Model Providers

Configure providers in the Models & Providers section of Settings → Server → AI Engine, or with the settings below. They apply when you run your own engine. With Stirling Cloud AI, Stirling Cloud manages models and provider keys.

Language models handle questions and editing instructions. Embedding models make document text searchable. You can use a different provider for each.

Language models#

Provider	Base URL	API key
`anthropic` (Claude)	Default service	Set in AI settings or `ANTHROPIC_API_KEY` on the engine.
`openai` (GPT)	Default service	Set in AI settings or `OPENAI_API_KEY` on the engine.
`ollama`	Required, for example `http://ollama:11434/v1`	Not required.
`custom`	Your OpenAI-compatible endpoint	Set if required by the endpoint.

The default is Anthropic with claude-haiku-4-5 for both Smart model and Fast model. Both models use the same provider. Enter model names without a provider prefix.

Embeddings#

Provider	Model and credentials
`voyageai`	Default model: `voyage-4`. Set `aiEngine.rag.embeddingApiKey` or `VOYAGE_API_KEY` on the engine.
`openai`	Set `OPENAI_API_KEY` on the engine.
`ollama`	Set the endpoint and embedding model, such as `nomic-embed-text`.
`custom`	Set the endpoint, embedding model, and any required key in AI settings.

For example, to set both provider keys in Stirling PDF:

yaml

aiEngine:
  models:
    provider: anthropic
    apiKey: your-anthropic-api-key
  rag:
    embeddingProvider: voyageai
    embeddingApiKey: your-voyageai-api-key

bash

AIENGINE_MODELS_PROVIDER=anthropic
AIENGINE_MODELS_APIKEY=your-anthropic-api-key
AIENGINE_RAG_EMBEDDINGPROVIDER=voyageai
AIENGINE_RAG_EMBEDDINGAPIKEY=your-voyageai-api-key

yaml

services:
  stirling-pdf:
    environment:
      AIENGINE_MODELS_PROVIDER: anthropic
      AIENGINE_MODELS_APIKEY: your-anthropic-api-key
      AIENGINE_RAG_EMBEDDINGPROVIDER: voyageai
      AIENGINE_RAG_EMBEDDINGAPIKEY: your-voyageai-api-key

You can instead set ANTHROPIC_API_KEY and VOYAGE_API_KEY on the engine and leave these keys empty.

Re-add stored documents after changing the embedding model. See Documents and Retrieval.

Local models#

Set the local provider and model names in AI settings, with aiEngine.pushConfigToEngine enabled. The language model must support structured output (JSON schema).

For example, with language and embedding models served at separate OpenAI-compatible endpoints:

yaml

aiEngine:
  models:
    provider: custom
    smartModel: Qwen/Qwen3-8B
    fastModel: Qwen/Qwen3-8B
    baseUrl: http://qwen3:8000/v1
    apiKey: ''
  rag:
    embeddingProvider: custom
    embeddingModel: Qwen/Qwen3-Embedding-0.6B
    embeddingBaseUrl: http://qwen3-embed:8000/v1
    embeddingApiKey: ''

bash

AIENGINE_MODELS_PROVIDER=custom
AIENGINE_MODELS_SMARTMODEL=Qwen/Qwen3-8B
AIENGINE_MODELS_FASTMODEL=Qwen/Qwen3-8B
AIENGINE_MODELS_BASEURL=http://qwen3:8000/v1
AIENGINE_RAG_EMBEDDINGPROVIDER=custom
AIENGINE_RAG_EMBEDDINGMODEL=Qwen/Qwen3-Embedding-0.6B
AIENGINE_RAG_EMBEDDINGBASEURL=http://qwen3-embed:8000/v1

yaml

services:
  stirling-pdf:
    environment:
      AIENGINE_MODELS_PROVIDER: custom
      AIENGINE_MODELS_SMARTMODEL: Qwen/Qwen3-8B
      AIENGINE_MODELS_FASTMODEL: Qwen/Qwen3-8B
      AIENGINE_MODELS_BASEURL: http://qwen3:8000/v1
      AIENGINE_RAG_EMBEDDINGPROVIDER: custom
      AIENGINE_RAG_EMBEDDINGMODEL: Qwen/Qwen3-Embedding-0.6B
      AIENGINE_RAG_EMBEDDINGBASEURL: http://qwen3-embed:8000/v1

Use addresses reachable from the engine container. Leave key fields empty only when the endpoint needs no authentication. Configure both providers locally to keep model requests within your infrastructure.

See AI Settings Reference for environment variables and Self-Hosting the AI Engine for container setup.

Edit this page on GitHub