AI Providers

DagFlux uses a large language model to translate plain-language requests into the operations that workflows execute. You bring your own key for your preferred provider and DagFlux handles the rest.

Supported providers

DagFlux supports the following AI providers for chat-style generation:

OpenAI

Industry-leading language models with excellent code generation capabilities.

Models: GPT-4, GPT-4 Turbo, GPT-3.5 Turbo

Anthropic Claude

Powerful models known for nuanced understanding and long context windows.

Models: Claude 3 Opus, Claude 3 Sonnet, Claude 3 Haiku

Google Gemini

Google's advanced multimodal AI with strong reasoning capabilities.

Models: Gemini Pro, Gemini Ultra

Embeddings providers

For text-to-vector embeddings (used in vector search and similarity operations), DagFlux supports:

OpenAI — text-embedding-ada-002, text-embedding-3-small, text-embedding-3-large

At a glance

DagFlux supports OpenAI, Anthropic Claude, and Google Gemini for chat-style generation, and OpenAI or Hugging Face for embeddings.
You configure one or more keys; DagFlux automatically picks an available provider at request time, with a default ordering you can override per request.
Decrypted keys never leave the background engine — the interactive layer always asks the engine to make the call on its behalf.
When a saved chat history would push a request past the model's context window, DagFlux drops the oldest entries automatically until the request fits, leaving the new prompt intact.

Section	What it covers
Providers	Which providers and models DagFlux supports, and how it picks one
Skill capabilities	What the AI assistant and generated code can use at run time
Embeddings	How text is turned into vectors

Security & Privacy

All API keys are encrypted using your operating system's secure storage (Keychain on macOS, Credential Manager on Windows, Secret Service on Linux). Keys are only decrypted in memory when making API calls and are never exposed in the UI or workflow files.

How DagFlux chooses a provider

When you configure multiple provider keys, DagFlux follows a default priority order. You can override this on a per-request basis if needed. If a provider fails or is unavailable, DagFlux automatically falls back to the next available provider.

Context window management

DagFlux automatically manages conversation history to stay within each model's context window. When a request would exceed the limit, older messages are dropped while preserving the most recent context and your new prompt. This ensures your workflows can have extended AI conversations without manual history management.