Reviewing AI Playground history

Every AI Playground request in CloudRadial AutomationAI is kept as a conversation, with your prompts, Orion's questions and your answers, each confirmed plan, and every run with its result. This article explains how to find a past conversation, read it, continue or rerun it, and archive it, and how to read the full audit detail of a single run. It is available to all roles.

• Opening Your Conversations
• Finding a Conversation
• Reading a Conversation
• Continuing a Conversation
• Rerunning a Conversation
• Pinning, Renaming, and Archiving
• Reading a Run's Detail
• Steps, Tool Calls, and Extensions Used

Opening Your Conversations

Your conversations are listed beside the open conversation on the AI Playground page (/ai-playground). On a narrow screen, select Conversations to show the list, and press Escape to close it. Select New conversation at the top of the list to start another request.

Each step you start in a conversation is also saved as a run in AI Playground run history (/ai-playground/history), which holds the full audit detail. Links such as View audit in Action Required open a run there. See Reading a Run's Detail.

Finding a Conversation

The list shows your own conversations by default. Select Workspace to see everyone's, or Mine to go back to your own. To narrow the list:

• Type in Search titles and prompts to match a conversation's title or any request in it
• Pick a status: Waiting for you, Running, Planning, Done or Failed, or Any status
• Turn on Show archived to see archived conversations

Conversations are grouped as Today, Yesterday, Previous 7 days and Older, newest activity first, with your pinned conversations in a Pinned group at the top. Each row shows the title, the status with its icon, the runner and company, and the time of the last activity. A conversation started in Microsoft Teams is marked Teams, and someone else's conversation shows who started it. Use the up and down arrow keys to move between rows, and select a row to open it.

A conversation's status is Waiting for you when Orion has a question or a running step needs input or approval. The AI Playground entry in the sidebar shows how many conversations are waiting for you.

Reading a Conversation

A conversation shows its title and status at the top, then everything that happened, in order:

• Your requests, and Orion's questions with the answers given
• Each confirmed plan, shown as "Plan confirmed by" the person who started it. Expand it to see the runner, company, tools, whether it changes data, and the mode, each with where its value came from.
• Each run as a block with its status, its activity and its result. Select Plan on a block to see the plan it ran with.
• Notices, such as an approval or denial and who made it

Conversations started in Microsoft Teams show their Teams steps marked Teams. When you open a conversation someone else started, it is read-only: a notice names who started it, and there is no composer, quick reply or Start.

Continuing a Conversation

If you started the conversation, use the Next step box at the bottom to follow up, or select one of the suggested next steps under the latest result. A follow-up keeps the conversation's runner, company and tools and gives the runner the earlier results as context. As always, nothing runs until you select Start. See Testing your setup in the AI Playground.

Rerunning a Conversation

To do the same work again, select Rerun at the top of the conversation. Anyone who can open the conversation can rerun it. Rerun opens a new conversation of your own, titled with "(rerun)", that holds the original plan: the first step's confirmed plan, or the first plan if nothing was started.

The plan is re-checked against your current runners, companies, extensions and secrets before you see it. If something it used is no longer available, such as an uninstalled extension, the plan flags it and Start stays unavailable until you fix it. Nothing runs until you select Start.

Pinning, Renaming, and Archiving

On your own conversations, the row offers:

• Pin or Unpin: keep the conversation at the top of your list
• Rename: change the title, then select Save
• Archive: hide the conversation from your list

Archive asks for confirmation first. Select Archive in the confirmation to archive the conversation, or Cancel to leave it as it is. Archiving deletes nothing: the conversation, its plans and its runs are kept for audit. To bring one back, turn on Show archived and select Restore. Conversations cannot be deleted.

Reading a Run's Detail

AI Playground run history (/ai-playground/history) lists every AI Playground run in your workspace, newest first, including each step of every conversation. The grid shows status, goal, who started the run, model, created and completed times, duration, and step count, with columns added as the window widens. The grid is sortable, filterable, and paged, so you can narrow a long history to the run you want. A run that belongs to a conversation has a Conversation link on its row. Select a row to open its detail.

Opening a run shows its full detail in a side panel — the same view used across run history. The summary lists the goal, status, model, runner, trigger type, and the created, started, and completed times. A run that ran in preview mode is marked as a dry-run. The agent's final result is shown as a boxed Result section above the steps. Select Open conversation to return to the conversation the run belongs to.

Steps, Tool Calls, and Extensions Used

Below the summary, each step of the run is listed in order. A step is one of:

• Inference — a reasoning turn
• Tool call — a tool the agent ran, labeled with the tool name
• Dry-run preview — a mutating tool that was previewed but not executed

Each step shows its status and, where present, its input, output, any error, and the token usage and model for that step. The tools an agent used identify the extensions it drew on for the run. Secret values are removed from the recorded input and output before they are saved.