Skip to content

Interaction Logs

The Logs page provides a detailed view of every individual AI interaction in your tenant. While the Usage page shows aggregated counts, the Logs page shows each conversation turn with full context — what the user asked, how the agent responded, whether an error occurred, and what feedback the user gave.

Logs list view showing user inputs, timestamps, and user columns

Each row on the Logs page represents a single AI interaction — one question from a user and one response from an agent. This gives you the granularity needed to investigate specific issues, review quality, and understand user behavior.

ColumnWhat it shows
TimestampWhen the interaction occurred
UserWho triggered the interaction (shown as user ID or email)
Chat sessionWhich conversation this interaction belongs to — useful for seeing multi-turn context
Response typeHow the agent responded (see below)
IntentWhat the user was trying to accomplish (see below)
ErrorWhether the interaction failed
ProtocolWhether the request came via the UI/API (http) or an MCP client (mcp)
Product areaWhich Alation product generated this interaction
TierWhether this was free or paid usage
SentimentUser feedback if given — thumbs up (positive) or thumbs down (negative)
EventsWhat happened during this request — which tools were called, how many times

The agent classifies each response into one of these categories:

TypeWhat it means
RAG answerThe agent retrieved context from the catalog and generated an answer grounded in real data
Connector troubleshootThe agent helped diagnose a connector issue
DirectThe agent answered directly from its knowledge without needing to search the catalog
ErrorSomething went wrong and the agent couldn’t produce a useful response

The system classifies what the user was trying to do:

IntentWhat it means
How-toThe user asked how to accomplish something (e.g., “How do I create a data product?”)
TroubleshootingThe user reported a problem (e.g., “My connector is failing with timeout errors”)
ConceptualThe user asked for an explanation (e.g., “What is a trust flag?”)

Each log entry includes a list of events that occurred during that interaction. For example, a single interaction might show:

  • search_catalog (tool) × 2 — the agent searched the catalog twice
  • get_data_schema (tool) × 1 — the agent retrieved a schema
  • alamigo (agent) × 1 — one agent run

This helps you understand the cost and complexity of individual interactions.

Expanded log entry showing the user query and full AI response

When you expand a log entry, you can see:

  • User query — the exact question the user typed
  • AI response — the full text of the agent’s reply

You can narrow the Logs page by:

FilterUse it to…
Chat sessionSee all interactions within a single conversation
Response typeFind only errors, only RAG answers, etc.
Error flagQuickly isolate failed interactions
Time rangeFocus on a specific period (e.g., “last Tuesday when the spike happened”)

When you see an unexpected spike on the Usage page, switch to the Logs page for the same time range. You can see exactly which users triggered the spike, what questions they asked, and which tools were invoked. This is the primary way to answer questions like “Was this spike from Ask Alation or from Agent Studio?”

Filter by negative sentiment to find interactions where users gave a thumbs-down. Review the user’s question and the agent’s response to identify patterns — are users asking questions the agent isn’t equipped to handle? Is context missing from the catalog?

Filter by is_error = true to see all failed interactions. Each entry includes a trace ID that can be shared with your support team for deeper investigation.

Review who is using AI features, what they’re asking, and how they’re interacting. The intent category and response type columns give you a high-level picture without needing to read every message.

RoleAccess
Server AdminFull access to all interaction logs, including message content
All other rolesNo access to the Logs page