AI Providers

AI powers chatbots, inbox auto-replies, knowledge-base answers (RAG) and the AI automation generator. WhatsMax supports three LLM providers:

Where keys are configured

There are two levels:

Platform level (you) — Admin → Configuration → AI Dashboard. Platform-wide provider keys, default models and usage monitoring across all workspaces. You can also set a fallback key in .env (AI_PROVIDER, OPENAI_API_KEY, OPENAI_MODEL — see Configuration).
Workspace level (your customers) — AI → Providers inside the app. Workspaces can bring their own keys, so their usage bills to their own provider account.

AI providers Workspace-level provider configuration.

All keys are encrypted at rest with APP_ENCRYPTION_KEY.

Create an API key on the provider console.
Enter it in Admin → AI Dashboard (platform-wide) or AI → Providers (single workspace).
Pick the default model — smaller models (e.g. gpt-4o-mini) are dramatically cheaper and fine for most support chatbots.
Save, then use the built-in test action to confirm the key works.

Feature	Runs where	Notes
AI chatbots	Worker process	Conversation replies via the chatbot builder.
Knowledge bases (RAG)	Worker process	Documents are chunked and embedded on upload; bots retrieve relevant chunks per question.
Inbox auto-reply	Worker process	Optional AI suggestions/automatic replies in conversations.
Automation generator	Web process	"Describe the automation" → builder graph.

Because AI jobs run in the worker, a stopped worker looks like "the bot never answers" — check pm2 logs whatsmax-worker first.

Admin → AI Dashboard shows token usage per workspace; Reports → AI Usage shows it to each workspace.
Plan limits let you cap AI usage per subscription tier (configure limits on the plan in Admin → Plans).
Prefer small/cheap models as defaults; let power users pick larger ones per chatbot.