Skip to content

11-analytics.md

Latest commit

 

History

History
92 lines (57 loc) · 3.28 KB

11-analytics.md

File metadata and controls

92 lines (57 loc) · 3.28 KB

Analytics Dashboard

Overview

The Analytics dashboard gives you insight into how your agents are being used — conversation volume, response times, popular questions, and knowledge base search performance. Access it by clicking the "Analytics" button on the dashboard.

Period Selector

At the top right you can choose the time window: 7d, 30d, or 90d. All metrics, charts, and tables update to reflect the selected period.

Summary Cards

Four cards at the top show key metrics at a glance:

Card What it Means
Total Conversations Number of unique conversations started in the period
Total Messages Total user + assistant messages exchanged
Avg Response Time Average time from user message to complete agent response
KB Search Hit Rate Percentage of knowledge base searches that returned at least one result

How to interpret

  • High response time (>3s): Consider switching to a faster model (deepseek-chat) or reducing Max Tokens in AI Response nodes.
  • Low KB hit rate (<50%): Your knowledge base may not cover the questions users are asking. Check the "Common Questions" table and add sources that address those topics.

Daily Conversations Chart

An area chart showing the number of new conversations per day. Use this to:

  • Spot usage trends (growing, declining, or seasonal)
  • Identify spikes after sharing the agent link or embedding the widget
  • Correlate with changes you made to the agent's flow or knowledge base

Top Agents

A ranked list of your agents ordered by conversation count. Each entry shows:

  • Agent name
  • Conversation count — how many unique conversations
  • Message count — total messages (user + assistant)

Agents with zero conversations are hidden from the list.

Common Questions

The most frequently asked first messages across all conversations. This helps you understand what users typically ask about.

  • Messages with 4+ consecutive digits are filtered out (privacy protection)
  • Long messages are truncated to 60 characters
  • The count shows how many times that exact first message appeared

How to use this

If you see a question that your agent handles poorly:

  1. Add relevant content to the knowledge base
  2. Adjust the system prompt to handle that topic
  3. Add a specific flow branch for common intents

Average Response Time Chart

A line chart showing daily average response time in milliseconds. This tracks end-to-end time from when the user sends a message to when the full response is delivered.

Factors that affect response time:

  • Model choice — GPT-4o is slower than deepseek-chat
  • Max Tokens — higher limits mean longer generation time
  • KB Search — re-ranking adds latency but improves quality
  • API calls / webhooks — external calls in the flow add latency

How Analytics Data is Collected

Analytics are collected automatically — no configuration needed:

  • Conversation and message counts are derived from existing database records
  • Response time is tracked via AnalyticsEvent records created on each chat API call
  • KB search stats are tracked each time a KB Search node executes

Analytics tracking is fire-and-forget — it never blocks or slows down the chat response.