System AI Tab

Use this tab to enter valid Azure OpenAI and Ollama credential information in PPM Administrator to be able to use the AI features in PPM applications, such as PM Compass and Cobra.

Note: This tab is available by default only to the SYSADMIN group. Access to this tab is controlled on the Groups Security tab.

Field	Description
AI Services Provider	It specifies the AI Service Provider to be used. You can select one of the following options: None Azure OpenAI Note: Take note that there is a token limit per minute when interacting with Azure OpenAI, depending on your company’s setup. If the number of users accessing the Narrative Score feature (specifically, clicking the Generate Narrative Score button) at the same time exceeds the assigned token limit per minute, PM Compass displays an error message. If you are using Azure OpenAI, Deltek recommends reviewing the AI model deployment in your subscription and adjusting the token usage/limit accordingly. If needed, submit a quota increase request to Microsoft. You can access the Request for Quota Increase form on the Microsoft Azure website. Ollama Note: Ollama is an open-source AI platform that lets you deploy large language models directly on your own servers or your private cloud. Unlike cloud-based AI services, Ollama keeps all your data within your organization— nothing is sent to external providers. You can choose from a variety of language models, including open-source options like Llama 3 and Mistral, proprietary models, or even custom-tuned models tailored to your specific business needs. This Bring Your Own Large Language Model (BYO LLM) approach gives you the flexibility to select the AI model that works best for your organization while maintaining complete control over privacy, security, and customization. Attention: While you can configure Ollama connections in PPM Administrator, you cannot use Ollama functionality until your specific PPM product (such as Acumen, Cobra, or PM Compass) releases its Ollama feature support. Actual Ollama availability depends on each product's release schedule and capabilities. For further information regarding the availability of Ollama within your product, see the relevant product's release documentation.
Endpoint URL	Use this field to enter the endpoint URL for your selected AI service provider. Enter a valid URL that does not exceed 900 characters. This field is enabled and required when you select Azure OpenAI or Ollama in the AI Service Provider field.
API Key	Use this field to enter the API key used to authenticate requests to the Azure OpenAI service provider. Enter a value that does not exceed 900 characters. This field is enabled and required when you select Azure OpenAI in the AI Service Provider field. It is disabled when you select None or Ollama in the same field.
Completion	Use this section to configure the AI models used by the application for generating responses based on your selected AI Service Provider. Deployment: Use this field to enter the name of the deployment you want to use when using the Azure OpenAI service provider. Enter a value between 2-64 characters. For the selected deployment, make sure that the Version Update Policy in Azure OpenAI is set to Once a new default version is available. This will ensure a consistent connection with PM Compass and Azure OpenAI. This field is enabled and required when you select Azure OpenAI in the AI Service Provider field. It is disabled when you select None or Ollama in the same field. Model: Use this field to enter the name of the model you want to deploy when using the Ollama service provider. Enter a value between 2-64 characters. This field is enabled and required when you select Ollama in the AI Service Provider field. This field is disabled when you select None or Azure OpenAI in the same field. Test Connection: Click this button to run the validation and test the AI configuration. If you enter a valid value in any of the fields, you can test the connection and save the changes. A successful test connection is required before saving.
Embeddings	Use this section to configure the AI models used by Ask Dela for generating responses and processing embeddings, based on your selected AI Service Provider. Deployment: Use this field to enter the name of the deployed Large Language Model (LLM) instance when using the Azure OpenAI service provider. Enter a value between 2-64 characters. This field is enabled and required when you select Azure OpenAI in the AI Service Provider field. It is disabled when you select None or Ollama in the same field. Model: Use this field to enter the name of the model used to generate text embeddings (for example, text-embedding-3-small)you want to deploy when using the Ollama service provider. Enter a value between 2-64 characters. This setting determines how text is converted into vectors and directly impacts semantic accuracy, search relevance, and retrieval quality. This field is enabled and required when you select Ollama in the AI Service Provider field. This field is disabled when you select None or Azure OpenAI in the same field. Test Connection: Click this button to run the validation and test the AI configuration. If you enter a valid value in any of the fields, you can test the connection and save the changes. A successful test connection is required before saving.
Maximum Concurrent Prompt Requests	This field sets the limit on the concurrent client requests sent to the provider. You can enter a value between 0 and 65,535. The default is set to 0, disabling concurrency limits and allowing simultaneous processing of all requests. This field is available when you select Azure OpenAI or Ollama in the AI Service Provider field. This field is disabled when you select None in the same field.
Custom HTTP Request Headers	This grid enables you to define custom HTTP headers that will be included in requests sent to the AI service provider. Enter a header name and value so the system automatically includes them in your request. This grid is available when you select Azure OpenAI or Ollama in the AI Service Provider field. This grid is disabled when you select None in the same field. For keys, enter keys that are valid HTTP headers based on RFC 7230 with at least 2 characters and a maximum of 900 characters. Values are optional. If a value is defined, enter values that are valid HTTP headers based on RFC 7230 with up to 4,096 characters. There is no limit on the number of rows on the grid. There is a limit of 16,384 characters to the entire HTTP headers. PPM Administrator stores the data in the AI_CUSTOM_REQUEST_HEADERS system preference.

To access this tab, see Display the System AI Tab.