---
title: "Quiesce"
method: POST
path: "/api/v1/cascade/quiesce"
tags: ["cascade"]
---

# Quiesce

`POST /api/v1/cascade/quiesce`

## Request

- The document declares no server URL.
- Auth: none declared

Drain the projection queue, then stop watcher + scanner + worker.

Returns when the index is a complete projection of the markdown on disk and
nothing further will rewrite it. That ordering is the point: stopping first
would freeze a *partial* index, which reads as a healthy store that quietly
fails to recall whatever had not been indexed yet.

Idempotent -- a second call reports ``quiesced=false`` rather than failing.

## Response `200`

Successful Response

- QuiesceResponse: What the final drain accomplished, and what it left behind. ``pending_before`` is the queue depth on arrival: a caller that expected the projection to be up to date can assert it was 0 and learn otherwise. Anything above 0 in ``pending_after`` means the drain could not finish -- the index is NOT a complete projection of the markdown, and searching it will silently under-recall.
  - `quiesced` boolean, required
  - `drained` integer, required
  - `pending_before` integer, required
  - `pending_after` integer, required
  - `failed_permanent` integer, required

## Changes

- **2026-09-08** `bf705e3f32e2`: 1 info
  - endpoint added

[Operation history](https://skmtc.dev/evermind-ai/apis/everos/changes/api/v1/cascade/quiesce/post.md)

---

[API](https://skmtc.dev/evermind-ai/apis/everos.md) - [All operations](https://skmtc.dev/evermind-ai/apis/everos/llms.txt) - [OpenAPI document](https://skmtc.dev/evermind-ai/apis/everos/revisions/8f6c6973c7e6?raw)