> For the complete documentation index, see [llms.txt](https://docs.artific.nl/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.artific.nl/en/for-administrators/assistants/session-logs.md).

# Chat history

A read-only record of the conversations people have had with an assistant. Found on the **Chat history** tab of an assistant's management area, next to **Settings**, **Runs** and **Triggers**.

## What it is for

The chat history is the feedback loop. It is the only reliable way to learn what people actually ask, as opposed to what you assumed they would ask.

Read it regularly, especially in the first weeks after launch. Almost every improvement to an assistant's instructions, knowledge and suggested questions comes from noticing a pattern here.

## Reading the chat history

The screen is a list on the left, a transcript on the right.

Each entry in the list shows the conversation's topic, a one-line summary, and when it was last active. Two markers can appear:

| Marker                  | Meaning                                                                   |
| ----------------------- | ------------------------------------------------------------------------- |
| **Initiated by {name}** | The conversation was started by another assistant delegating to this one. |
| **Project chat**        | The conversation happened inside a project.                               |

Scrolling loads more; the list tells you when you have reached the end.

Select a conversation to read the full transcript, including any files the user attached. Above it you see when the conversation started, when it was last updated, and identifiers for the session and the user, each with a copy button, useful when you need to refer to a specific conversation in a support request.

## Steps behind an answer

Under each reply in the transcript you can open the steps the assistant took to reach it: which knowledge it searched and which tools it used. It is the same record as for [runs](https://docs.artific.nl/en/for-administrators/assistants/runs.md), so you can trace why an answer came out the way it did.

**Technical details** exports everything that explains the conversation: its messages, the steps under each reply, what each model call received and returned, and the tools on offer. Useful when you investigate a report in detail.

## What to look for

* **Questions that got a poor answer.** Usually a gap in the instructions or the knowledge, not a fault of the model.
* **Questions outside the assistant's scope.** If many people ask the same off-topic thing, either extend the assistant, add a sub-agent, or say clearly in the opening message that it is not covered.
* **The same question, phrased differently.** Good candidates for [suggested questions](https://docs.artific.nl/en/for-administrators/assistants/suggested-questions.md).
* **Conversations that ended abruptly.** Often a sign the assistant misunderstood and the user gave up.
* **Requests to speak to a person**, where Direct Chat is on. A rising number means the assistant is not covering something it should.

## Availability and privacy

The chat history contains whatever your users typed, which may include personal or sensitive information. Treat access to this tab as a privileged permission and grant it deliberately.

{% hint style="info" %}
The chat history can be hidden for an organisation. If the tab is not there, it has been switched off for your organisation. Your account manager can explain the setting. People always keep seeing their own recent chats, whatever this setting.
{% endhint %}

Conversation history can also be deleted automatically after a set period. That is configured centrally under [Organization settings](https://docs.artific.nl/en/for-administrators/organization/settings.md).
