> For the complete documentation index, see [llms.txt](https://seekspeed.gitbook.io/seekspeed-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://seekspeed.gitbook.io/seekspeed-docs/regression-guard/overview.md).

# Overview

Regression Guard turns SeekSpeed into a continuous QA layer for agent endpoints. You pin a run as the **baseline**, put the same benchmark on a **schedule**, and Guard runs it in the background and raises an **alert** when latency, cost, throughput, or error rate regresses beyond a configurable threshold.

It exists because "we shipped a prompt change, is the p95 still fine?" should be a check that runs on its own, not a Slack ping to whoever benchmarked last.

## Concepts

| Concept       | What it is                                                                                                                                                                                           |
| ------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Baseline**  | Frozen snapshot of a completed run's summary: p50 / p95 / mean latency, mean tok/s, cost per run, error rate, and the per-iteration total-latency array (kept so we can run Welch's *t*-test later). |
| **Schedule**  | A benchmark set + cadence (`15m` / `hourly` / `daily` / `weekly`) + baseline + threshold list + notify config. Executed by a public cron endpoint.                                                   |
| **Threshold** | A `(metric, op, value, severity)` tuple. e.g. `p95_ms +15% → warn`, `error_rate > 1% → critical`.                                                                                                    |
| **Alert**     | Row emitted when a scheduled run breaches ≥1 threshold. Rendered in the Guard feed with unacknowledged-first ordering; also POSTed to the configured webhook.                                        |

## Why bother

A one-off benchmark answers "how fast is my agent right now". Regression Guard answers three questions that are much harder to keep track of:

* "Did today's model rev slow us down?" — cadence + p95 threshold.
* "Are we drifting toward the SLA?" — trend across many scheduled runs.
* "Did prompt change #47 quietly increase spend?" — cost\_usd threshold on the same prompt set.

## What Guard *doesn't* do

* It does **not** replace load testing. Guard runs your benchmark set once per tick; it is not a sustained-load generator.
* It does **not** grade output quality. If you need pass-rate on top of latency, wire a golden-dataset harness on top of the same schedules.
* It does **not** back-fill history. Only runs executed after Guard was enabled show up in the timeline.


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://seekspeed.gitbook.io/seekspeed-docs/regression-guard/overview.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.