AI Providers
Activate AI providers per environment and reuse their credentials across Universal AI components, the Knowledge Base, and other AI features.
AI Providers is the catalog where an administrator activates the AI providers (OpenAI, Anthropic, Mistral, and others) used by ByteChef's AI features. You enter a provider's API key once, and every AI feature that needs a model reads the credentials from this catalog instead of from a per-workflow connection.
The page requires the ADMIN authority. Keys are stored per environment, so Development, Staging, and Production can each use their own providers and credentials. Self-hosted deployments can also supply keys through environment variables such as BYTECHEF_AI_PROVIDER_OPEN_AI_API_KEY. See Self-hosting: static keys.

Activate a provider
-
Go to Settings → AI → Providers.
-
Select the environment (Development, Staging, or Production) in the environment selector in the left sidebar.
-
Turn on the provider's toggle. If the provider has no credentials yet, its row expands to the credentials form.
-
Enter the credentials and click Save:
- Most providers: API Key.
- Azure OpenAI: API Key and Endpoint (for example
https://<resource>.openai.azure.com). - Ollama: Base URL (default
http://localhost:11434) and an optional API Key for Ollama's hosted API. Clearing the Ollama key removes it.
A saved API key is shown masked. Leave it unchanged to keep it.
After you save, the toggle stays on. To review or change a credential, expand the row and click Edit. To stop using a provider, turn off its toggle.

The API key is encrypted at rest and is only shown in obfuscated form. Repeat the steps for each environment: activating OpenAI in Development does not activate it in Production.
Available providers
The catalog lists Anthropic, Azure Open AI, Groq, Mistral, NVIDIA, Open AI, Stability, Vertex Gemini, Perplexity, DeepSeek, and Ollama. Each row has badges for the capabilities the provider supports. Hover a badge to see its description.
| Badge | Meaning | Providers |
|---|---|---|
| Text | Chat models for text generation, used by AI components and agents. | All except Stability |
| Image | Image generation from text prompts. | Azure Open AI, Open AI, Stability, Vertex Gemini |
| Embeddings | Embedding models for knowledge bases and semantic search. | Mistral, Open AI, Ollama |
| Copilot Docs | The provider whose embeddings Copilot uses to search the ByteChef documentation. | The provider currently configured for it |
What uses activated providers
A provider activated for an environment is available to:
- Universal AI components - workflow steps where you pick a provider from the catalog and then a model, without a separate connection.
- The Copilot model picker (coming soon) - choose which activated provider and model answers a Copilot conversation.
- The AI Hub (coming soon) - the same catalog backs the AI Hub's model selection.
- The Knowledge Base - document embedding uses the first activated embedding provider in the environment, checked in the order Mistral, OpenAI, Ollama.
Each feature reads the key for the current environment, so a workflow uses that environment's credentials as it moves from Development to Production.
Default models
Each provider has a default model per kind (chat, embedding, and, for OpenAI, image generation), defined in configuration as bytechef.ai.provider.{chat,embedding,image}.<provider>.options.model and overridable through the matching environment variables. See the chat, embedding, and image model references. The AI Providers page stores credentials and enablement, not model names.
The page decides which default applies. Features where ByteChef picks the model (Knowledge Base embedding and, once available, the Copilot default, the workflow converter, the property Copilot, and AI Hub image generation) use the first enabled provider in the current environment, or the provider set in BYTECHEF_AI_COPILOT_PROVIDER when it is enabled, together with that provider's configured model. Enabling or disabling a provider takes effect at the next request, without a restart. Universal AI workflow steps are not affected, because their authors pick the model per step.
Two providers work differently:
- Azure OpenAI has no default model, and the default resolution skips it until you set one. The model is your deployment name: set
BYTECHEF_AI_PROVIDER_CHAT_AZURE_OPEN_AI_OPTIONS_MODELto it. - Ollama uses the models installed on your Ollama server. The AI model picker lists its chat models, read from the server. Unless you set
BYTECHEF_AI_PROVIDER_CHAT_OLLAMA_OPTIONS_MODEL(for exampleqwen3:8b), the default chat model is the first local chat model the server reports. Ollama lists the most recently pulled model first. Cloud models (names ending in:cloud) are listed after local ones and never become the default. Knowledge Base embedding works the same way with the installed embedding models, or withBYTECHEF_AI_PROVIDER_EMBEDDING_OLLAMA_OPTIONS_MODEL(for examplenomic-embed-text).
Choosing an embedding model
Documents already stored in knowledge bases keep the embeddings produced by the model that was active at upload time. Switching to a different embedding model degrades semantic search over existing documents until they are uploaded again.
Embedding models
Embeddings are the vectors behind Knowledge Base search. Three providers in the catalog offer them:
| Provider | Default model | Configuration key |
|---|---|---|
| OpenAI | text-embedding-3-small | BYTECHEF_AI_PROVIDER_EMBEDDING_OPEN_AI_OPTIONS_MODEL |
| Mistral | mistral-embed | BYTECHEF_AI_PROVIDER_EMBEDDING_MISTRAL_OPTIONS_MODEL |
| Ollama | (first embedding model installed on the server) | BYTECHEF_AI_PROVIDER_EMBEDDING_OLLAMA_OPTIONS_MODEL |
The corresponding property form is bytechef.ai.provider.embedding.<provider>.options.model. For the full list of embedding keys, see the AI Embedding Model configuration reference.
Anthropic publishes no embedding models. A deployment that uses Anthropic for chat still needs OpenAI, Mistral, or Ollama activated for embeddings.
Models are configuration
The model is always set in configuration. The AI Providers page stores credentials and enablement, not model names. The exception is Ollama with no model configured: ByteChef then picks one of the models installed on the server. The API key can also be set in configuration (for example BYTECHEF_AI_PROVIDER_OPEN_AI_API_KEY), so a deployment can set both the embedding model and its key through properties.
Vector dimensions are fixed per deployment
Embedding models produce vectors of different widths (text-embedding-3-small produces 1536 dimensions, text-embedding-3-large produces 3072), and a pgvector column has one width. The vector store is configured for 1536 dimensions by default (bytechef.ai.vectorstore.pgvector.dimensions, or BYTECHEF_AI_VECTORSTORE_PGVECTOR_DIMENSIONS). When you switch to a model of a different width, change that property as well. A mismatch fails at write time.
Changing the embedding model
Chunks are embedded once, when their document is ingested. There is no re-embedding operation. Changing the model affects documents added afterwards, and existing documents keep their vectors. Retrieval quality degrades until the older documents are deleted and uploaded again. Plan a model change as a re-ingestion, per environment.
Self-hosting: static keys
A self-hosted deployment can supply provider keys as configuration instead of, or in addition to, the UI. For example, set BYTECHEF_AI_PROVIDER_OPEN_AI_API_KEY or BYTECHEF_AI_PROVIDER_ANTHROPIC_API_KEY. See the environment variables reference.
Precedence: a provider configured through the UI for an environment always wins. The configuration key is used only as a fallback when the provider has no settings saved for that environment.
The Copilot documentation corpus is an exception: it is shared by the whole instance and embeds with its own key rather than the per-environment catalog. See AI Copilot → Documentation embeddings.
How is this guide?
Last updated on