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

# How It Works

> SLNG runs three stages between your orchestrator and the models, one for each part of the voice pipeline. Each removes a category of unnecessary compute.

### STT Performance layer `PRIVATE BETA`

The STT performance layer cleans up and routes audio in real time to make transcription faster and more accurate. It cancels background noise before speech reaches the transcription engine, detects the spoken language on the fly to send audio to the right model, and separates individual speakers within a conversation, all working together so customers get cleaner, better attributed, more accurate transcripts without adding extra steps to their pipeline.

See [STT Performance layer](/execution-layer/stt-performance-layer) for details.

### Context Router (LLM)

The Context router sits in front of the customer's LLM as a single drop-in endpoint (Open AI compatible) for every call in a voice conversation. It speeds up and cuts the cost of LLM calls by answering predictable turns instantly, caching repeated responses so they skip the frontier models entirely, and routing each call to the right provider and model based on the stage of the conversation, all working together so customers get faster, cheaper, more reliable LLM responses without changing how they integrate.

Intelligence is allocated where it is needed and skipped where it is not.

See [Context Router](/execution-layer/context-router) for details.

### TTS Path Optimization

SLNG's TTS Path Optimization layer serves audio that's already been generated so common or repeated phrases don't need to be synthesized again, cutting both latency and cost for customers running voice agents at scale.

See [TTS Path Optimization](/execution-layer/tts-path-optimization) for details.

## Regional execution

All three stages run across a global edge network. By default, requests route to the lowest-latency region. You can pin to specific clusters for compliance or performance.

## Continuous improvement

As more calls flow through the layer, routing decisions sharpen, cache coverage grows, and tier allocation becomes more accurate. Every call makes the next one more efficient.