Choosing and configuring your AI provider

Choosing and configuring your AI provider

Comparison of OpenAI, Google Gemini, DeepSeek, and self-hosted Ollama — and how to wire each one up.

Picking a provider

EnvaTicket's chatbot is provider-agnostic. Pick the option that fits your budget and infrastructure:

ProviderStrengthsTrade-offsTypical monthly cost
OpenAIBest instruction-following; both chat and embeddings on one key.Paid from message one.Under $2 for most authors.
Google GeminiGenerous free tier via AI Studio.Requires the Custom-endpoint config.$0 within the free tier.
DeepSeekCheapest per token for chat.No embeddings API — pair with OpenAI or Ollama for retrieval.Pennies per month.
OllamaSelf-hosted, zero recurring cost.Needs a server with enough RAM (8 GB+) for a 7B–8B model.$0 (host cost only).

Setting up OpenAI

  1. Go to Admin → Settings → AI.
  2. Set AI Provider to OpenAI (hosted).
  3. Paste your API key into API Key.
  4. Set AI Model Name to gpt-4o-mini.
  5. Leave Embeddings Provider as Same as AI Provider.
  6. Save, then click Reindex Now in the Knowledge Index card.

Setting up Google Gemini

  1. Set AI Provider to Custom OpenAI-compatible endpoint.
  2. Custom Endpoint Base URL: https://generativelanguage.googleapis.com/v1beta/openai
  3. API Key: your AI Studio key.
  4. AI Model Name: gemini-2.0-flash.
  5. Save and Reindex Now.

Setting up DeepSeek with OpenAI embeddings

DeepSeek has no embeddings API of its own, so split the configuration:

  1. AI Provider: DeepSeek, model deepseek-chat, with your DeepSeek key.
  2. Embeddings Provider: OpenAI.
  3. Embeddings API Key: your OpenAI key.
  4. Save and Reindex Now.

Setting up Ollama (self-hosted)

  1. On your server: ollama pull llama3.1:8b and ollama pull nomic-embed-text.
  2. AI Provider: Ollama, model llama3.1:8b.
  3. Leave the API Key blank.
  4. Save and Reindex Now.
3 views Thanks for your feedback!
Was this helpful?
Next Article Understanding the Knowledge Index
→