> For the complete documentation index, see [llms.txt](https://help.tillit.cloud/tillit/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://help.tillit.cloud/tillit/tools/aura/cycles.md).

# Cycles

## How a Cycle works

Aura operates in recurring cycles such as daily, per shift, or weekly depending on your configuration. Each cycle follows three stages.

<figure><img src="/files/bjrqSfsn460Wcz6MfDGJ" alt=""><figcaption></figcaption></figure>

<figure><img src="/files/vPn3s9z2bjMcd8lWuJY3" alt="" width="375"><figcaption></figcaption></figure>

### Stage 1 > Scan

Aura reads every table in scope, profiles the data, and identifies which investigations are worth running in the current cycle. It determines what to look at before any queries are written.

### Stage 2 > Hypothesise

For each candidate investigation, Aura forms a specific, testable hypothesis while narrowing the focus to the asset, shift, or SKU where the evidence is strongest.

<figure><img src="/files/Wia4SBbCOvcfOKczvMtS" alt=""><figcaption></figcaption></figure>

### Stage 3 > Investigate

Aura writes and runs SQL against your data. For each finding it produces a recommendation, a full evidence notebook, and a confidence score.

<figure><img src="/files/LCaA4g7wpOhz5BwSTSby" alt=""><figcaption></figcaption></figure>

Cycles run without human input. When a cycle completes, findings appear on the discovery board.

See [Findings](/tillit/tools/aura/findings.md).

## Role-based Perspectives

The same data point reads differently depending on who is looking at it. A machine stoppage, for example, is a reliability issue for maintenance, a throughput issue for operations, a schedule risk for planning, and a quality risk all at the same time.

Aura runs five agent perspectives simultaneously so each role receives findings relevant to their decisions:

* **Maintenance** — asset reliability and failure patterns
* **Operations** — throughput, downtime, and performance loss
* **Quality** — rejects and rework by product variant
* **Planning** — schedule risk and forecast deviation
* **Supply chain** — inventory and SKU-level issues

{% hint style="info" %}
If you would like more than 5 specialised roles, contact us and we cal discuss a solution.
{% endhint %}

## Audit Trail and Reporting

If you want to see exactly what each of your agent has done. Click on 'Download Reports (zip)'

<figure><img src="/files/SePrwxvoy3vOVmyMfzya" alt=""><figcaption></figcaption></figure>

You will get a zip file containing this content:

<figure><img src="/files/snsijGAzSerWlfsozfs8" alt="" width="375"><figcaption></figcaption></figure>

* `queue.json` - the queue history of agents running analyis
* `sessions` folder
  * `*.log` files containing different analysis sessions run by the agents containing information about the model, data used etc
* `reports` folder
  * `*.json` files - the analysis that the agent had done in json format. This is used to form the equivalent pdf reports.
  * `*.pdf` files - the formatted PDF report about the analysis the agent had done.

<figure><img src="/files/UTII2PWr5Tg3R0LMl6O7" alt=""><figcaption></figcaption></figure>