Configuration Guide
AI Generation Config
How to configure the prompts used for AI-powered JD, competency, and question generation.
URL: /admin/ai-generation
What This Page Controls
The system and task instruction prompts for all AI generation features:
JD Generation
- Persona (system):
jd_system_prompt— how the AI identifies itself for JD generation - Task Instructions (user):
jd_user_prompt— what the AI should do. Tokens:{title},{context},{document}
JD Enhancement
- Persona:
jd_enhance_prompt - Task Instructions:
jd_enhance_user_prompt— Tokens:{title},{jd_text},{context},{document}
Competency Generation
- Persona:
competency_system_prompt - Task Instructions:
competency_user_prompt— Tokens:{jd_text},{context},{document},{count}
Question Spine Generation
- Persona:
question_system_prompt - Task Instructions:
question_user_prompt— Tokens:{jd_text},{competencies},{count},{context},{document}
Token Pills
Each prompt field shows available tokens as clickable pills. Click a pill to insert it at your cursor position. The UI warns if you use an invalid token.
Labels
The admin UI uses "Persona" and "Task Instructions" labels (not "system prompt" and "user prompt") to be more intuitive for non-engineers.
Impact
Changes here affect the quality of AI-generated competencies and questions that employers use. If generated competencies have poor anchors, it may be the prompts rather than the AI model.