Skip to main content
Before initializing Prisme.ai AI Products, you need to ensure that your environment meets all necessary prerequisites and configurations. This guide will walk you through the required services, configuration steps, and verification processes.

​
Core Requirements

A fully operational Prisme.ai environment is required, including both open-source and private microservices.
  • Core Services
  • Apps Services
The following microservices must be running:
  • prismeai-api-gateway
  • prismeai-console
  • prismeai-pages
  • prismeai-workspaces
  • prismeai-events
  • prismeai-runtime
For offline environments, additional prerequisites may be required. Please refer to the offline environment setup guide.

​
Network Requirements

All microservices should be able to communicate with each other within your environment. Ensure that:
  • The runtime service can reach all other services (functions, crawler, searchengine, etc.)
  • The functions service can reach external package repositories for npm/pip dependencies
  • The console service is accessible to administrators

​
Hardware Requirements

Minimum Requirements

  • 8 CPU cores
  • 16GB RAM
  • 100GB storage
  • Suitable for testing and small-scale deployments

Recommended Requirements

  • 16+ CPU cores
  • 64GB+ RAM
  • 250GB+ storage
  • Suitable for production deployments
Hardware requirements may vary depending on your specific usage patterns, the number of concurrent users, and the amount of data processed.

​
Storage requirements

We recommend using:
  • MongoDB: For applications collections
  • Vector Database: one of Opensearch (recomended) or Redis Stack (with JSON and SEARCH modules)
Ensure these databases are properly configured and accessible to the prismeai-functions microservice.
prismeai-functions and prismeai-llm microservices require a RWX volume with at least 50GB and attached to a Kubernetes PVC.

​
Offline environments

If the installation environment does not have access to internet, ensure that the following prerequisites are always met :
  • If using app microservices, prismeai-functions need access to a NPM registry. If the default https://registry.npmjs.org/ is not available, you must configure your own npm registry using NPM_CONFIG_REGISTRY environment variable inside prismeai-functions microservice
    • Depending on your NPM proxy, it might require additional configuration for authentication
  • If using internal / self-signed TLS certificates, they must be configured inside almost all microservices to avoid HTTPS errors
    • prismeai-searchengine and prismeai-crawler are Python microservices : once mounted, the certificate file location can be configured with REQUESTS_CA_BUNDLE prismeai-llm does not need the custom certificate
    • Every other prismeai-* services are NodeJS : once mounted, the certificate file location can be configured with NODE_EXTRA_CA_CERTS

​
Configuration Requirements

Several environment variables must be configured in the prismeai-runtime microservice to ensure proper functioning of installed Apps. If your deployment uses custom API ports modified by a container orchestrator or configuration, you’ll need to adapt these values accordingly.

​
Environment Variables

APP_CONFIG_CustomCode_apiUrl='http://FUNCTIONS_HOST:80'
Where FUNCTIONS_HOST refers to the domain from which the prismeai-runtime microservice can reach prismeai-functions.
WORKSPACE_CONFIG_collection_mongoURI='mongodb://MONGO_HOST/collections'
Where MONGO_HOST should be replaced with the host that allows the prismeai-functions microservice to connect to MongoDB (e.g., mongo if prismeai-functions and mongo are launched from the same docker-compose).
APP_CONFIG_Crawler_apiUrl='http://SEARCHENGINE_HOST:80'
Where SEARCHENGINE_HOST is the domain from which prismeai-runtime can reach prismeai-searchengine.
APP_CONFIG_NLU_apiUrl='http://NLU_HOST:80'
Where NLU_HOST is the domain from which prismeai-runtime can reach prismeai-nlu.
The NLU App is not required for the operation of AI Products, but provides advanced intent recognition capabilities if needed.
APP_CONFIG_SendMail_apiKey='a mailgun API key'
APP_CONFIG_SendMail_region='https://api.eu.mailgun.net/'
APP_CONFIG_SendMail_domain='the mailgun domain'
The SendMail App is currently compatible only with MailGun. It remains optional and is not required for the proper functioning of the AI products.

​
AI Products Initialization Process

Once you’ve confirmed that all prerequisites have been met, you can proceed with importing the Prisme.ai App Store and associated product workspaces. The initialization process involves importing archives in the correct sequence and verifying functionality after each step.
It’s recommended to perform archive imports using a studio account dedicated to this task to avoid cluttering the dashboard of an administrator account with the multiple app workspaces.

​
Import Process Overview

The initialization process follows a specific sequence of imports, with testing performed after each step:
1

Import Package 1: Base Apps

Import the first archive containing foundational apps required by other components.
2

Test Base Apps

Verify the functionality of apps imported in the first step.
3

Import Package 2: Extended Base

Import the second archive with additional core services.
4

Test Extended Base Apps

Verify the functionality of apps imported in the second step.
5

Import Package 3: AI Products

Import the main AI product workspaces.
6

Configure AI Products

Complete final configuration steps for AI products.

​
Importing Archives

To import an archive:
  1. Navigate to the platform
  2. Hover over the “Create a workspace” button
  3. Click on the three-dot menu
  4. Select “Import a workspace from an archive” from the dropdown
  5. Select the archive file
  6. Wait for the import to complete
    1. Namely, the workspaces pod will issue a log Terminated bulk import
    2. The log should have an internal field errors: []
  7. Refresh the page to view the newly imported workspaces

​
Package 1: apps-base1.zip

This archive contains foundational apps required by many other components.

​
Key Components

Custom Code

Enables execution of custom JS/Python codeRequires the functions microservice to be online and configured

Prisme.ai API

Enables API calls to Prisme.ai from Automations (YAML files)

​
Verification Tests

After importing the first package, run these tests to verify functionality:
  1. Create a test workspace
  2. Install the Custom Code App in your test workspace
  3. Create a new function named hello with the following code:
import fetch from 'node-fetch';
const res = await fetch('https://httpbin.org/headers')
return await res.json();
  1. Test the function using the green play button
  2. Verify that the node-fetch package has been properly installed and that you receive the expected website response without any import or internal errors

​
Package 2: apps-base2.zip

This archive contains extended base applications that build upon the first package.

​
Key Components

Collection

MongoDB integrationRequires the Custom Code App to work correctly

Crawler

Web page indexing & search engineRequires the crawler and searchengine microservices

NLU

NLU engineRequires the nlu microservice

RedisSearch

Interface with the redis-stack-server database

​
Verification Tests

After importing the second package, run these tests to verify functionality:
  1. Open the PlatformTests workspace that was created during import
  2. Navigate to the Collection/TestCollection automation
  3. Run the automation from the top-right corner play button
  4. Verify that you receive a valid JSON response including a “Toulouse” city object
  1. Open the Crawler/Crawl page automation and run it with the top-right corner play button
  2. Open the Activity tab and regularly reload it until you see Crawler.notifications event has been fired
    • If this never occurs, check prismeai-crawler container logs (apps namespace) to resolve issues
  3. Open the Crawler/Search page automation and run it
  4. Verify that you receive the freshly indexed page result

​
Package 3: apps-extended.zip

This final archive contains the main Prisme.ai AI products and is larger than the previous packages. The import is complete when you see a log similar to: 
{
  "level":30,
  "time":1691420134138,
  "app":"core-prismeai-workspaces",
  "msg":"Terminated bulk import",
  "result":{
    "imported":[...],
    "errors":[...],
    "createdWorkspaceIds":[...],
    "updatedWorkspaceIds":[],
    "publishedApps":[...]
  }
}
Several new workspaces should appear when you refresh the page after seeing this log.

​
Post-Import Configuration

After successfully importing all three packages, you need to proceed with the configuration of each AI Product to complete the setup process. It is advised to configure the products in the following minimum order :
  • Configure AI Governance
    • 1st, Generate a long-term token
    • Configure the adminAccessToken setting
    • Configure the Ops Manager and upgrade the core helm deployment
  • Configure super admins
    • The AI Governance shortcut has appeared, elect a PlatformAdmin and PlatformManager
  • Configure LLM providers
    • In the AI Builder > AI Knowledge > Settings section
  • Create an AI SecureChat agent
    • Finally, add this agent as a left-menu shortcut in the AI Governance > Products and Workspaces > Shortcuts section

​
Troubleshooting

If you encounter issues during the initialization process, here are some common problems and their solutions:
Several critical AI products and Apps (including AI Knowledge) are built on Custom Code. If issues arise:
  1. Monitor the logs of the prismeai-functions service
  2. Look for these error messages:
    • Failed npm install
    • Dependencies are still being installed from
    • Some dependencies are missing
    • TimeoutError
    • Failed running function
  3. If a dependency is missing or still being installed after several minutes, re-save the Custom Code App for the affected workspace from the studio. This will prompt prismeai-functions to restart the installation (Installing N modules...).
  4. If the issue concerns functions of an App (indicated by the app. prefix in the logs), you’ll need to re-publish the App.
If the import process times out or seems stuck:
  1. Check the logs of the prismeai-workspaces service to see if the import is still in progress
  2. Wait for the completion message (Terminated bulk import) even if the UI appears frozen
  3. Refresh the page after a few minutes to see if new workspaces have appeared
  4. If the import failed, try importing again or check for network issues
If tests fail due to connectivity issues:
  1. Ensure all required microservices are running
  2. Verify that the environment variables are correctly set
  3. Check network connectivity between services
  4. Look at the logs of the specific services mentioned in the error messages

​
Next Steps

Create your SecureChat

Create your AI Secure Chat

AI Store

Learn how to manage and deploy agents in the Store

AI Knowledge

Organize collections and document-based agents

Create an agent

Learn more about agent creation