This workflow uses an AI agent to orchestrate the entire conversation. The agent determines which knowledge source (expert graph) to use based on the user's spoken question and then generates a personalized answer.
Step-by-Step Process:
- Voice Input
The user asks a question through a voice interface powered by ElevenLabs. - Prompt Delivery
The question is forwarded via a webhook, including a session ID to maintain conversation history and context. - Agent Decision-Making
The AI agent receives the prompt and evaluates which knowledge tool or "expert" is best suited to answer. Each expert is backed by a unique knowledge graph and described using metadata generated by InfraNodus. - Prompt Optimization
The agent may refine or rephrase the question to better align with the selected expert’s format. - Knowledge Graph Query
The updated query is sent to the chosen GraphRAG-powered knowledge source, which analyzes the graph and extracts relevant insights. - Expert Response
The expert provides a rich, context-aware answer, along with supporting statements retrieved using a combination of Retrieval-Augmented Generation (RAG) and GraphRAG methods. - Answer Synthesis
The AI agent collects responses from the experts and composes a final, cohesive answer. - Voice Output
The completed response is sent back to the webhook. ElevenLabs receives it, transforms the text into a natural-sounding voice reply, and delivers it back to the user in conversation.