> ## Documentation Index
> Fetch the complete documentation index at: https://docs.changeguard.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Engineering Experience

> Your workspace's own engineering history, recalled into an opinion only when it changes the decision: past fix outcomes on the same target and strong similar incidents. Generally available with the Engineering Opinion.

**Engineering Experience** is institutional memory: what happened the last time a fix was applied to this target, and what strong similar incidents in your workspace looked like. It is recalled into an [Engineering Opinion](https://docs.changeguard.ai/intelligence/engineering-opinion.md) and shown as **Seen before**, only when the history actually changed the decision.

It is one input to a decision, never a decision. Similarity is stated as similarity, never as identity.

## What it recalls

| Memory | Source | Shown as |
| - | - | - |
| Terminal outcomes of past fixes on the same target | Remediation records in your workspace: verified, rolled back, failed, with the sample size (this incident excluded) | Exact counts from the backend, for example "2 of 3 fixes on this target rolled back", with citations to the real remediation records |
| Strong similar past incidents | Incidents in your workspace whose stored embeddings are at least 0.55 similar | Cited incidents, marked as similar, with the backend's count |

Recall has a restraint gate. It is eligible only when there is a failure or rollback on record, at least two terminal outcomes on the target, or a strong similar incident. Below that, nothing is recalled and nothing is shown.

## What "tenant memory" means for your data

* Memory is **per workspace**. It is built only from your own incidents and remediation outcomes; it never reads another tenant's history.
* It is **retrieval, not retraining**. No model, threshold, policy or prompt is modified by an outcome. The embeddings used for similarity are stored per tenant and read under the same row-level isolation as everything else.
* It only exists where there is history. A workspace with no reported outcomes has nothing to recall, and the product says so by staying quiet.

## Prerequisites and limits

| | |
| - | - |
| Plan | Professional or Enterprise (plan feature `ai`). |
| Setting | **Settings > Engineering assistance > Engineering Experience**; requires the Engineering Opinion to be on. |
| History | Recorded terminal outcomes and stored incident embeddings in your workspace. Report outcomes to build it. |
| Limits | Recall is per workspace and only from recorded outcomes; similarity is embedding proximity, stated as such. |

## Related

* [Engineering intelligence overview](https://docs.changeguard.ai/intelligence/overview.md)
* [Engineering Reasoning](https://docs.changeguard.ai/intelligence/engineering-reasoning.md)
* [Security: data handling](https://docs.changeguard.ai/security/data-handling.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.