Skip to content

AI Providers

Taufinity Studio supports multiple AI providers out of the box. Each widget can use a different provider and model — configured per-widget in the AI Configuration section.

Built-in Providers

OpenAI

Standard OpenAI API. Supports GPT-4o, GPT-4, and other OpenAI models.

SettingExample
Provideropenai
Modelgpt-4o, gpt-4o-mini, o1
CredentialsOpenAI API key

Anthropic

Direct access to Claude models via the Anthropic API.

SettingExample
Provideranthropic
Modelclaude-sonnet-4-20250514, claude-3-5-haiku-20241022
CredentialsAnthropic API key

Google (Gemini)

Google's Gemini models via the Generative AI API.

SettingExample
Providergoogle
Modelgemini-2.0-flash, gemini-pro
CredentialsGoogle AI API key

Mistral

Mistral AI models including Mistral Large and Codestral.

SettingExample
Providermistral
Modelmistral-large-latest, mistral-small-latest, codestral-latest
CredentialsMistral API key

OpenRouter

Access 200+ models from multiple providers through a single API key. OpenRouter acts as a unified gateway — one key unlocks models from Meta, Cohere, Perplexity, DeepSeek, and more.

SettingExample
Provideropenrouter
Modelopenrouter/meta-llama/llama-3.1-405b-instruct, openrouter/openai/gpt-4o
CredentialsOpenRouter API key

Model Names

OpenRouter models use the format openrouter/<provider>/<model>. Browse available models at openrouter.ai/models.

Amazon Bedrock

AWS-managed access to Claude, Llama, Mistral, and other models through your existing AWS account. Uses AWS IAM credentials with SigV4 signing — no separate API keys needed.

SettingExample
Providerbedrock
Modelbedrock/anthropic.claude-3-5-sonnet-20241022-v2:0, bedrock/meta.llama3-1-70b-instruct-v1:0
CredentialsAWS Access Key ID + Secret Access Key + Region

EU Regions

Bedrock is available in EU regions (eu-west-1, eu-central-1). The region is configured as part of the credentials — each organization can use a different region.

Temporary Credentials

If using AWS STS temporary credentials (with a session token), note that these expire. Update the credentials in Secret Manager before expiry. The system caches credentials for 5 minutes, so expired credentials are automatically refreshed on the next request.

Custom Provider

Use any OpenAI-compatible API endpoint as a custom provider. Configure the endpoint URL, authentication, and model name in Settings > Custom AI Providers.

See the Admin Guide for managing custom providers.

Credential Management

Provider credentials are stored per-organization in Google Secret Manager. Each organization brings its own API keys — there are no platform-level keys shared across tenants.

To configure credentials, go to Settings > AI Providers in the dashboard.

Agent Mode Compatibility

All providers support agent mode (tool calling) when enabled on a widget. The system automatically adapts the tool calling format for each provider:

  • OpenAI, OpenRouter, Mistral — OpenAI function calling format
  • Anthropic — Claude tool use format
  • Google — Gemini function calling format
  • Bedrock — AWS Converse API tool format
  • Custom — OpenAI function calling format (most custom endpoints are OpenAI-compatible)