AI Agent - Revision history

Ollvihe: /* Take AI Agent into Use */

2025-12-03T14:32:39Z

Take AI Agent into Use

← Older revision		Revision as of 14:32, 3 December 2025
Line 6:		Line 6:

	== Take AI Agent into Use ==		== Take AI Agent into Use ==
	To use the AI Agent, Snowflake permissions may need to be set, so that the Snowflake language models are available for the AI Agent. By default, all those language models can be used that are available in the same region. The permission is given to the public role in Snowflake, so make sure the Snowflake user has the public role or a custom role that has the needed permissions (more information: https://docs.snowflake.com~~/en~~/user-guide/snowflake-cortex/~~llm~~-~~functions#required~~-~~privileges~~). In addition, it's possible to restrict which language models can be used (nore information: https://docs.snowflake.com/en/user-guide/snowflake-cortex/llm-functions#control-model-access). To use Snowflake language models that are available in other Snowflake regions, the Cross-region Inference feature can be used (more information: https://docs.snowflake.com/en/user-guide/snowflake-cortex/cross-region-inference).		To use the AI Agent, Snowflake permissions may need to be set, so that the Snowflake language models are available for the AI Agent. By default, all those language models can be used that are available in the same region. The permission is given to the public role in Snowflake, so make sure the Snowflake user has the public role or a custom role that has the needed permissions (more information: https://docs.snowflake.com/user-guide/snowflake-cortex/aisql#cortex-user-database-role). In addition, it's possible to restrict which language models can be used (nore information: https://docs.snowflake.com/en/user-guide/snowflake-cortex/llm-functions#control-model-access). To use Snowflake language models that are available in other Snowflake regions, the Cross-region Inference feature can be used (more information: https://docs.snowflake.com/en/user-guide/snowflake-cortex/cross-region-inference).

	Optionally, the OpenAI language models can be used by the AI Agent if the OpenAI connection is configured. To use OpenAI default model, leave the LLM name field empty. See [[Snowflake_Connection_Configuration#Snowflake_Cortex\|Snowflake Cortex]] or [[AI_Assistant_for_QPR_ProcessAnalyzer#Take_AI_Assistant_into_Use\|OpenAI related settings]] respectively.		Optionally, the OpenAI language models can be used by the AI Agent if the OpenAI connection is configured. To use OpenAI default model, leave the LLM name field empty. See [[Snowflake_Connection_Configuration#Snowflake_Cortex\|Snowflake Cortex]] or [[AI_Assistant_for_QPR_ProcessAnalyzer#Take_AI_Assistant_into_Use\|OpenAI related settings]] respectively.

TeeHiet: TK-61376

2025-10-08T07:53:49Z

TK-61376

← Older revision		Revision as of 07:53, 8 October 2025
Line 13:		Line 13:

	== General Settings ==		== General Settings ==
			Snowflake Cortex language model queries are run in the Snowflake connection where the QPR ProcessAnalyzer model selected for the AI Agent is located.
	* '''System Prompt''' (optional): Enter a system prompt for the language model if desired. This can help guide the model's responses but is not required. The system prompt can be used give general instructions how the LLM should behave when responding to the user.		* '''System Prompt''' (optional): Enter a system prompt for the language model if desired. This can help guide the model's responses but is not required. The system prompt can be used give general instructions how the LLM should behave when responding to the user.
	* '''User Prompt''': Provide a user prompt for the language model. This field must contain valid input. Usually, the user prompt is used give the specific task or question the language model should respond to.		* '''User Prompt''': Provide a user prompt for the language model. This field must contain valid input. Usually, the user prompt is used give the specific task or question the language model should respond to.
Line 20:		Line 21:
	** Reordering datasets: You can change the order of datasets by using the mouse to drag and drop them into your preferred sequence.		** Reordering datasets: You can change the order of datasets by using the mouse to drag and drop them into your preferred sequence.
	** Deleting datasets: To remove a dataset, click the Delete icon button next to the dataset you wish to delete.		** Deleting datasets: To remove a dataset, click the Delete icon button next to the dataset you wish to delete.
			** Each prompt dataset can use a different model and the prompt dataset queries are run in the Snowflake connection where the selected model is located.
	* '''Use charts in dashboard as prompt''': Select this checkbox to have the datasets for the other charts in the dashboard to be provided as prompt for the AI Agent. Only the following visualizations are included (others are ignored): Table, Pivot table, KPI Card, Root causes, and charts.		* '''Use charts in dashboard as prompt''': Select this checkbox to have the datasets for the other charts in the dashboard to be provided as prompt for the AI Agent. Only the following visualizations are included (others are ignored): Table, Pivot table, KPI Card, Root causes, and charts.
	* '''LLM Name''': Select a language model from the predefined list of Snowflake Cortex models or enter a custom model name (for example your own fine-tuned model). If the OpenAI connection is available and no LLM name is specified, the OpenAI language model will be used by default.		* '''LLM Name''': Select a language model from the predefined list of Snowflake Cortex models or enter a custom model name (for example your own fine-tuned model). If the OpenAI connection is available and no LLM name is specified, the OpenAI language model will be used by default.

Ollvihe at 12:19, 12 September 2025

2025-09-12T12:19:18Z

← Older revision		Revision as of 12:19, 12 September 2025
Line 1:		Line 1:
	The AI Agent is an innovative dashboard component designed to display textual responses generated by language models based on user-configured prompts. This component seamlessly integrates with both case-centric and object-centric Snowflake models, to provide comprehensive insights.		The AI Agent is an innovative dashboard component designed to display textual responses generated by language models based on user-configured prompts. This component seamlessly integrates with both case-centric and object-centric Snowflake models, to provide comprehensive insights. Key features of the AI Agent include:
	~~<br>~~		* Versatile Model Support: The AI Agent is compatible with language models from both OpenAI and Snowflake Cortex, offering flexibility and a wide range of capabilities to meet diverse analytical needs.
	~~'''~~Key features of the AI Agent include~~'''~~:		* Dynamic Response Generation: By configuring specific prompts, users can tailor the AI Agent to generate relevant and context-specific textual responses, enhancing decision-making processes.
	* ~~'''~~Versatile Model Support~~'''~~: The AI Agent is compatible with language models from both OpenAI and Snowflake Cortex, offering flexibility and a wide range of capabilities to meet diverse analytical needs.		* Integration with Snowflake: The AI Agent supports Snowflake's case-centric and object-centric models, allowing users to harness the full potential of their data within the Snowflake ecosystem.
	* ~~'''~~Dynamic Response Generation~~'''~~: By configuring specific prompts, users can tailor the AI Agent to generate relevant and context-specific textual responses, enhancing decision-making processes.
	* ~~'''~~Integration with Snowflake~~'''~~: The AI Agent supports Snowflake's case-centric and object-centric models, allowing users to harness the full potential of their data within the Snowflake ecosystem.
	This component is ideal for users seeking to leverage advanced language models to extract meaningful insights and drive data-driven strategies directly from their dashboards.		This component is ideal for users seeking to leverage advanced language models to extract meaningful insights and drive data-driven strategies directly from their dashboards.

Ollvihe: /* Generic Requirements for AI Systems */

2025-06-26T21:35:11Z

Generic Requirements for AI Systems

← Older revision		Revision as of 21:35, 26 June 2025
Line 81:		Line 81:

	== Generic Requirements for AI Systems ==		== Generic Requirements for AI Systems ==
	This chapter describes how the AI Agent fulfills ~~common~~ generic requirements for AI-powered systems.		This chapter describes how the AI Agent fulfills some generic requirements for AI-powered systems.

	=== Human Supervision of AI-generated Actions ===		=== Human Supervision of AI-generated Actions ===

Ollvihe: /* Generic Requirements for AI Systems */

2025-06-26T21:34:42Z

Generic Requirements for AI Systems

← Older revision		Revision as of 21:34, 26 June 2025
Line 87:		Line 87:

	Furthermore, the interface is user-friendly, with clear and accessible controls that allow users to easily review and override AI-generated recommendations as necessary. This setup not only facilitates effective collaboration between the AI and its human operators but also ensures that human judgment and expertise remain central to the operational workflow.		Furthermore, the interface is user-friendly, with clear and accessible controls that allow users to easily review and override AI-generated recommendations as necessary. This setup not only facilitates effective collaboration between the AI and its human operators but also ensures that human judgment and expertise remain central to the operational workflow.

	~~=== AI Models Lifecycle Management ===~~
	QPR ProcessAnalyzer AI Agent is based on the foundational large language models (LLMs) from selected providers (such as Meta, Claude and Mistral AI). When a new AI Agent is configured, the foundational LLM is selected, the input instructions are written, and the source datasets and analyses used as prompt are selected. By default, the foundational LLMs are not subjected to further training during this process but only the context window of the request is used as input for the LLM.

	=== Output Labeling to Distinguish AI-generated Content ===		=== Output Labeling to Distinguish AI-generated Content ===
Line 96:		Line 93:
	=== Masking of Data Processed by AI ===		=== Masking of Data Processed by AI ===
	In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.		In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.

			=== AI Models Lifecycle Management ===
			QPR ProcessAnalyzer AI Agent is based on the foundational large language models (LLMs) from selected providers (such as Meta, Claude and Mistral AI). When a new AI Agent is configured, the foundational LLM is selected, the input instructions are written, and the source datasets and analyses used as prompt are selected. By default, the foundational LLMs are not subjected to further training during this process but only the context window of the request is used as input for the LLM.

	[[Category: QPR ProcessAnalyzer]]		[[Category: QPR ProcessAnalyzer]]

Ollvihe: /* Masking for Data Processed by AI */

2025-06-26T21:34:18Z

Masking for Data Processed by AI

← Older revision		Revision as of 21:34, 26 June 2025
Line 94:		Line 94:
	QPR ProcessAnalyzer AI Agent functions as a separate, clearly identifiable component within the dashboards. This design makes it straightforward for users to distinguish between AI-generated content and human-generated content, fulfilling the requirement for transparency.		QPR ProcessAnalyzer AI Agent functions as a separate, clearly identifiable component within the dashboards. This design makes it straightforward for users to distinguish between AI-generated content and human-generated content, fulfilling the requirement for transparency.

	=== Masking ~~for~~ Data Processed by AI ===		=== Masking of Data Processed by AI ===
	In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.		In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.

	[[Category: QPR ProcessAnalyzer]]		[[Category: QPR ProcessAnalyzer]]

Ollvihe: /* Generic requirements for AI systems */

2025-06-26T21:34:10Z

Generic requirements for AI systems

← Older revision		Revision as of 21:34, 26 June 2025
Line 80:		Line 80:
	* If the AI Agent doesn't need to follow the dashboard filters, disable the "Follow Dashboard Filters" settings. When the settings is enabled, the AI Agent is rerun and the language model called every time filters are changed.		* If the AI Agent doesn't need to follow the dashboard filters, disable the "Follow Dashboard Filters" settings. When the settings is enabled, the AI Agent is rerun and the language model called every time filters are changed.

	== Generic ~~requirements~~ for AI ~~systems~~ ==		== Generic Requirements for AI Systems ==
	This chapter describes how the AI Agent fulfills common generic requirements for AI-powered systems.		This chapter describes how the AI Agent fulfills common generic requirements for AI-powered systems.

	=== Human ~~supervision~~ of AI-generated ~~actions~~ ===		=== Human Supervision of AI-generated Actions ===
	To ensure robust human oversight, the AI Agent has been designed with several key features. Firstly, all insights and recommendations generated by the AI are intended to support human decision-making rather than replacing it. The AI does not autonomously execute any actions, ensuring that human operators are always involved in the final decision-making process.		To ensure robust human oversight, the AI Agent has been designed with several key features. Firstly, all insights and recommendations generated by the AI are intended to support human decision-making rather than replacing it. The AI does not autonomously execute any actions, ensuring that human operators are always involved in the final decision-making process.

	Furthermore, the interface is user-friendly, with clear and accessible controls that allow users to easily review and override AI-generated recommendations as necessary. This setup not only facilitates effective collaboration between the AI and its human operators but also ensures that human judgment and expertise remain central to the operational workflow.		Furthermore, the interface is user-friendly, with clear and accessible controls that allow users to easily review and override AI-generated recommendations as necessary. This setup not only facilitates effective collaboration between the AI and its human operators but also ensures that human judgment and expertise remain central to the operational workflow.

	=== AI ~~models lifecycle management~~ ===		=== AI Models Lifecycle Management ===
	QPR ProcessAnalyzer AI Agent is based on the foundational large language models (LLMs) from selected providers (such as Meta, Claude and Mistral AI). When a new AI Agent is configured, the foundational LLM is selected, the input instructions are written, and the source datasets and analyses used as prompt are selected. By default, the foundational LLMs are not subjected to further training during this process but only the context window of the request is used as input for the LLM.		QPR ProcessAnalyzer AI Agent is based on the foundational large language models (LLMs) from selected providers (such as Meta, Claude and Mistral AI). When a new AI Agent is configured, the foundational LLM is selected, the input instructions are written, and the source datasets and analyses used as prompt are selected. By default, the foundational LLMs are not subjected to further training during this process but only the context window of the request is used as input for the LLM.

	=== AI ~~output labeling~~ ===		=== Output Labeling to Distinguish AI-generated Content ===
	QPR ProcessAnalyzer AI Agent functions as a separate, clearly identifiable component within the dashboards. This design makes it straightforward for users to distinguish between AI-generated content and human-generated content, fulfilling the requirement for transparency.		QPR ProcessAnalyzer AI Agent functions as a separate, clearly identifiable component within the dashboards. This design makes it straightforward for users to distinguish between AI-generated content and human-generated content, fulfilling the requirement for transparency.

	=== ~~Data~~ Masking for AI ~~handled data~~ ===		=== Masking for Data Processed by AI ===
	In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.		In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.

	[[Category: QPR ProcessAnalyzer]]		[[Category: QPR ProcessAnalyzer]]

Ollvihe: /* Data Masking */

2025-06-26T21:32:43Z

Data Masking

← Older revision		Revision as of 21:32, 26 June 2025
Line 94:		Line 94:
	QPR ProcessAnalyzer AI Agent functions as a separate, clearly identifiable component within the dashboards. This design makes it straightforward for users to distinguish between AI-generated content and human-generated content, fulfilling the requirement for transparency.		QPR ProcessAnalyzer AI Agent functions as a separate, clearly identifiable component within the dashboards. This design makes it straightforward for users to distinguish between AI-generated content and human-generated content, fulfilling the requirement for transparency.

	=== Data Masking ===		=== Data Masking for AI handled data ===
	In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.		In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.

	[[Category: QPR ProcessAnalyzer]]		[[Category: QPR ProcessAnalyzer]]

Ollvihe: /* Fulfilling AI system requirements */

2025-06-26T21:32:06Z

Fulfilling AI system requirements

← Older revision		Revision as of 21:32, 26 June 2025
Line 80:		Line 80:
	* If the AI Agent doesn't need to follow the dashboard filters, disable the "Follow Dashboard Filters" settings. When the settings is enabled, the AI Agent is rerun and the language model called every time filters are changed.		* If the AI Agent doesn't need to follow the dashboard filters, disable the "Follow Dashboard Filters" settings. When the settings is enabled, the AI Agent is rerun and the language model called every time filters are changed.

	== ~~Fulfilling AI system~~ requirements ==		== Generic requirements for AI systems ==
	~~=== Human oversight of~~ AI ~~generated insights and decision =~~==		This chapter describes how the AI Agent fulfills common generic requirements for AI-powered systems.
	~~To ensure robust human oversight,~~ the ~~QPR ProcessAnalyzer~~ AI Agent ~~has been designed with several key features. Firstly, all insights and recommendations generated by the AI are intended to support human decision-making rather than replacing it. The~~ AI ~~does not autonomously execute any actions, ensuring that human operators are always involved in the final decision~~-~~making process~~.

	~~Furthermore~~, the ~~interface is user-friendly~~, ~~with clear~~ and ~~accessible controls that allow users~~ to ~~easily review, modify, or override AI~~-~~generated recommendations as necessary~~. ~~This setup~~ not ~~only facilitates effective collaboration between the AI and its~~ human operators ~~but also ensures that human judgment and expertise remain central to~~ the ~~operational workflow~~.		=== Human supervision of AI-generated actions ===
			To ensure robust human oversight, the AI Agent has been designed with several key features. Firstly, all insights and recommendations generated by the AI are intended to support human decision-making rather than replacing it. The AI does not autonomously execute any actions, ensuring that human operators are always involved in the final decision-making process.

	=== AI ~~model~~ lifecycle management ===		Furthermore, the interface is user-friendly, with clear and accessible controls that allow users to easily review and override AI-generated recommendations as necessary. This setup not only facilitates effective collaboration between the AI and its human operators but also ensures that human judgment and expertise remain central to the operational workflow.
	QPR ProcessAnalyzer AI Agent is based on the foundational large language models (LLMs) from selected providers (such as Meta, Claude and Mistral AI). When a new AI Agent is configured, the foundational LLM is selected, the input instructions are written, and the source datasets and analyses used as prompt are selected. ~~The~~ foundational LLMs are not subjected to further training during this process but only the context window of the request is used as input for the LLM.
			=== AI models lifecycle management ===
			QPR ProcessAnalyzer AI Agent is based on the foundational large language models (LLMs) from selected providers (such as Meta, Claude and Mistral AI). When a new AI Agent is configured, the foundational LLM is selected, the input instructions are written, and the source datasets and analyses used as prompt are selected. By default, the foundational LLMs are not subjected to further training during this process but only the context window of the request is used as input for the LLM.

	=== AI output labeling ===		=== AI output labeling ===
Line 94:		Line 96:
	=== Data Masking ===		=== Data Masking ===
	In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.		In QPR ProcessAnalyzer AI Agent, robust data masking strategies are implemented by allowing users to carefully select and configure only the relevant datasets required for analysis when creating a new AI Agent. In QPR ProcessAnalyzer, users have access to the entire process mining model (eventlog) but the data masking techniques can additionally be used to hide any irrelevant or sensitive information. This approach aligns with the best practices of minimizing data sharing and protecting sensitive information.

	[[Category: QPR ProcessAnalyzer]]		[[Category: QPR ProcessAnalyzer]]

Ollvihe: /* Cost of Using Language Models */

2025-06-26T06:34:36Z

Cost of Using Language Models

@@ Line 79: / Line 79: @@
 * Write a prompt that instructs the language model to give shorter responses because the number of output tokens affect the costs. If you want to make sure that the response is no longer than a certain limit, the ''max_tokens'' parameter can be used (see [[AI_Agent#General_Settings|LLM Parameters]] setting above).
 * If the AI Agent doesn't need to follow the dashboard filters, disable the "Follow Dashboard Filters" settings. When the settings is enabled, the AI Agent is rerun and the language model called every time filters are changed.
 [[Category: QPR ProcessAnalyzer]]