Skip to main content
History is your audit trail for everything your flows do. Use it to read a customer’s full chat transcript, follow how a single run moved through the canvas, and debug failures. Menu: Insights → History Route: /admin/transcripts (admin_transcripts_conversations) Permission: comerix.insights.history History has two surfaces that work together:
  • Conversations — the customer-facing thread: every message exchanged, the variables collected, and telemetry, grouped by visitor.
  • Executions — the engine-facing record: each time a flow ran, node by node.
Conversations vs. executionsA conversation is one ongoing chat with a customer. An execution is a single run of a flow. One conversation can contain many executions — each trigger that fires while the customer chats starts a new execution against the same conversation. Start in Conversations to see what the customer experienced; open Executions to see how the engine behaved. See Lifecycle for the full status model and Concepts for the vocabulary.
A conversation transcript. An execution with its event timeline.

Browse conversations

The conversations list with search and filters. The conversations list (/admin/transcripts) is paginated at 25 per page. Use the filters to narrow it down:
Empty conversations are hiddenThe list only shows conversations that carry at least one chat message. A visitor who opened the widget and produced only telemetry (page views, scrolls) but never exchanged a message has nothing to replay, so that telemetry-only conversation is omitted from the list. It still contributes its signals to Goals and analytics.
FilterWhat it does
FlowShow only conversations that ran a specific flow
Statusrunning, waiting_input, waiting_time, completed, failed, aborted
Date rangeLimit to conversations seen within a window
Columns in the list:
ColumnMeaning
CustomerThe identified customer, or (anonymous) for unknown visitors
ChannelWhere the chat happened (e.g. the web widget)
MessagesNumber of messages in the thread
Latest runThe most recent execution for this conversation
Last seenWhen the customer was last active

Search across messages

Use the full-text search box to find a conversation by what was said in it. Search runs across the text of all messages. The results panel shows matching message snippets with the matched words highlighted. Click a result to jump straight into that conversation, positioned at the matched message — handy when you remember a phrase but not the customer.

Read a conversation

Open any row to reach the conversation detail (/admin/transcripts/conversations/{id}). It shows the complete thread in order:
  • Transcript — user, agent, and system messages, each with a timestamp.
  • Form requests — inline form prompts shown to the customer, with the field values they submitted.
  • Token usage — where a message was produced by a model, a small badge under the bubble shows its token count and cost (for example 42 tokens · $0.0012). Hover the badge to see which model generated the message. Messages with no recorded usage show no badge.
  • Telemetry — widget telemetry events folded into the timeline where they occurred.
The context column on the right summarises the conversation — customer, channel, locale, and first/last seen — and lists the conversation variables collected during the chat (the conversation-scope values your flows set), with a “No variables.” note when none were recorded.
Telemetry events here are the same signals that Goals match against. If a goal didn’t fire when you expected, the transcript shows whether the triggering event was actually recorded.

Browse executions

The executions list. Switch to the executions list (/admin/transcripts/executions, admin_transcripts_executions) to see flow runs directly. It is paginated at 50 per page and supports the same flow / status / date-range filters.
ColumnMeaning
FlowThe flow that ran — links to its editor
StatusCurrent run status (see Lifecycle)
Trigger typeWhat started the run
StartedWhen the run began
DurationHow long it took
Error codePopulated only for failed runs

Debug a run

Open an execution to reach its detail page (/admin/transcripts/executions/{id}). This is the place to diagnose a failure. The summary lists the flow, status, version, trigger, timestamps, duration, and error (if any), followed by the trigger payload that started the run. Below that is the ordered event timeline — the full step-by-step record of the run:
EventWhen it appears
NodeStarted / NodeCompletedA node began / finished successfully
NodeFailedA node raised an error
NodeSkippedA node was bypassed by branch logic
WaitScheduled / WaitFiredA wait was queued / resumed
FormDisplayed / InputReceivedA form was shown / answered
ExecutionPausedThe run paused (e.g. waiting on input or time)
ExecutionCompleted / ExecutionFailed / ExecutionAbortedThe run ended
Working with the timeline:
  • Click any row to expand it and see its output or error JSON.
  • Failed rows auto-expand so the error is visible immediately.
  • Each node event deep-links to the node in the canvas, so you can open the exact step that misbehaved.
Tracing a failure
  1. Filter executions by failed status and find the run.
  2. Open it and read the error and error code in the summary.
  3. Scan the timeline to the auto-expanded NodeFailed row and read its error JSON.
  4. Click through to the node in the canvas to fix it.
Look up what an error code means in Errors reference. For queue and worker behaviour behind waits and retries, see Operating reference.

Permissions

ActionPermission
View conversations, executions, and all detail pagescomerix.insights.history
Everything under History is gated by comerix.insights.history. Users without it won’t see the History entry in the Insights menu.