Product Description zapAgent Wilma
Overview of zapAgent Wilma's functionality and license conditions
Wilma provides a conversational AI interface for working with analytics results generated by other zapliance agents. Wilma enables users to explore, interpret, and follow up on analytical findings in a structured, chat-based workflow. Wilma operates within the contractually agreed zapHome setup and can be used in both project-based one-off scenarios and recurring data analysis workflows.
1. Purpose
Wilma is the conversational AI agent within the zapliance solution. Her purpose is to support users in understanding, interpreting, prioritizing, and following up on analytics results produced by other zapliance agents. Wilma does not perform SAP data extraction or indicator-based processing herself. Instead, she complements analytical workflows by providing an interactive layer for analysis, reasoning, and investigation.
2. System Context
2.1 Architecture
Wilma operates exclusively in the analysis layer of the zapliance architecture. She has no SAP connection, does not execute RFC calls, perform or initiate SAP data extraction, and cannot expand the approved data scope.
Instead, Wilma works exclusively on already extracted and prepared data and attachments stored in a project database, which she access via MCP. Every Wilma action is triggered by a user.
For more information, see SAP Governance Data Extraction with zapliance.
Wilma operates within the contractually agreed zapHome setup and requires a deployed and operational LibreChat environment. The users access Wilma's web-based chat interface via their web browser.
Depending on the agreed setup, Wilma may utilize:
- Azure OpenAI Services
- MCP Servers
- Customer-provided documents
- MCP tools defined in Tool Bundles
- External systems connected through APIs
2.2 Model Selection
Wilma uses the customer-provided LLM through the customer's own Azure OpenAI Services.
Two recommended models are supported: GPT-5.4 and GPT-5.6.
The appropriate model depends on the customer's specific use case and requirements. zapliance can provide guidance for the decision-making based on the use case.
GPT-5.4 is recommended as the standard model for most use cases, providing a suitable balance between performance and cost.
GPT-5.6 provides stronger capabilities for use cases that require more advanced reasoning but typically results in higher usage costs.
3. Scope
This product description defines:
- the purpose and functional scope of generic zapAgent Wilma / LibreChat features
- the applicable license conditions for the use of zapAgent Wilma
- general customer cooperation obligations
- the features that can be activated through Tool Bundles
- detailed feature descriptions
- feature-specific customer cooperation obligations
- the feature scope of individual Tool Bundles
- customer-specific prompts, agents, tools, MCP Servers, or workflows
- generic features or requirements of zapHome
- Customer Success services
- Training and Consulting services
- Custom Development services
4. License Conditions
The customer acquires a license for the use of a Wilma instance, with the following license components:
- Service Start Date: The service start date regulates when the use of zapAgent Wilma can begin. If the offer includes a restriction on the service start date, this restriction applies to the specified period.
- Service Period: The service period defines the usage and billing period for Wilma and the activated Tool Bundles. If a subscription is agreed, the service period automatically renews for an additional term unless canceled in writing. If no subscription is agreed, the service period ends automatically at the end of the agreed term.
- Minimum Term: If a minimum term is defined in the offer, cancellation is only possible at the end of this term. The service period is effective from the service start date.
- Tool Bundles: Tool Bundles include contractually agreed licensing conditions and define the features, tools, capabilities, and integrations available within Wilma. If the customer-specific offer does not include specific tools or capabilities, the details of the specified Tool Bundle apply.
- Tool Bundle Updates: The number of Tool Bundle updates determines the maximum number of Tool Bundles configured specifically for the customer within a service period and delivered to the customer.
5. General Cooperation Obligations
5.1 General Prerequisites for Operating Wilma
- Provision of a server or virtual machine (VM) to serve as the host environment for the contractually agreed zapHome variant.
- Deployed and operational zapHome setup.
- Administrator rights for installing LibreChat on a virtual machine or server.
- Provision of user access via web browser.
- Availability and operationability of customer-provided Azure OpenAI Services resource (LLM) as specified in the following article: Setup Guide Azure OpenAI Setup
- Integration of created Azure OpenAI Services resource into LibreChat to be used by zapAgent Wilma.
- Provision of required endpoints and API keys for integration.
- Availability and operationability of MCP server launched by zapAnalytics, storing the project results.
- Integration of required MCP server and selection of assigned MCP slot to connect project results to Wilma and enable user interaction.
- The customer is responsible for security requirements and measures for operating the system. Security requirements for the server and system environment should meet the SAP data security standard.
- The customer ensures the logical and physical security of the IT infrastructure on which the digital products, the data generated by digital products, and/or Wilma operate or where the digital products and data are transported or transferred. The customer is guided by their already applied measures concerning the SAP productive system.
- Additionally, the customer ensures the interface security between the SAP systems, Wilma, and the web browser used for access. zapliance GmbH is not liable, subject to the provisions in section VII of the General Terms and Conditions, for damages due to lack of access protection, lack of security measures, insufficient archiving, and lack of encryption of the customer’s data.
- The customer ensures that (network) access to the customer’s SAP systems – or direct use of Wilma on the customer’s systems – and possibly on an SAP test system (with up-to-date original data) is enabled for Wilma. The customer is responsible for taking appropriate measures to prevent damage (e.g., virus infestation, system crash) due to access or software use on the customer’s systems.
- The customer is responsible for complying with all legal (especially data protection and labor law) regulations during software use, including the data protection officer and works council if necessary.
- In addition to the general cooperation obligations, the customer must fulfil the feature-specific cooperation obligations detailed in section 7.2 for the features activated by the selected Tool Bundle and Agent Team.
5.2 Technical Requirements
To operate Wilma, the contractually agreed zapHome instance must be deployed and operational.
A Linux-based virtual machine (VM) or server that meets the specified requirements must be prepared as the host environment for Docker.
The requirements, including recommended hardware and software configurations, are specified in this article: Setup Guide Virtual Machine / Server
We recommend Docker as the container management tool. For other deployment scenarios, please refer to this article.
6. Generic LibreChat Features Used by Wilma
-
Browser-based Chat Interface
- User access through web browser
- Conversational interaction
- Support for multiple chat sessions
-
User Creation
-
Creation of user accounts for accessing Wilma
-
-
User Login (with optional SSO)
- Authentication via user credentials
- Optional Single Sign-On (SSO)
-
Password Reset on Login Screen
-
Password reset functionality directly from the login screen
-
- Change of Login Credentials (E-Mail & Password)
- Modification of user e-mail addresses and login passwords
- Prompt Library
- Creation, management, sharing, and execution of reusable prompts
- Optional usage of variables
- Agent Builder
- Creation and management of AI agents
- Configuration of names, prompts, models, capabilities, and settings
- Agent Selection
- Selection of created AI agents
- MCP Slot Selection
- Use of connected MCP Servers within agents
- File Upload in Chat Session
- Upload of documents and files for analysis within chat conversations
- Chat History and Search
- Storage and retrieval of previous chat sessions
- Chat Session Export
- Export of chat sessions in various formats, e.g. for support cases
- Memory Management
- Creation and management of memory entries
- Filtering and maintenance of stored memories
7. Features Activated by Tool Bundles
7.1 Available Features by Agent Team
This is a list of all Wilma features activated by Tool Bundles and their availability per Agent Team.
The features activated in a specific setup depend on the selected Tool Bundle. See the applicaple Tool Bundle Description for detailed information on the features activated by each Tool Bundle.
|
Feature |
Wolfgang + Wilma |
Werner + Wilma |
Wickie + Wilma |
|
LLM Data Analysis on Request1 |
✅ |
✅ |
✅ |
|
LLM Attachment Analysis on Request2 |
✅ |
✅ |
✅ |
|
Automatic Selection of MCP Tools3 |
✅ |
✅ |
✅ |
|
Pseudonymization of Results4 |
✅ |
✅ |
✅ |
|
Link to SAP Documents (WebGUI/Fiori) in Chat on Request5 |
✅ |
✅ |
✅ |
|
Custom MCP Tools6 |
✅* |
✅* |
✅* |
|
Web Search (via Azure OpenAI)7 |
❌ |
❌ |
✅ |
|
VAT ID Validation via Endpoint Connection (BZSt & VIESS)8 |
❌ |
❌ |
✅ |
Legend:
✅ = Feature available for this Agent Team
❌ = Feature not available for this Agent Team
✅* = Custom feature available for this Agent Team as part of a Custom Tool Bundle. Only activated if explicitly included in the commercial agreement. May require a custom development project that is charged separately.
7.2 Feature Descriptions and Cooperation Obligations
1. LLM Data Analysis on Request
Analyzes and explains, at the user’s request, specific SAP data already stored in the project database. This feature does not extract data from SAP.
2. LLM Attachment Analysis on Request
Analyzes and explains, at the user’s request, specific SAP attachments already stored in the project database. This feature does not download attachments from SAP.
Cooperation Obligations
- create and deploy additional Azure OpenAI Services resource (Document Intelligence/OCR)
- provide required endpoints and API keys for integration
- see details in Setup Guide Document Intelligence
3. Automatic Selection of MCP Tools
Wilma automatically selects the most appropriate available MCP tools required to answer the user's request and analyze the relevant data.
4. Pseudonymization of Results
Project results made available to Wilma via the MCP server are provided in pseudonymized form according to zapliance pseudonymization logic.
5. Link to SAP Documents (WebGUI/Fiori) in Chat on Request
Allows users to access individual SAP documents directly from the chat interface without additional effort of searching for the specific item in SAP.
The feature utilizes SAP WebGUI or SAP Fiori. It is only used to access individual SAP documents from the Wilma chat after data extraction and processing have been completed. It is technically separate from RFC-based data extraction and cannot influence its scope or execution.
Cooperation Obligations
Conduct all necessary preparations listed in the article Setting Up SAP Web GUI, including:
- provide users with required SAP user access to view documents
- provide users with access rights to required transaction codes:
- FB03: Display accounting documents
- ME53: Display purchase requisitions
- ME23: Display purchase orders
- VL03: Display deliveries
- VA03: Display sales orders
- VF03: Display billing documents
- provide WebGUI or Fiori URL
- ensure that user's devices have access to the provided WebGUI or Fiori URL
6. Custom MCP Tools
Enables Wilma to use customer-specific MCP tools developed by zapliance according to the customer's requirements.
Custom MCP Tools require a separate development engagement.
Cooperation Obligations
- define the required business and technical requirements
- participate in specification and validation activities
- provide all information required for implementation and testing
7. Web Search (via Azure OpenAI)
Utilizes the customer's Azure OpenAI Web Search tool to conduct targeted web search, for example for VAT ID validation and company detail research.
It works exclusively on already extracted and processed data and has no connection to SAP or RFC-based data extraction. It cannot access SAP, initiate or modify an extraction, or influence the approved extraction scope.
Cooperation Obligations
- set up Azure OpenAI Services
- activate the web_search tool in the Responses API
- see details in Setup Guide Azure OpenAI
8. VAT ID Validation via Endpoint Connection (BZSt & VIESS)
Utilizes the customer's Azure OpenAI Services for data validation via targeted endpoint connections, for example for VAT ID verification.
It works exclusively on already extracted and processed data and has no connection to SAP or RFC-based data extraction. It cannot access SAP, initiate or modify an extraction, or influence the approved extraction scope.
Cooperation Obligations
- set up Azure OpenAI Services
- see details in Setup Guide Azure OpenAI
In addition to this Product Description, the Product Descriptions of all components included in the offer, contractual or commercial agreement apply.
The General Terms and Conditions (GTC) of zapliance in their currently valid version apply.