Create chatbot guide
Build an education-focused chatbot with a practical workflow from sign-in through preview readiness.
Who this guide is for
This guide is for educators, training teams, and support teams creating a chatbot that answers questions from course, school, program, or internal learning materials. It focuses on the creator workflow before a chatbot is published for learners or visitors.
Before you start
- Make sure you can sign in to the StadiTA workspace where the chatbot will live.
- Write the audience, topic boundaries, and top questions the chatbot should handle.
- Gather current source files for knowledge packs, and remove outdated duplicates before upload.
- Identify a teammate or subject-matter reviewer who can test the chatbot in preview mode.
What you can configure in StadiTA
Basics
Set your chatbot name and short description so users immediately understand what it helps with.
Persona instructions
Define tone, guardrails, audience, scope boundaries, and what the chatbot should do when it is unsure.
Welcome message
Use the opening message to set expectations, name the topic area, and suggest a strong first prompt.
Knowledge packs
Upload trusted files into knowledge packs so answers can be grounded in your materials instead of generic assumptions.
Model and reasoning settings
Choose model and reasoning settings based on the balance you want between answer depth, response speed, and AI Credit usage.
Step-by-step setup
1) Sign in
Sign in before creating or previewing a chatbot. Preview work belongs on the authenticated preview path, not on the public live route.
2) Create a chatbot
Start a new education-focused chatbot for one clear audience and learning job, such as course Q&A, onboarding guidance, study support, or internal knowledge search.
3) Add name, description, persona instructions, and welcome message
Use a plain-language name and description. Then write persona instructions that define tone, boundaries, answer format, and fallback behavior. Finish with a welcome message that helps users ask useful first questions.
4) Choose model and reasoning settings
Use stronger model or reasoning settings when depth and precision matter, and lighter settings for routine high-volume questions. Compare options by answer depth, response speed, and AI Credit usage.
5) Upload knowledge-pack files
Upload current source files into knowledge packs. Keep files organized by topic, remove stale versions, and test whether answers reflect the right material.
6) Preview before publishing
Use preview mode for realistic conversations, ambiguous prompts, and out-of-scope requests. Confirm the chatbot asks clarifying questions and avoids inventing unsupported facts before you publish it.
Common mistakes and troubleshooting
Vague persona instructions
Replace broad instructions like “answer everything” with audience, scope, tone, fallback, and formatting guidance.
Stale or duplicated knowledge files
Remove outdated files from knowledge packs so retrieval does not mix old and current answers.
Skipping preview checks
Preview and live chat are separate modes. Test internally on the preview route before sharing a live chatbot link.
Overpromising expertise
Keep the chatbot focused on education support and uploaded materials. Add fallback instructions for topics outside its scope.
Pre-publish quality checklist
- Top expected user questions produce useful and consistent answers.
- Responses reflect your uploaded knowledge packs, not model guesswork.
- Unsafe, irrelevant, or unsupported requests trigger safe fallback behavior.
- The welcome message matches brand voice and nudges users to high-quality prompts.
- A teammate or trusted reviewer validates behavior in preview mode.
Next step
Once this checklist passes, continue with the Publish chatbot guide to move from authenticated preview work to a shareable live chatbot route.