Configure Editorial AI | PagibleAI Documentation

Configure tasks separately

PagibleAI does not require one provider to handle every editorial task. Configure the service and model for writing, refinement, translation, descriptions, transcription, and image operations independently. Only add credentials for the features you enable.

Keep secrets in environment variables. Commit the provider and model choices only when they are safe defaults for every deployment.

CMS_AI_WRITE=gemini
CMS_AI_WRITE_MODEL=gemini-2.5-flash
CMS_AI_WRITE_API_KEY=your-key

CMS_AI_TRANSLATE=deepl
CMS_AI_TRANSLATE_API_KEY=your-key

CMS_AI_TRANSCRIBE=openai
CMS_AI_TRANSCRIBE_MODEL=whisper-1
CMS_AI_TRANSCRIBE_API_KEY=your-key

Operational checks

Check
Why it matters
Owner
Provider terms
Content may leave your infrastructure
Legal or security
Model choice
Quality, cost, and context limits vary
Product owner
Key scope
Limits damage from a leaked credential
Platform team
Human review
Generated work remains a draft
Publisher
Fallback behavior
Editors need a clear error when a service is unavailable
Development team

Keep the publishing decision human

Security controls used to protect an editorial workspace

Generated text, translations, and media enter the same draft and version workflow as manual changes. The provider does not publish a page. Your CMS permissions still determine who can approve the result and make it public.

Test the configured task with representative content before enabling it for a wider editorial group.