Providers

Providers abstract external AI services used throughout the system. They are shared entities (not project-scoped) and can be referenced by multiple projects, stages, classifiers, transformers, and tools.

Structure

Field	Description
id	Unique identifier
name	Display name
description	Optional description
providerType	Service type: llm, tts, asr, embeddings, or storage
apiType	Implementation identifier (e.g., openai, azure, anthropic)
config	Provider-specific connection configuration
createdBy	Operator who created the provider
tags	Searchable labels for organization
version	Optimistic locking version

Provider Types

LLM Providers

Used for: response generation (stages), classification (classifiers), data extraction (transformers), tool execution (tools).

API Type	Description
openai	OpenAI Responses API (GPT-5, GPT-4o, o-series)
openai-legacy	OpenAI Chat Completions API (legacy)
anthropic	Anthropic API (Claude models)
gemini	Google Gemini API (also supports Vertex AI endpoints)
groq	Groq Cloud (fast inference, Llama / Mixtral models)
mistral	Mistral AI API
deepseek	DeepSeek API
openrouter	OpenRouter (unified gateway to many providers)
together-ai	Together AI
fireworks-ai	Fireworks AI
perplexity	Perplexity AI
cohere	Cohere API
xai	xAI (Grok models)

Configuration example (OpenAI):

{
  "apiKey": "sk-...",
  "organizationId": "org-...",
  "baseUrl": "https://api.openai.com/v1"
}

LLM Settings (per-reference, not in the provider config):

Each entity referencing an LLM provider can customize settings. Available fields depend on the provider's API type — see the Swagger UI for the full per-provider schema (e.g., OpenAILlmSettings, AnthropicLlmSettings). Common fields:

{
  "model": "gpt-4o",
  "defaultMaxTokens": 2048,
  "defaultTemperature": 0.7,
  "defaultTopP": 1.0
}

To discover which models are available for a configured provider, use GET /api/providers/:id/models.

TTS Providers

Used for: voice synthesis in agents.

API Type	Description
elevenlabs	ElevenLabs text-to-speech
openai	OpenAI TTS
azure	Azure Cognitive Services Speech
deepgram	Deepgram Aura TTS
cartesia	Cartesia Sonic TTS

See Agents for TTS settings configuration per voice.

ASR Providers

Used for: speech-to-text transcription at the project level.

API Type	Description
azure	Azure Speech Services
elevenlabs	ElevenLabs speech recognition
deepgram	Deepgram Nova ASR
assemblyai	AssemblyAI Universal Streaming
speechmatics	Speechmatics real-time speech-to-text

Configured in the project's asrConfig.

Storage Providers

Used for: persisting conversation artifacts (audio recordings, transcripts).

API Type	Description
s3	Amazon S3 or S3-compatible storage
azure-blob	Azure Blob Storage
gcs	Google Cloud Storage
local	Local filesystem storage

Configured in the project's storageConfig.

Provider Catalog

The system includes a built-in Provider Catalog that lists all available provider types, their API types, and configuration schemas. Access it via:

GET /api/provider-catalog

This is useful for building operator UIs that dynamically render provider configuration forms.

Where Providers Are Referenced

Entity	Provider Types Used
Project	ASR (speech-to-text), Storage (artifacts)
Agent	TTS (voice synthesis)
Stage	LLM (response generation)
Classifier	LLM (intent classification)
Context Transformer	LLM (variable population)
Tool	LLM (tool execution)

Provider Inheritance

If a stage, classifier, transformer, or tool does not specify its own llmProviderId, the system may fall back to project-level defaults. However, it's recommended to explicitly set providers for clarity and predictability.

Security

Provider configurations contain sensitive data (API keys, connection strings). Only operators with provider:read or provider:write permissions can view or modify providers. The config is stored securely and not exposed in API responses beyond what's necessary.