INPUTS
Triggers, merchant context, and time history enter the decision layer.
Vera
Context-aware merchant growth assistant that ranks triggers, grounds messages in sourced fact packets, rotates model keys, and validates every outbound WhatsApp draft.
VERA
MERCHANT GROWTH ENGINE
Context-aware merchant growth assistant that ranks triggers, grounds messages in sourced fact packets, rotates model keys, and validates every outbound WhatsApp draft.
SYSTEM / MEDIA LOOP
THE PIPELINE
Vera separates deterministic decisioning from generative drafting, then validates every outbound message before it ships.
SYSTEM NOTES
HOW IT WORKS
The pipeline keeps research, creation, publishing, and learning inspectable as one operating loop.
DECISION ENGINE
Seven hard gates and priority ranking determine whether outreach should exist.
FACT PACKET
Sourced context is grounded and provenance tagged before drafting.
MESSAGE COMPOSER
Gemini Flash drafts a WhatsApp message after deterministic decisions.
OUTPUT VALIDATOR
Grounding, safety, tone, and a single CTA are checked before shipping.
KEY POOL
Model keys rotate with cooldown and failover as part of the service.
TECHNICAL CONTEXT
THE SYSTEM, BRIEFLY
Context-aware merchant growth assistant that ranks triggers, grounds messages in sourced fact packets, rotates model keys, and validates every outbound WhatsApp draft.
Vera separates deterministic decision intelligence from generative drafting. Hard gates, priority ranking, suppression, and time tracking decide whether a message should exist; a grounded composer writes it; a validator checks tone, claims, and CTA safety before it ships.
The deterministic system decides. The LLM writes.
PythonFastAPIGemini FlashOpenRouterPydanticJSONLRedis