Requests

Overview

The Requests page provides a comprehensive view of all platform interactions, allowing administrators and users to track, analyze, and manage request history. This detailed interface displays critical information about each request, including timestamps, request content, responses, model usage, token consumption, project details, user information, and performance metrics.

Accessing the Requests Page

1. Navigate to the left sidebar navigation panel
2. Click on "Monitoring" in the menu options
3. Select "Requests" from the submenu (indicated by the arrow icon)
4. The Requests interface will load, displaying a tabular view of all request data

Requests Interface Components

Time Range Selection

Located near the top of the interface, this control allows filtering of request data by different time periods:

• 24H: Displays requests from the past 24 hours
• 7D: Displays requests from the past 7 days
• 1M: Displays requests from the past month
• 3M: Displays requests from the past 3 months
• All: Displays all historical request data

Export Functionality

• EXPORT ALL DATA: Button located next to time filters that allows exporting the current view of request data for external analysis or reporting

View Controls

Located at the top-right corner of the table:

• Search: Magnifying glass icon to search through requests
• Table View: Toggle between different table view modes
• Fullscreen: Expand the request table to fullscreen mode

Request Table

The main component displaying detailed request information in tabular format. The table includes various columns that can be customized and sorted:

Request Information Columns

Column	Description	Example Value
Selection Checkbox	Allows selection of individual or multiple requests	☐
Created At	Timestamp of when the request was created	MM/DD/YYYY, H:MM PM
Request	Text of the user's request/query	"system:You are an expert at..."
Response	Text of the system's response	"Feature description..."
Model Name	AI model used to process the request	[Model identifier]
Token Name	Type of token used for the request	[Token type]
Total Units	Number of tokens consumed by the request	500
Cost	Cost of the request	$0.00XX
Latency	Response time in seconds	1.25s

Additional Available Columns

Column	Description	Example Value
Project Name	Name of the project the request belongs to	[Project name]
Project Type	Category or type of the project	[Project type]
User	User who initiated the request	[User identifier]
Feedback	User feedback on the response	No Feedback
Reaction	User reaction to the response	No Reaction

Request Details Modal

When clicking on a specific request, a detailed view appears in a modal window showing:

• Created At: Exact timestamp of the request
• Embedding Model: Model used for embeddings
• Completion Model: Model used for generating responses
• Tokens: Number of tokens used
• Latency: Response time in seconds
• Status: Request status (e.g., Success)
• User: User identifier

The modal also includes a conversation view showing:

• User Message: The original query
• Assistant Response: The system's response

A "Close" button at the bottom allows closing the modal to return to the request list.

Table Controls

• Column Headers: Each column has sortable headers (indicated by arrows)
• More Options: Each column has a menu with additional options (indicated by "..." icon)
• Pagination Controls: Located at the bottom of the table, showing:
  • Rows per page selector (default: 10)
  • Page navigation buttons (Previous, Next, First, Last)
  • Current page indicator
  • Total number of pages

Request Data Details

Request Types

Various request types may appear in the table:

• System Prompts: Requests beginning with "system:" followed by instructions
• Feature Queries: Requests about features of a product or system
• Getting Started Queries: Requests about how to begin using a product
• Product Inquiries: Requests about specific products or services

Response Patterns

Several response patterns may be observed:

• Information Responses: Responses providing features or getting started information
• Out-of-Scope Responses: Responses indicating the topic is beyond the system's knowledge
• Educational Responses: Responses about educational topics
• Product Information: Responses about products and services

Token Consumption

Token usage typically varies across requests based on:

• Complexity of the query
• Length of the response
• Model type used
• Conversation context

Performance Metrics

The table shows various performance indicators:

• Latency: Response time in seconds (can range from milliseconds to several seconds)
• Cost: Monetary cost of each request based on token usage and model type
• Success Rate: Indicated by the status field in request details

Using the Requests Page Effectively

Filtering and Searching

• Use the time range selectors (24H, 7D, 1M, 3M, All) to focus on specific time periods
• Use the search functionality to find specific requests by content, response, or other criteria
• Sort columns to identify patterns, such as:
  • Highest/lowest cost requests
  • Requests with longest/shortest latency
  • Most/least token-intensive requests

Analyzing Request Details

To inspect a specific request in detail:

1. Click on any row in the request table
2. Review the detailed information in the modal that appears
3. Examine the conversation flow between user and assistant
4. Check technical details like embedding model, completion model, and token count
5. Note the status and performance metrics

Data Analysis

• Cost Analysis:

  • Sort by the "Cost" column to identify the most expensive requests
  • Compare cost with token usage to understand cost efficiency
  • Identify request patterns that lead to higher costs

• Performance Analysis:

  • Sort by "Latency" to identify slow-responding requests
  • Correlate token usage with latency to understand performance impacts
  • Look for patterns in request types that lead to higher latency

• Usage Analysis:

  • Review request content to understand common user queries
  • Analyze response patterns to identify areas for improvement
  • Track token usage across different request types

• User Behavior Analysis:

  • Filter by specific users to analyze their usage patterns
  • Check feedback and reaction columns to gauge user satisfaction
  • Identify common queries from specific users or user groups

Exporting Data

Use the "EXPORT ALL DATA" button to:

• Generate reports for billing and cost analysis
• Perform detailed offline analysis of request patterns
• Create backups of request history for record-keeping

Best Practices for Request Monitoring

Regular Monitoring

• Check the Requests page daily for unusual patterns or issues
• Review high-cost or high-latency requests to identify optimization opportunities
• Monitor response quality for error patterns that may indicate knowledge gaps

Performance Optimization

• Identify request patterns with high token usage and consider reformulating prompts
• Look for requests with high latency and investigate potential causes
• Compare different model performances for similar request types

Cost Management

• Regularly review cost metrics to ensure budget alignment
• Identify and optimize high-cost queries
• Monitor token usage efficiency (input vs. output token ratio)

User Experience Improvement

• Pay attention to requests with "No Feedback" or negative reactions
• Analyze patterns in requests that receive error responses
• Use insights to improve prompt engineering and model selection

Troubleshooting Common Issues

Issue	Possible Cause	Solution
High latency	Complex queries, system load	Optimize prompt design, check system resources
Knowledge gap responses	Information limitations, unclear queries	Refine prompts, update knowledge base
Excessive token usage	Verbose prompts, inefficient queries	Streamline prompts, implement token limits
Cost spikes	High-volume usage, inefficient prompts	Review usage patterns, optimize prompts
Missing data	Filtering issues, processing delays	Adjust filters, refresh the page

Integration with Other Monitoring Features

The Requests page is part of a broader monitoring ecosystem that includes:

• Dashboard: Provides overview metrics of platform performance
• Audit Trail: Tracks all system changes and administrative actions
• User Analytics: Analyzes token usage and costs at the user level
• Platform Monitoring: Monitors technical aspects of the platform infrastructure

These components work together to provide comprehensive visibility into all aspects of platform usage and performance.