---
title: Planner
description: The planner decision log at
---

# Planner



## Purpose

PLANNER (route `#/decisions`) audits how the capacity planner graded each compute launch attempt and why, so operators can answer "is the planner doing the right thing?" without replaying router logs.

Deep link: `ops.useargus.co/#/decisions` (nav group COMPUTE; the nav label is PLANNER).

## What it shows

* A filter toolbar:
  * WINDOW chips: 1h, 6h, 24h, 7d, 30d.
  * GRADE chips: ALL, A, B, C, D, F.
  * OUTCOME chips: ALL, OK (success), NO CAP (capacity\_unavailable), ERROR (driver\_error).
* A summary block for the selected window.
* DECISION AUDIT: the decision rows (up to 200), with an inline legend: GRADE is the A to F score the planner assigned each launch attempt, where A is a clean allocation match and F is failed or forced. A row can be expanded inline for its detail, and the expansion survives refresh ticks.

Data sources: `GET /compute/planner/decisions` and `GET /compute/planner/decisions/summary`.

## Actions

Filtering only; this view performs no mutations. Changing a filter drops the cached payload and refetches for the new window, grade, or outcome.

## Matching CLI

```bash
argusctl compute planner
```

See [argusctl compute](/docs/cli/compute).

## Notes

* Polled on the 15 second timer. Summary and rows are fetched together as one atomic payload so the panel is never half stale.
* The toolbar lives outside the polled panel, so filter state and chip handlers survive every repaint.


---

For a semantic overview of all documentation, see [/sitemap.md](/sitemap.md)

For an index of all available documentation, see [/llms.txt](/llms.txt)

For agent-facing discovery, including API and MCP surfaces, see [/agents.md](/agents.md)