AI Chat History¶

Overview ¶

The Chat History feature allows each authenticated user of the F5 Insight AI Assistant to maintain persistent, server-side conversation threads. Your chat history survives page refreshes, tab closes, and session logouts.

Each thread retains full message history up to a configurable limit, with automatic AI-generated summarization when threads grow large. Your conversations expire after 24 hours of inactivity.

Key concepts ¶

Concept	Description
Thread	A named conversation between one user and the AI Assistant. Isolated per user. No cross-user access.
TTL	Rolling 24-hour window from the last message. F5 Insight automatically soft-deletes threads with no activity for 24 hours (hidden from the API). Soft-deleted threads are permanently hard-deleted after 30 additional days.
Message limit	200 messages per thread (hard cap after summarization).
Summarization	At 250 messages, the oldest messages are folded into a single AI-generated summary, bringing the thread back to 200. Summarization runs in the background and doesn’t block the current response.
Thread limit	50 active threads per user. When the limit is reached, the oldest thread is evicted silently to make room for the new one.
TTL worker	Background job that runs every 24 hours to soft-delete expired conversations.

Role-based access ¶

Chat History endpoints require the chatManagement permission group.

Role	Read	Write	Delete
Viewer	No	No	No
Operator	Yes	Yes	Yes
Admin	Yes	Yes	Yes

Note

The Viewer role has no access to chat history. Both Operator and Admin roles have full access.

Use the web console ¶

The AI Assistant opens with a collapsible conversation sidebar on the left. The sidebar lists all saved conversations for the current user, sorted by most recent activity.

The sidebar header shows a live count (for example, 3/50). The count turns orange (warning) above 40 threads and red (critical) above 45, indicating that F5 Insight evicts the oldest thread when you reach the limit.

The sidebar can be collapsed to a narrow icon strip using the toggle button in the header.

Start a new conversation ¶

Select the New conversation icon (speech bubble with +) in the sidebar header. F5 Insight doesn’t create a thread immediately. It creates one automatically when you send your first message and the title is derived from that message.

Resume a conversation ¶

Select any conversation row in the sidebar. The full message history loads into the chat area.

Note

If a background response is still generating for that thread, a spinner icon appears on the row.

Delete a conversation ¶

Hover over any conversation row and select the delete icon (bin) that appears on the right.

Warning

Deletion is immediate and permanent. You can’t recover the conversation or its messages.

Export as PDF ¶

There are two PDF export options in the UI:

Export an entire conversation ¶

Open the conversation you want to export.
Select the ⋮ (three-dot menu) icon in the chat header.
Select Export chat as PDF.
The full conversation thread is exported to a PDF file.

Export a single response ¶

Open the conversation containing the response you want to export.
Hover over the assistant message bubble to reveal its action menu.
Select Export response as PDF.
The selected AI response and the user message that preceded it are exported to a PDF file.

Note

Both export options are available only when a saved conversation is open (not on the homepage).
Both options require the chatManagement.post permission.

Use the REST API ¶

All chat history tasks available in the web console are also available through the F5 Insight REST API, enabling automation and integration with existing tooling.

Authentication ¶

All chat history endpoints require a valid JWT access token:

Authorization: Bearer <token>

Responses use standard HTTP status codes.

List conversations ¶

GET /api/chat/history/conversations

Response (200 OK)

{
  "conversations": [
    {
      "id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
      "user_id": "user123",
      "title": "My investigation",
      "created_at": "2026-07-01T09:00:00Z",
      "last_msg_at": "2026-07-01T11:30:00Z",
      "expires_at": "2026-07-02T11:30:00Z",
      "message_count": 14
    }
  ]
}

Create a conversation ¶

POST /api/chat/history/conversations
Content-Type: application/json

{
  "title": "My investigation"
}

title is optional (max 500 characters). If omitted, the UI derives a title from the first message.

Returns 201 Created with the conversation object.

Get a conversation (with messages)¶

GET /api/chat/history/conversations/{id}

Response (200 OK)

{
  "conversation": {
    "id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
    "user_id": "user123",
    "title": "My investigation",
    "created_at": "2026-07-01T09:00:00Z",
    "last_msg_at": "2026-07-01T11:30:00Z",
    "expires_at": "2026-07-02T11:30:00Z",
    "message_count": 14
  },
  "messages": [
    {
      "id": "...",
      "conversation_id": "xxxxxxxx-xxxx-xxxx-xxxx-xxxxxxxxxxxx",
      "user_id": "user123",
      "role": "user",
      "content": "How many virtual servers...",
      "content_metadata": {},
      "created_at": "2026-07-01T09:00:00Z",
      "is_summarized": false
    }
  ],
  "context_summarized": false
}

Response field reference

Field	Description
`context_summarized: true`	Indicates the earliest messages have been replaced by an AI-generated summary.
`is_summarized: true`	On an individual message, marks that message as the summary message itself.

Delete a conversation ¶

DELETE /api/chat/history/conversations/{id}

Returns 204 No Content. No response body.

Add a message to a conversation ¶

POST /api/chat/history/conversations/{id}/messages
Content-Type: application/json

{
  "role": "user",
  "content": "How many virtual servers are on device-01?",
  "content_metadata": {}
}

Parameters

Parameter	Required	Description
`role`	Yes	Must be `"user"` or `"assistant"`
`content`	Yes	Message content. Max 100 KB.
`content_metadata`	No	Arbitrary JSON object (for example, chart data)

Returns 201 Created with the saved message.

Export a conversation as PDF ¶

POST /api/chat/export/pdf
Content-Type: application/json

Parameters

Parameter	Required	Description
`type`	Yes	`"conversation"` - full export; `"single"` - one assistant response and its preceding user message
`conversation_id`	Yes	UUID of the conversation
`message_id`	Only when `type="single"`	UUID of the specific assistant message
`session_id`	No	Optional session identifier for audit tracking

Returns a PDF file as an application/pdf download.

Configuration ¶

Variable	Default	Description
`CHAT_MAX_CONVERSATIONS_PER_USER`	`50`	Maximum active threads per user. When reached, the oldest thread is evicted silently.
`CHAT_SUMMARIZATION_TRIGGER_COUNT`	`250`	Message count that triggers automatic summarization.

Limits ¶

Limit	Value
Max messages kept per thread	200
Summarization trigger	250 messages
Max active threads per user	50 (configurable)
Thread TTL (inactivity)	24 hours rolling; soft-deleted, then hard-deleted after 30 days
Max message content size	100 KB
Max conversation title length	500 characters

Troubleshooting ¶

Thread disappeared after 24 hours ¶

Chat threads are soft-deleted after 24 hours of inactivity and hidden from the API. They are permanently hard-deleted 30 days later.

Note

The TTL worker runs every 24 hours, so a thread might remain visible for up to 24 hours past its expiry time.

Orange (warning) at 40 threads
Red (critical) at 45 threads

Tip

Export important threads via PDF before they are evicted. See Export as PDF for instructions.

Previous Next

AI Chat History¶

Overview ¶

Key concepts ¶

Role-based access ¶

Use the web console ¶

Start a new conversation ¶

Resume a conversation ¶

Delete a conversation ¶

Export as PDF ¶

Export an entire conversation ¶

Export a single response ¶

Use the REST API ¶

Authentication ¶

List conversations ¶

Create a conversation ¶

Get a conversation (with messages)¶

Delete a conversation ¶

Add a message to a conversation ¶

Export a conversation as PDF ¶

Configuration ¶

Limits ¶

Troubleshooting ¶

Thread disappeared after 24 hours ¶

Accessing an expired conversation returns 410 Gone ¶

Summarized context: AI response lacks earlier details ¶

Thread limit reached ¶