# Inference

> Invoke and stream model requests through explicit descriptors, codecs, retries, and a local gateway.

- Path: `Modules > Inference - Model Inference`
- Human: https://looprig.com/docs/modules/inference
- Machine index: https://looprig.com/llms.txt

## Repository

| Field | Value |
| --- | --- |
| Repository | `github.com/looprig/inference` |
| Version | `v0.12.0` |
| GitHub | [looprig/inference](https://github.com/looprig/inference) |

## Description

Invoke and stream model requests through explicit descriptors, codecs, retries, and a local gateway.

## Where it fits

Inference is useful on its own when a Go application needs provider-neutral model requests, streaming, codecs, retries, routing, or a local model gateway. Within Looprig, it supplies the model boundary for [Harness](/docs/modules/harness.md), [Eval](/docs/modules/eval.md), [Tools](/docs/modules/tools.md), and [Workflows](/docs/modules/workflows.md). Inference moves model data; it does not own agent sessions or provider credential catalogs.

## Dependencies

- [Core](/docs/modules/core.md)
- [Credentials](/docs/modules/credentials.md)
- [Secrets](/docs/modules/secrets.md)

## Dependents

- [Classifiers](/docs/modules/classifiers.md)
- [Eval](/docs/modules/eval.md)
- [Foreign Loops](/docs/modules/foreignloops.md)
- [Harness](/docs/modules/harness.md)
- [LLM](/docs/modules/llm.md)
- [MCP](/docs/modules/mcp.md)
- [Tools](/docs/modules/tools.md)
- [TUI](/docs/modules/tui.md)
- [Workflows](/docs/modules/workflows.md)