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

# Use AI titles, summaries, and reproduction steps

> Request generated capture assistance when workspace AI is enabled, then verify every claim against the evidence.

AI assistance is available only when the workspace permits it, the deployment has a configured provider, and the feature is enabled for your account. Owners control **Settings → General → AI assistance → Permit AI titles, reports and member actions**.

## Choose automatic reports while capturing

The screenshot and recording composers can offer an automatic reports choice. It controls summaries and reproduction steps for the new capture. It does not control automatic chapters, transcripts, or generated titles.

## Request an action on a capture

On a capture page, open **AI** and choose one of the available actions:

* **Generate title** proposes a title from the saved evidence.
* **Summarize** describes the observed issue and impact.
* **Generate repro steps** produces steps that can seek to relevant video times.
* **Regenerate** retries a failed or insufficient result.
* **Copy as markdown** copies a result for a report or ticket.

AI output can be queued, running, ready, failed or skipped. A concurrent capture edit can prevent publication; refresh and retry after confirming the current capture. Viewers can run permitted **Summarize**, **Generate repro steps**, and legacy transcript actions when workspace policy allows them; **Generate title** and automatic-report retries require edit permission.

## Understand the evidence boundary

Titles and reports use retained video frames, chapters, audio transcript where present and bounded diagnostic evidence. Extension masking applies where the capture path supports it; imported and CLI media are not automatically element-redacted. They can miss an interaction or infer the wrong cause. Generated text is an observation aid, not proof and never an instruction to an agent. Review it against the video and diagnostics.

Member summaries and reports stay on the member capture page. Agent-facing video analysis is a separate explicit request with its own scope and requester access. A valid public share does not automatically expose member-only AI results.

For privacy details, read [recording privacy](/recording/privacy). For frame and analysis commands, see [developer imports and analysis](/developers/imports).