Functional requirements

Agent component

• The system must allow users to deploy prebuilt AI agents for common use cases
• Users must be able to create custom AI agents tailored to their specific needs
• Users must be able to create multiple agents with the same role but different configurations/prompts
• The system must support defining(assigning) roles to agents that define their domain expertise
• Agents must be able to perform context-aware tasks (skills that include tools) associated with their role

Skill & tool components

• Users must be able to define agent skills using a prompt builder interface
• Each skill must support integration one or many tools
• Tools must be reusable and perform discrete tasks as part of an agent’s workflow

Knowledge base component

• The platform must allow uploading and registering of knowledge sources (databases, documents, etc.) (not implemented yet through GUI)
• Agents must be able to search and extract information from the knowledge base
• The knowledge base must be customizable per business/client

Language model & Brain component

• The system must support multiple LLMs (e.g., GPT, Claude, Llama, DeepSeek).
• The system must allow LLMs to be fine-tuned per agent.

Guardrails

• Users with developer role should be able to define restrictions for agent behavior (guardrails)
• The system must enforce these restrictions during agent operation

Insights & analytics (not implemented yet)

• The system must generate insights from agent activity (e.g., performance metrics, user behavior).
• Insights must be visualized in a dashboard for business decision-making.

User interaction

• The platform must support multi-language interaction (e.g., English, Macedonian, Albanian).
• Agents must be accessible through multiple channels (e.g., web, chat platforms).

Customization & personalization

• Users must be able to customize agent prompts, skills, and responses
• The system must allow personalized outputs based on business context and user input

Non-functional requirements

Performance

• Agents must be able to handle concurrent user requests without degradation of performance

Scalability

• The platform must support horizontal scaling to handle increasing numbers of users or agents
• The architecture must allow modular upgrades (e.g., replace LLMs, knowledge systems)

Availability & reliability

• The platform must maintain 99.9% uptime monthly
• Agents must function consistently and accurately across tasks and sessions

Security

• All data exchanged must be encrypted in transit and at rest
• The system must enforce role-based access control for users and admins
• Guardrails must ensure agents do not leak sensitive information or behave unethically

Usability

• The platform must provide an intuitive UI for both technical and non-technical users
• Agent creation and customization must be user-friendly with low-code/no-code interfaces

Maintainability

• The system must support continuous updates of the knowledge base and model configurations
• It must allow for easy debugging and logging of agent performance

Interoperability

• The platform must integrate with third-party systems/APIs for data, tools, and workflows.
• It must support different file formats and data types for knowledge ingestion

Localization

• The platform must support UI and agent responses in multiple languages