🇬🇧 EN

3.1 AI Providers

AI Providers are the connections to large language model services that power the Agent, Schema Extractor, Image Generation nodes, and AI Voice Call questionnaires. You must configure at least one provider before using any AI-powered node in your workflows.

AI Providers settings page — Settings → AI Providers — empty state with the "Add First Provider" prompt. Once providers are configured, they appear as a list here.

Supported Providers

Provider	Driver name	Notes
OpenAI	`openai`	GPT-4o, GPT-4 Turbo, o1, and all OpenAI API models.
Anthropic	`anthropic`	Claude 3, Claude 3.5, Claude 4 model families.
Ollama	`ollama`	Self-hosted local models. Point the Base URL to your Ollama instance.
Any OpenAI-compatible API	`openai`	Works with Groq, Together AI, LM Studio, etc. Set the Base URL to the endpoint.

Adding a Provider

Open AI Providers settings Navigate to Settings → AI Providers in the sidebar.
Click "New Provider" A modal form slides open on the right side of the screen.
Fill in the provider details Complete the fields as described in the table below.
Save Click Save Provider. The new provider will appear in the list immediately.

Add AI Provider modal form — Add AI Provider form — Display Name, Driver, Default Model, Base URL, API Key, Capabilities checkboxes, and a "Set as default provider" option.

Provider Fields Reference

Field	Required	Description
Provider Name	Required	A human-readable label for this provider (e.g. "OpenAI Production", "Local Ollama"). This name appears in node configuration dropdowns.
Driver	Required	The API format this provider uses. Choose `openai` for OpenAI and compatible APIs, `anthropic` for Anthropic, or `ollama` for Ollama.
Base URL	Optional	Override the default API endpoint. Required for self-hosted models (Ollama) or third-party compatible APIs (e.g. `http://localhost:11434` for Ollama). Leave blank for standard OpenAI or Anthropic endpoints.
API Key	Conditional	Your secret API key. Required for cloud providers (OpenAI, Anthropic). Not required for local Ollama instances running without authentication. The key is stored encrypted.
Default Model	Required	The model identifier that will be pre-selected when you add a new AI node and choose this provider. Examples: `gpt-4o`, `claude-sonnet-4-6`, `llama3.2`.
Set as Default	Optional	When checked, this provider is automatically selected in new Agent nodes. Only one provider can be the default at a time; toggling another provider as default will unset the previous one.

Capabilities

Each provider can declare which capabilities it supports. Capabilities appear as checkboxes in the provider form and as badges on provider cards.

Capability	Description
Tool Calling	The model can invoke external tools/functions. Required for Agent nodes that use tools.
JSON Mode	The model can produce structured JSON output. Used by Schema Extractor and some Agent nodes.
Vision	The model can process images. Required for image-related workflows.
Voice	The model is suitable for voice call questionnaires. When checked, additional voice-specific configuration fields appear (see below).

Voice Capability Settings

When the Voice capability is enabled, an additional configuration section appears in the provider form. These settings control how the provider behaves when used for AI Voice Call questionnaires.

Field	Description
Voice Model	The model to use for voice answer classification. Defaults to the provider's default model if left blank. Consider using a fast model (e.g. `gpt-4o-mini`) for lower latency.
Supports Realtime	Whether this provider supports native audio I/O (e.g. OpenAI Realtime API). For future use — currently all voice processing uses Twilio for speech.
Transcription Strategy	`twilio` (default) — Twilio handles speech-to-text. `provider` — reserved for future native transcription.
Synthesis Strategy	`twilio` (default) — Twilio handles text-to-speech. `provider` — reserved for future native synthesis.
Parsing Mode	How the AI classifies spoken answers: `classifier_prompt` (default, uses a classification prompt), `direct_structured`, or `hybrid`.
Default Turn Timeout	Default seconds to wait for caller response before re-prompting. Can be overridden per questionnaire in the Voice Call channel settings.

AI Providers with voice capability — AI Providers settings page — showing provider cards with capability badges including Voice.

Editing & Deleting Providers

To edit a provider, click the Edit (pencil) icon next to it in the list. The same form opens with existing values pre-filled. Change any field and click Save Provider.

To delete a provider, click the Delete (trash) icon. A confirmation prompt will appear. Note that deleting a provider will not remove it from existing workflow node configurations — those nodes will simply fail to execute until you update them with a valid provider.

Before deleting a provider: Search your workflows for any Agent, Schema Extractor, Image Generation, or Voice Call nodes configured to use that provider, and update them first.

Common Setup Examples

OpenAI

Driver: openai
Base URL: (leave blank)
API Key: Your key from platform.openai.com → API Keys
Default Model: gpt-4o

Anthropic Claude

Driver: anthropic
Base URL: (leave blank)
API Key: Your key from console.anthropic.com → API Keys
Default Model: claude-sonnet-4-6

Local Ollama

Driver: ollama
Base URL: http://localhost:11434 (adjust for your host)
API Key: (leave blank)
Default Model: llama3.2 (or any model you have pulled)

Tip: You can register multiple providers and switch between them per-node. For example, use GPT-4o for critical agents and a local Ollama model for lightweight classification tasks to reduce cost.