Environment setup

You must first verify that your system and workspace meet the necessary standards, and then perform the required installation and configuration steps to successfully run a Sample-Based Learning project.

This page contains the following sections:

Sample-Based Learning system requirements

Before you can begin to use Sample-Based Learning you must have the following:

Adding the Sample-Based Learning application

To install Assisted Review in your workspace, import the Assisted Review application from the application library. To do this, you must have the appropriate admin rights.

To install the Assisted Review application in your workspace:

  1. Navigate to the Relativity Applications tab.
  2. Click New Relativity Application.
  3. Select Select from Application Library.
  4. Click ellipsis button on Choose from Application Library.
  5. Select Assisted Review, and click OK.
  6. Click Import.
  7. Once the import completes, Relativity creates an Assisted Review tab, from which you can use Assisted Review in your workspace. You may need to refresh your browser to see this new tab.

Agent configuration

Sample-Based Learning functionality depends on two types of agents:

  • Assisted Review Worker - does the heavy lifting. The more worker agents there are, the faster some Assisted Review processes are. However, there is a limit to how many are useful, and using too many may cause unnecessary load on some Relativity setups.
  • Assisted Review Manager - does some organization and initialization work, specifically overseeing distribution of processes to the worker agents. This agent works on all projects regardless of Resource Group. It does no intensive work. You can have only one manager agent per environment hosting Assisted Review, as opposed to one manager agent per agent server. If you install more than one manager agent in the environment, Assisted Review fails to operate in Relativity.

Note the following about RAR agents:

  • One worker and one manager agent are installed automatically during the initial installation of Assisted Review.
  • Never put RAR agents on a web server.

    Note: You must install the Assisted Review Manager on an agent server in your workspace's resource pool.

  • Always disable your RAR agents when upgrading Relativity, or else you will encounter issues with your RAR project(s) in your new Relativity environment.

Agent recommendations

The following guidelines are a good place to begin when setting up your Relativity environment for Assisted Review:

Environment Size # of Worker Agents Comments
Tier 1 (Single Server Deployments) 2 More than two Assisted Review agents may introduce system wide performance and stability related issues.
Tier 1 - Tier 2 < 10 Dependent on available resources.
Tier 3 10 - 30 When running many Assisted Review projects on many workspaces, going past ten agents is beneficial. Work with Client Services to determine your needs. If you have more than ten Worker agents, you can raise the run interval of the additional agents from .5 seconds to 5 seconds.

See Tier level definitions for more information.

Note the following about categorization in Sample-Based Learning:

  • The most resource-intensive Sample-Based Learning processes are categorization and saving categorization results.
  • Categorization does not utilize Assisted Review agents; it uses the Analytics Categorization Manager.
  • Assisted Review requires the use of an Analytics Categorization Manager. There should be no more than two Analytics Categorization Manager Agents per Relativity resource pool. See Agents for more information regarding the Analytics Categorization Manager.

In consultation with Relativity Client Services, you may want to consider adding worker agents if:

  • You have many more Assisted Review projects than you have Assisted Review Worker agents.
  • You commonly save the categorization results of multiple projects at the same time.
  • The saving of categorization results takes a long time.

Relevant instance setting table values

RAR uses the following instance setting table values to retrieve data:

The following EDDS instance setting table values control Assisted Review error retries:

For more information, see Instance setting values.

Manually adding worker agents

Depending on your environment setup, adding more worker agents may increase performance of Assisted Review. To add worker agents:

  1. Click your name in the upper right corner of Relativity and click Home.
  2. Navigate to the Agents tab.
  3. Click New Agent.
  4. Complete the fields on the agent form. See Fields for more information.
  5. Click Save. You receive a confirmation that the agent was successfully added.


The agent form contains the following fields:

  • Agent Type - the type of agent you're adding. Click ellipsis button to display a list of available agents. Select Assisted Review Worker and click OK.
  • Number of Agents - the number of agents you want to add. Environments hosting Assisted Review should have multiple worker agents installed to avoid performance issues when executing a project. Even if only a single project is running in the environment, you should have at least two worker agents installed.
  • Agent Server - the server where you want to add the agent(s). Click ellipsis button to display a list of available servers. Select the server and click OK.
  • Run Interval - the number of seconds between each time the agent checks for available jobs. When you select Assisted Review Worker as the agent type, this value automatically defaults to 0.5. Changing this value impacts performance.
  • Logging level of event details - determines what events the agent logs. For Assisted Review, it is recommended that you keep this at the default value of Log critical errors only.
  • Enabled - determines whether the agent is enabled once you add it. For Assisted Review, it is recommended that you keep this at the default value of Yes.

See Agents for more information.