Providers
A provider is the connection between Delphi and an AI vendor (OpenAI, Azure, Pythia, Vodafone TOBi, …). Every voice agent ultimately calls a provider to do speech-to-text, generate a response, and synthesise speech.
There is no longer a top-level /providers page in the sidebar. Providers are configured inside the Flow Designer — on the flow's default provider (Flow Settings) or as a per-node override on a voice agent. This page is a concept reference; for editing, open the Flow Designer.
Two ways a flow can use providers
Each flow picks one of two pipeline modes for its default provider:
| Mode | What it does |
|---|---|
| Realtime | One provider runs STT, LLM and TTS in a single bidirectional connection (lowest latency; e.g. OpenAI Realtime, Azure Realtime). |
| Pipeline | Separate STT + LLM + TTS providers chained together (more flexibility; e.g. Azure STT → OpenAI Chat → OpenAI TTS). |
A voice-agent node can override the flow default with its own provider configuration — useful when one part of the conversation needs a different vendor (e.g. a different language).
The catalogue
The set of providers available to a team depends on the platform's feature flags and what the team admin has unlocked.
OpenAI
- Realtime —
gpt-realtime-*family. Picks one of the OpenAI voices, a transcription model, a VAD configuration, and audio noise-reduction. - Chat (LLM) —
gpt-*chat models for pipeline mode. - Transcription (STT) — Whisper-family transcription models.
- TTS —
tts-*voices.
Azure
- Realtime — Azure OpenAI Realtime deployments. Pick the deployment name and domain suffix.
- Chat (LLM) — Azure OpenAI deployments.
- STT — Azure Speech recognition with selectable region and recognition mode (interactive, dictation, conversation), plus language-detection mode.
- TTS — Azure neural voices.
Pythia
Open-source / self-hosted STT and TTS endpoints maintained by KI Kombinat for cases where data residency or cost rules out hosted vendors.
Vodafone TOBi (managed LLM)
A managed LLM integration: the LLM is provided by Vodafone TOBi, with managed transfer / hangup actions and channel-data forwarding. Rendered on the canvas as the dedicated Managed LLM node.
Audio preprocessing
For voice flows the provider config also picks the noise-reduction / VAD profile applied to the inbound audio.
Credentials
Provider credentials (API keys, deployment names) are not stored on the flow. They live as team variables and secrets and are referenced by name inside the provider config. This keeps flow definitions portable across environments — pasting the same JSON into a different platform picks up the destination's credentials.
Where to edit
Open the Flow Designer for an app and:
- Click the Default Provider chip (or the Start node) to edit the flow-default provider config — STT, LLM, TTS, or Realtime.
- Click any voice agent node and use its Provider override tab to depart from the default for that step.
See Inspectors → Provider config editor for the full editor UI.