> ## Documentation Index > Fetch the complete documentation index at: https://docs.hubtalk.ai/llms.txt > Use this file to discover all available pages before exploring further. # Incidents > Review and investigate incidents affecting HubTalk AI v4. # Incidents The **Incidents** page lists events that may affect agents, calls, campaigns or integrations. Use it to understand what happened, when the problem started, and whether it is still active. ## What incidents are Incidents are events detected by background observers during conversations. They help the team find situations that require review, such as an agent going off topic or a caller's speech becoming unintelligible. The page is used for triage and review. It does not change the conversation by itself; it shows the detected event and the current reaction or correction status. ## What the user sees At the top of the page, the user sees the total number of incidents and filters: * **Search hint or type** — search by the incident hint or event type. * **All agents** — show incidents for a selected agent. The list holds live agents only: deleted and archived ones are gone from it. * **Any severity** — filter by severity. * **Status** — filter open or other incident states. * **Real calls only** — exclude incidents from test calls. * **Date range** — select the period to review. Each incident row shows: * the agent name, date, time, and call mode such as live * the incident label and event key, such as `off_topic_counter` or `unintelligible_speech` * the severity badge, for example high * the observer's hint or recommended instruction * the reaction status, such as no correction * whether the incident is open * a control to open the incident details Use the list to prioritize open high-severity incidents, understand the observer's reasoning, and decide whether the conversation or agent configuration needs correction. Incidents list with filters Expand a row to see its full detail — actions taken, detector confidence, the trigger quote, who reviewed it, and controls to mark it resolved, flag it as a false positive, or reopen it. Expanded incident row with actions, confidence, trigger quote, and reviewer ## Incidents inside a call Open a call's details to see incidents in the context of the conversation: * An incident sits **at the very turn that triggered it** — as its own block between the utterances, not in a list of its own. The block shows the incident type, the severity, the observer's hint, and whether a correction reached the agent. * The **Background observer** panel gathers all observer activity for that call: the number of checks performed, violations detected, and incidents raised, together with a timeline of every observer event during the call. ## Incident review For each incident, review: * the affected service or workflow * incident severity and current status * start time and latest update * affected agents, calls, or campaigns * the reason and available resolution details Use incidents together with **History & reports** when investigating failed calls, delayed analytics, or webhook delivery problems.