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

# Memory

> See what Vidoc learned from your team, and supersede, reactivate, or edit the entries.

<Info>
  Memory needs the Enterprise plan on Vidoc Cloud. Included in self-hosted installations.
</Info>

**Memory** shows what Vidoc learned from your team and the entries that apply to the current project. It has two types of entries:

* A **learning** comes from feedback on one finding.
* A **memory entry** is a general fact about your code or your policy.

The validation agent uses Memory when it checks later findings. The "find issues" stage does not use Memory. An entry does not remove a finding directly. The validation agent weighs the entry with the code evidence and makes the decision. For scope and details, refer to [How Vidoc works](/how-it-works#learnings-and-memory).

## Where entries come from

The Memory page does not create entries. Entries come from these sources:

| Source                      | Type         | How it is created                                                                                                                                               |
| --------------------------- | ------------ | --------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| **Manual**                  | Learning     | A user sets a finding to Ignored or False positive with a reason, in the web app or through the REST API. Refer to [Findings](/web-app/findings#change-status). |
| **False positive**          | Learning     | A developer replies to a Vidoc PR comment and says that the finding is not real.                                                                                |
| **Confirmed true positive** | Learning     | A developer replies to a Vidoc PR comment and confirms the finding.                                                                                             |
| **Pull request**            | Memory entry | A developer mentions the Vidoc bot in a PR and asks it to remember a fact.                                                                                      |
| **Slack**                   | Memory entry | Vidoc Cloud only. A user mentions Vidoc in Slack and asks it to remember a fact. Refer to [Slack](/web-app/slack#memory-from-slack).                            |

Replies and mentions in PRs work on GitHub and GitLab only. For PR replies, Vidoc must link the platform user to a Vidoc user. Each user sets this link in [SCM usernames](/web-app/settings#scm-usernames). For the PR commands, refer to [Feedback in pull requests](/pull-requests/feedback) and [Commands and questions](/pull-requests/commands).

## Find entries

* Filter by source (**All sources**), by repository (**All codebases**), and by status (**Any status**: active or superseded).
* Use **Search memory…** to search the text.

The table shows the entry, the source, who created it, the scope, the date, and the status.

Click an entry to open its panel. The panel shows the full text (**Full learning**) and the origin (**Where this came from**), for example the original comment and a link to the source.

## Change entries

You need the Admin or Security Engineer role. Other roles can only view Memory.

| Action              | What it does                                                                                                                    |
| ------------------- | ------------------------------------------------------------------------------------------------------------------------------- |
| **Mark superseded** | Vidoc stops using the entry.                                                                                                    |
| **Reactivate**      | Vidoc uses a superseded entry again.                                                                                            |
| **Edit**            | Active learnings only. Change the **Title**, the **Context**, the **Reasoning**, and **When this memory applies to a finding**. |

To change a memory entry, teach the bot the correct fact again, then supersede the old entry.

### When this memory applies to a finding

This setting tells the validation agent how to use a learning:

| Option                                 | Effect                                                                                      |
| -------------------------------------- | ------------------------------------------------------------------------------------------- |
| **Treat as false positive (suppress)** | The agent uses the learning as strong evidence that a matching finding is a false positive. |
| **Confirm as a real issue**            | The agent uses the learning as evidence that a matching finding is a real issue.            |
| **Flag for human review**              | The agent uses the learning as context only. Vidoc does not create a review task.           |

## Review Memory regularly

Learnings from PR replies and saved memory entries from PRs and Slack are used without a separate approval step. For example, a false-positive reply on a PR ignores the issue and creates a learning at once.

We recommend that the AppSec team examines new entries every week:

1. Filter by source: **False positive**, **Confirmed true positive**, **Pull request**, and **Slack**.
2. Open each new entry and read it.
3. Click **Mark superseded** for each entry that is wrong.

<Warning>
  Superseding a learning does not reopen the issue that it ignored. To reopen the issue, open [Findings](/web-app/findings), set **State** to **Ignored**, find the issue, and set it to **Open**.
</Warning>