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

# Conversations

> Browse every conversation an agent has handled, inspect the message thread, and trigger analysis on demand.

The **Conversations** page is the per-agent inventory of every conversation Insights knows about, with an analysis-status filter, message-level inspection, and on-demand analyze controls.

<Frame>
  <img src="https://mintcdn.com/prismeai/C1y4E8zihxFJ6QR0/images/insights/conversations.png?fit=max&auto=format&n=C1y4E8zihxFJ6QR0&q=85&s=67d9f0f4b5525be2e20c7bc88d14e46e" alt="Conversations page with the status filter strip, conversation table, and detail panel showing messages" width="1440" height="900" data-path="images/insights/conversations.png" />
</Frame>

## Layout

The page is a two-pane split:

* **Left** — A status filter strip and the conversations table.
* **Right** — A detail panel for the selected row, with metadata, the full message thread, and the analyze button.

## Analysis statuses

Every conversation is in one of five states:

| Status           | What it means                                                                                                       |
| ---------------- | ------------------------------------------------------------------------------------------------------------------- |
| **Not analyzed** | Not yet picked up by the analysis pipeline. Either too new for the batch pass, or never queued.                     |
| **Pending**      | Analysis is queued and will be processed soon.                                                                      |
| **Processing**   | Analysis is currently in progress.                                                                                  |
| **Analyzed**     | Analysis complete. Click the row to read the result on this page or use [Insights](/products/ai-insights/insights). |
| **Failed**       | The analysis pass errored. The detail panel shows the error message. Click **Re-analyze** to retry.                 |

The filter strip at the top of the table lets you scope the table to **All**, **Not Analyzed**, **Pending**, **Processing**, **Analyzed**, or **Failed**, with counts per status.

## Per-row controls

Each row in the table shows:

* A short conversation ID and message count.
* A preview of the last message.
* A status badge.
* An action button — **▶ Analyze** for new conversations, **↻ Re-analyze** for those already analyzed.

Click the row to load the conversation in the detail panel on the right.

## The detail panel

Open a conversation and the right pane shows:

* The conversation ID and current status.
* A metadata grid: **Messages** count, **Created**, **Score** (when analyzed, /100, color-coded), **Analyzed At**.
* The message thread with sender avatars (user / bot / system), timestamps, and per-message token counts where present.
* A **Load earlier messages** button if the thread is long.
* An **Analyze** / **Re-analyze** button at the bottom.

If a previous analysis attempt failed, an error block appears above the analyze button with the failure reason returned by the pipeline.

## Batch analysis

To analyze multiple conversations in one pass:

<Steps>
  <Step title="Tick the rows you want">
    Use the per-row checkboxes, or the header checkbox to select all visible rows.
  </Step>

  <Step title="Click 'Analyze N selected'">
    The button appears at the top of the table once at least one row is selected. The label includes the count.
  </Step>

  <Step title="Wait for completion">
    Each conversation transitions through Pending → Processing → Analyzed (or Failed). The analyze button shows **Analyzing…** during the run.
  </Step>
</Steps>

<Frame>
  <img src="https://mintcdn.com/prismeai/C1y4E8zihxFJ6QR0/images/insights/conversations-batch.png?fit=max&auto=format&n=C1y4E8zihxFJ6QR0&q=85&s=bf5ba658f6775137261de2335f586b2d" alt="Conversations table with multiple rows selected and the Analyze N selected button visible" width="1440" height="900" data-path="images/insights/conversations-batch.png" />
</Frame>

The maximum batch size is set per workspace — the default is **100 conversations per batch**.

## How automatic analysis works

You don't have to click anything for the bulk of analysis to happen. Insights runs a recurring pass that picks up conversations matching all of the following:

* The conversation has been **inactive for at least 2 hours** (no new messages).
* It contains **at least 2 messages**.
* It has not been analyzed yet, or has grown by 5+ messages since the last analysis (re-analysis trigger).

The agent's **default daily analysis budget** is 1,000 conversations — past that, conversations queue up and continue the next day. See [Queue](/products/ai-insights/queue) for monitoring.

<Note>
  Whether real-time analysis (triggered as soon as a conversation goes inactive) runs in addition to batch is a workspace-level setting. By default, analysis runs in batch only.
</Note>

## When you'd analyze on demand

* **Recently changed criteria.** Re-evaluate a slice of conversations against your new criteria without waiting for the batch pass.
* **Failed conversations.** Click **Re-analyze** after fixing a configuration issue.
* **Spot-check.** Pull one specific conversation forward to validate that your criteria produce sensible answers.

## Where to go next

<CardGroup cols={2}>
  <Card title="Read the analyzed result" icon="magnifying-glass-chart" href="/products/ai-insights/insights">
    See the criterion answers, summary, key moments, and token usage.
  </Card>

  <Card title="Tune evaluation criteria" icon="list-check" href="/products/ai-insights/criteria">
    Add or refine the questions the analysis pass answers.
  </Card>

  <Card title="Monitor the queue" icon="gauge" href="/products/ai-insights/queue">
    Watch backlog and throughput, especially after a bulk reanalyze.
  </Card>
</CardGroup>