> ## Documentation Index
> Fetch the complete documentation index at: https://voicera.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Model server

> Run STT, TTS, and LLM on your own GPUs behind one gateway.

An optional, self-contained subsystem for running speech and language models on your own hardware. One gateway on `:8100` fronts three slots — STT, TTS, LLM — and each slot runs whichever model you selected, without the gateway knowing which one.

<Note>
  The model server is **optional**. VoicEra runs fine against cloud providers alone. Deploy this when data residency, per-minute cost, or Indic-language coverage makes self-hosting worth the GPUs.
</Note>

## Start here

| Page                                 | What it answers                                           |
| ------------------------------------ | --------------------------------------------------------- |
| [Overview](overview)                 | What it is, what it replaces, and its current state.      |
| [Slots and models](slots-and-models) | The slot/model split, and switching models with one line. |
| [Gateway API](gateway-api)           | Every endpoint, including `WS /v1/asr/ws`.                |

## The models

| Page                     | Covers                                                                           |
| ------------------------ | -------------------------------------------------------------------------------- |
| [STT models](stt-models) | `indic-conformer`, `indic-transcribe`, and partial transcripts versus streaming. |
| [TTS models](tts-models) | `indic-parler`, `orpheus`, `indic-mio`, and audio format negotiation.            |
| [LLM models](llm-models) | `qwen3.5-4b` and the model-id agreement across files.                            |

## Operating it

| Page                              | Covers                                                 |
| --------------------------------- | ------------------------------------------------------ |
| [Adding a model](adding-a-model)  | The container contract, and the two steps to add one.  |
| [Running on GPUs](gpu-operations) | Device selection, MPS, shared HuggingFace cache, disk. |

## Honest status

<Note>
  Two gaps are documented rather than hidden. The LLM slot has not yet been run on real hardware, and a full call routed from the voice runtime through a self-hosted model has not been verified end to end. Both are called out on the pages concerned.
</Note>

## Related

* [Self-hosted models](../../guides/deployment/self-hosted-models) — wiring the model server to the runtime
* [Provider registry](../reference/provider-registry) — how agents select a provider