:
  • System prompt: General tone, identity, rules
  • Routing instructions on product manual: “When answering from this source, include part numbers and compatibility info”
  • Routing instructions on pricing page: “When answering from this source, always mention the current promotion”