Gen AI + AI agents.

AI Connectors in the Integration Layer define how Rayven connects to LLM provider APIs. Gen AI + AI Agents in the Execution Layer defines how those connections are used within workflow logic - for document processing, structured extraction, conversational interfaces, agent-style automation + RAG. The connectors are the pipe; this page is what flows through it.

Yes. Rayven's LLMOps capability allows custom LLMs to be trained on your specific operational data, documents + historical records. The trained model is deployed within the platform and continuously updated as new data accumulates - delivering responses grounded in your operational context.

A policy-aware agent is configured with specific action permissions, prompt constraints + governance controls. For example, an agent can be permitted to raise a work order or generate a report - but not modify a configuration or access data outside its assigned Label scope. Policies are defined per agent and enforced at the platform level.

Yes. AI agent outputs can trigger any output node in a workflow - writing to a database, calling an external API, sending an alert, raising a work order or triggering a control command. The agent determines the action based on its configured prompt + policy set; the workflow executes it.

Retrieval-augmented generation grounds LLM responses in your actual operational data by retrieving relevant context from Cassandra + MySQL before the model generates a response. This improves accuracy, reduces hallucination risk + ensures answers reflect current, governed data rather than generic model knowledge.

Yes. Rayven is unique in supporting multiple AI nodes within a single workflow. This enables multi-step AI pipelines - for example, one model extracts structured data from a document, a second classifies the output, and a third generates a summary. Each step passes its output to the next.

Yes. Conversational analytics widgets are configurable components that can be embedded in any Rayven interface, including white-label applications built for clients. The interface is trained on client-specific data and presented under the partner's own branding - with no visible Rayven infrastructure.

All LLM interactions are logged with input context, prompt, output + timestamp in Cassandra. Prompt management, memory limits, rate controls + policy-aware action constraints are configured per agent or LLM node. Full audit trail is accessible via the Inspect Data tab.

Yes. A GenAI workflow can be configured to query relevant data from Cassandra or MySQL, pass it to an LLM node with a report generation prompt + write the structured output to a Dynamic Report widget or export as a document. Scheduled report generation on a cron trigger is fully supported.

Yes. Custom LLMs are continuously updated using real-time data ingestion. As new operational records, documents + events accumulate in Cassandra, the model's training context expands. Retraining frequency is configurable to balance model freshness against compute requirements.