Configuring Products

Initialize and configure the new Prisme.ai product stack in a self-hosted environment

Use this guide after the core platform is deployed and reachable. It covers the product setup sequence for the new Prisme.ai stack: Governe, LLM Gateway, Storage, Agent Creator, Insights, Builder, and helper agents.

Legacy Knowledges and AI Store settings are only relevant when migrating an existing installation. For migration-specific steps, use Migration v27.

Configuration flow

1. Verify the platform

   Confirm that core and app services are healthy before importing product workspaces.

2. Import product workspaces

   Use the Platform workspace bulk import to install the required workspace groups in order.

3. Configure Governe

   Set the administration token, organization defaults, menu, roles, and access rules.

4. Configure LLM Gateway

   Add providers, secrets, models, default models, and model routing rules.

5. Configure Storage

   Connect the vector store and file storage services used by knowledge, files, and agent capabilities.

6. Enable products

   Validate Agent Creator, Insights, Builder apps, and helper agents.

Product stack

Product	What you configure	Where to configure it
Governe	Organizations, roles, SSO access, model governance, menu, appearance, observability.	Govern app and Governe workspace secrets.
LLM Gateway	Provider credentials, model catalog, default models, routing, quotas, and provider tests.	Govern > Models and LLM Gateway workspace secrets.
Storage	Vector store driver, credentials, index prefix, file and vector store connectivity.	Govern > Infrastructure and Storage workspace secrets.
Agent Creator	Agent creation, capabilities, templates, and published agent access.	Agent Creator product and Governe roles.
Insights	Conversation analytics, evaluation access, and monitoring dashboards.	Insights product and Governe roles.
Builder	Workspaces, apps, automations, templates, custom code, and product helper wiring.	Builder product and workspace settings.
Helper Agents	Form-assistance agents used by products.	Helper Agents workspace and Governe product configuration.

Prerequisites

Before configuring products, make sure the base deployment is ready:

• You can sign in as a super admin.
• The Platform workspace is visible in Builder.
• Core services are running: prismeai-api-gateway, prismeai-console, prismeai-pages, prismeai-workspaces, prismeai-events, and prismeai-runtime.
• App services required by product workspaces are running when used: prismeai-functions, prismeai-crawler, and prismeai-searchengine.
• Object storage, uploads storage, workspace storage, and databases are configured.
• Elasticsearch or OpenSearch is available if you use vector stores.
• LLM provider credentials are available as platform secrets or environment variables.

Readiness check

Run the platform readiness endpoint before importing or configuring products:

curl -H "Authorization: Bearer YOUR_TOKEN" https://api.your-domain.com/v2/readiness

You can also open the readiness URL from a browser already authenticated as a super admin.

All required services should return a ready status. Resolve readiness errors before continuing.

Offline or private-network deployments

For restricted environments, check these points before product initialization:

• prismeai-functions needs access to an npm registry when Custom Code dependencies are installed. If the public npm registry is not reachable, set NPM_CONFIG_REGISTRY to your internal registry.
• If internal or self-signed certificates are used, mount the CA bundle in services that call internal HTTPS endpoints.
• Python services such as prismeai-crawler and prismeai-searchengine use REQUESTS_CA_BUNDLE.
• Node.js services use NODE_EXTRA_CA_CERTS.
• LLM providers, vector stores, and crawler targets must be reachable from the relevant pods or containers.

Runtime app endpoints

Some product workspaces call app microservices through the runtime. With the standard Helm charts, these values are normally initialized for you.

If you run a custom deployment, verify the runtime configuration:

APP_CONFIG_CustomCode_apiUrl=http://FUNCTIONS_HOST:80
APP_CONFIG_Crawler_apiUrl=http://SEARCHENGINE_HOST:80

FUNCTIONS_HOST must be reachable from prismeai-runtime and should point to prismeai-functions.

SEARCHENGINE_HOST must be reachable from prismeai-runtime and should point to prismeai-searchengine.

If you use Collections from Builder apps, also verify the collection storage configuration:

COLLECTIONS_STORAGE_DRIVER=mongodb
COLLECTIONS_STORAGE_HOST=mongodb://MONGO_HOST/collections

For Helm deployments, prefer the chart values for runtime storage and secrets instead of hardcoding these values in manifests.

Import product workspaces

Product workspaces are imported from the Platform workspace. Only super admins can access this workspace.

Import the groups in this order:

Order	Group	Contains
1	base1	Foundation apps such as Custom Code and Prisme.ai API.
2	base2	Extended apps such as Crawler, NLU, and search integrations.
3	extended	Legacy workspaces kept for compatibility or migration.
4	one-product	New product workspaces such as Governe, LLM Gateway, Storage, and Agent Creator.

Always wait for the workspaces.bulkImport.completed event with no errors before importing the next group.

Import from the UI

1. Open Builder as a super admin.
2. Open the Platform workspace.
3. Go to Settings -> Versions -> Platform Pull.
4. Select the release repository.
5. Select the group to import.
6. Start the import and monitor the Activity feed.
7. Repeat for the next group only after the completion event appears.

Import from the API

You can trigger the same import through the API:

curl -X POST "${API_URL}/v2/workspaces/platform/versions/latest/pull" \
  -H "Authorization: Bearer ${ACCESS_TOKEN}" \
  -H "Content-Type: application/json" \
  -d '{"repository": {"id": "release"}, "groups": ["base1"]}'

Repeat with base2, extended, then one-product.

Re-importing a group can overwrite workspace changes made in Builder since the previous import. Use it intentionally.

Configure Governe first

Governe is the control plane for organizations, roles, product access, models, and platform appearance. Configure it before opening products to end users.

Administration token

1. Sign in as a super admin.
2. Generate a long-lived access token for the Governe workspace.
3. Open the Governe workspace in Builder.
4. Open Settings -> Secrets.
5. Store the token in the adminAccessToken secret.

See Configuring Governe for the detailed token command and workspace setup.

Organization baseline

In Govern, configure:

• The organization name and technical identifier.
• Join rules for users who authenticate through local auth or SSO.
• Default organization roles.
• Groups used for product access.
• Appearance, legal links, and menu entries.

For end-user navigation, import or recreate the platform menu so users can reach Agent Creator, Governe, Insights, Builder, and other enabled products.

Configure LLM Gateway

LLM Gateway centralizes model providers and exposes a consistent model catalog to products.

Provider secrets

Store provider credentials either:

• In environment variables using the WORKSPACE_SECRET_llm-gateway_* prefix.
• Directly in the LLM Gateway workspace under Settings -> Secrets.

Examples:

WORKSPACE_SECRET_llm-gateway_openaiApiKey=...
WORKSPACE_SECRET_llm-gateway_awsBedrockAccessKey=...
WORKSPACE_SECRET_llm-gateway_awsBedrockSecretAccessKey=...

Providers and models

1. Open Govern.
2. Go to Models -> Providers.
3. Add each provider and reference the matching secret names.
4. Save the provider configuration.
5. Switch to Models.
6. Add the completion, embedding, multimodal, and tool-capable models you want to expose.
7. Test each model from the UI.

For embedding models, set dimensions explicitly when the provider requires it.

Organization model policy

After the model catalog is valid:

1. Open the target organization.
2. Go to Agent controls.
3. Select the allowed models.
4. Choose default completion and embedding models.
5. Save.

Products will only expose models allowed by the organization policy.

Configure Storage

Storage handles files, vector stores, and knowledge-backed retrieval for the new product stack.

1. Open Govern.
2. Go to Infrastructure.
3. Select the vector store driver: Elasticsearch or OpenSearch.
4. Reference the credential secrets.
5. Set the vector index prefix.
6. Save and click Test.

Store vector credentials either:

• In environment variables using the WORKSPACE_SECRET_storage_* prefix.
• Directly in the Storage workspace under Settings -> Secrets.

Examples:

WORKSPACE_SECRET_storage_opensearch_host=...
WORKSPACE_SECRET_storage_opensearch_username=...
WORKSPACE_SECRET_storage_opensearch_password=...

Use a clear index prefix per environment, such as prod_rag, preprod_rag, or dev_rag.

Enable product access

Once Governance, LLM Gateway, and Storage are configured, validate user-facing products.

Agent Creator

Use Agent Creator for creating, testing, and publishing agents.

1. Open Agent Creator.
2. Create a small test agent.
3. Select an allowed model.
4. Add a simple instruction.
5. Test it in the playground.
6. Confirm the agent appears only for users or groups that should access it.

Product usage docs: Agent Creator.

Insights

Use Insights after conversations start flowing through agents.

1. Open Insights.
2. Confirm dashboards load.
3. Check that access is restricted to admin, analytics, or governance users.
4. Verify that conversation and evaluation data appears for test agents.

Configuration details: Configuring Insights.

Builder

Use Builder to configure product workspaces, install apps, run automations, and maintain custom workflows.

Validate:

• The Platform workspace can import release groups.
• Custom Code can run a simple function.
• Crawler and search apps work if website ingestion is enabled.
• Product workspaces are visible only to the right administrators.

Product usage docs: Builder.

Helper agents

Helper agents assist users while they fill forms in Prisme.ai products.

1. Configure HELPER_AGENT_ENDPOINT.
2. Open the Helper Agents workspace.
3. Run the initialization automations.
4. Configure generated helper credentials in Governe.
5. Test helper availability from a product form.

Configuration details: Configuring Helper Agents.

Validation checklist

Before handing the platform to users, verify:

• Readiness API returns ready for required services.
• All product import groups completed without errors.
• Governe has a valid adminAccessToken.
• Organization join rules assign the expected roles and groups.
• Menu entries point to the enabled products.
• At least one completion model and one embedding model are tested and allowed.
• Storage infrastructure test passes.
• Agent Creator can create and run a test agent.
• Insights access is restricted and dashboards load.
• Custom Code and Crawler tests pass if those apps are enabled.

Troubleshooting

A product is missing from the menu

Check the Governe menu editor and organization access rules. Also confirm the product workspace was imported by the one-product group.

Models do not appear in Agent Creator

Confirm that the model exists in Govern > Models, the provider is saved, the model test succeeds, and the model is allowed for the user’s organization.

Model tests fail

Check provider secret names, network egress to the model provider, provider API versions, deployment names, and rate limits.

Storage test fails

Check WORKSPACE_SECRET_storage_* secret names, vector store host reachability, credentials, TLS settings, and index permissions.

Bulk import reports errors

Open the workspaces.bulkImport.completed event in the Platform workspace Activity feed. Check the errors payload, resolve missing dependencies or slug conflicts, then re-import the failed group.

Custom Code apps fail after import

Check prismeai-functions logs for dependency installation errors. If dependencies are still installing or missing, re-save the affected Custom Code app to restart installation.

Next steps

Configure Governe

Set up organizations, roles, model governance, and platform access.

Model governance

Manage providers, models, defaults, and routing behavior.

Agent Creator

Create and test agents once models and storage are configured.

Install from scratch (v27)

Fresh-install walkthrough — Helm values, bulk import, providers, models, vector store.

Migration v27

Migrate legacy Knowledges and AI Store settings to the new stack.