Agent Node

Represents an autonomous AI entity capable of securely reasoning, planning, and interacting with authorized tools or knowledge vaults to accomplish a given enterprise objective.

Functionality

This node uses an LLM acting as a "brain" to dynamically decide a sequence of unpredictable actions. Based on the human's goal — provided via messages/input — it can proactively choose to use available pre-defined Tools or semantically query Document Stores to gather missing information or execute operations on external servers. It strictly manages its own internal reasoning cycle and handles memory threads beautifully.

Configuration Parameters

• Model: Specifies the underlying AI model driving the reasoning execution loop.
• Messages: Define the initial conversational input, prime directive, or context for the agent pipeline, structuring it across defined roles (System, User, Assistant).
• Tools: Specify an array of explicitly registered Lyntaris Tools the agent is legally authorized to execute.
  • Optional: Require Human Input flags can be toggled per tool, guaranteeing the agent pauses entirely to await manual manager intervention before performing sensitive tasks like writing to a database.
• Document Stores: Choose highly secure Document Store partitions from which the agent can semantically retrieve company information arrays.
• Describe Knowledge: Provide an explicit natural language system instruction regarding the contents of designated Document Stores, effectively teaching the Agent when it would be semantically profitable for it to search that vault.
• Vector Embeddings: Configure robust ad-hoc connections to external, isolated vector clusters if Document Stores aren't utilized.
• Memory Capabilities: Supports complete chat memory context retrieval, token-windowing configurations, and custom Input Messages mapping.
• Return Response: Define if the final un-tooled generation represents an Assistant or User generation for downstream flows.
• Update Flow State: Bind extracted agent reasoning logic securely directly into $flow.state keys for subsequent pipeline usages.

Inputs & Outputs

• Inputs: Expects natural language directives or variables via {{ variable }}.
• Outputs: Streams or finalizes the deterministic result the agent formulated after all internal tool loops sequence conclusions.