Persona
Teach a Persona with knowledge data and generated prompt responses.
The Persona Knowledge Base stores training data from text, pages, attachments, and sources, then uses it to create prompt-response knowledge.
Model
What it stores
- Training data
- A titled knowledge object with token count and active status.
- Pages
- URLs or page lists used to collect source knowledge.
- Attachments
- Files uploaded as part of a knowledge object.
- Text
- Pasted knowledge such as FAQs, emails, transcripts, or notes.
- Scope
- Private or public data scope for the stored knowledge.
Workflow
Add training data
- 01
Open Knowledge Base
Open it from a trained Persona instance. - 02
Add a training object
Paste text, add URLs, attach files, or combine sources. - 03
Choose language and scope
The UI supports Persona language selection and public/private data scope. - 04
Review stored pages
BHuman can stream page collection progress for reviewed sources. - 05
Generate prompt responses
Use the knowledge object to auto-generate prompt-response pairs.
Sources
Sources
- Use URLs when Persona needs to learn from a public page or domain.
- Use attachments when source material is a file.
- Use pasted text for transcripts, emails, FAQs, policies, or notes.
- Use clear titles so training data is searchable and maintainable.
- Keep pasted training objects under the product word limit.
Training data is not a dumping ground
Add focused, current source material. Large, stale, or contradictory knowledge makes prompt-response review harder.
Prompts
Generated prompts
If no prompts exist and training data is available, Persona can generate prompt-response pairs from the knowledge base. Review them before relying on them in public or embedded experiences.