AI Providers
WhatsMine's AI chatbots and AI features are powered by large language models. You can set a platform-wide key (so AI works for everyone out of the box) and/or let each customer use their own key.
Supported providers
| Provider | Used for | Credentials |
|---|---|---|
| OpenAI (default) | Chat + embeddings (semantic search) | API Key (and optional Organization ID) |
| Anthropic (Claude) | Chat | API Key |
| Google Gemini | Chat | API Key |
Embeddings
Knowledge-base semantic search needs embeddings, which OpenAI provides. If you use Claude or Gemini for chat, OpenAI (or the built-in fallback) is still recommended for embeddings.
System-level setup (admin)
Location: Admin → Integrations
📸 Screenshot: The AI provider integration form.
- Open the provider (e.g. OpenAI).
- Enter the API key.
- Test the connection.
- Enable it (and optionally set as default).
This makes AI available to all your customers using your account/billing — a great way to offer AI as a built-in feature and meter it via plan limits.
Getting keys
- OpenAI — platform.openai.com → API keys.
- Anthropic — console.anthropic.com → API keys.
- Google Gemini — Google AI Studio → API key.
Workspace-level setup (customer)
Location (customer): AI → Providers (/app/ai/providers)
Customers can enter their own API key to use their own AI account instead of your platform default. When a workspace has its own key, it takes priority; otherwise it falls back to your system key.
This is ideal for customers who want to control their own AI spend or use a specific model.
Vector database (optional)
For semantic search over knowledge bases, WhatsMine can use Qdrant if you configure it:
QDRANT_URL=https://your-qdrant-host:6333
QDRANT_API_KEY=your-key
If Qdrant isn't configured, WhatsMine uses an efficient in-database cosine-similarity fallback — so AI knowledge bases work with no extra services. The Admin AI Dashboard shows Qdrant health when configured.
Choosing a model
The default chat model is cost-effective (gpt-4o-mini). You can change models per chatbot. For the latest model recommendations and pricing, consult your provider's documentation.
Monitoring cost
AI usage (tokens, runs, cost, latency) is tracked and visible in:
- Admin — platform-wide in the AI Dashboard.
- Customer — per workspace in Reports.
Usage maps to the ai_tokens_per_month plan limit, so you can price AI fairly.
➡️ Next: Payments.