Back to Projects
Institutional Knowledge Engine

CISNR Voice-Activated RAG Agent

n8nPineconeElevenLabsGemini 2.5 ProGoogle Drive API

The Problem

Institutional knowledge at CISNR was trapped in static PDF reports and legacy documents stored in Google Drive. Researchers and staff struggled to find specific information quickly, leading to wasted time and underutilized data assets.

The Solution

I created an 'Always-On' ingestion engine using n8n. The system watches Google Drive for new files, automatically chunks the text, creates vector embeddings, and stores them in Pinecone. I then built a Voice Agent using ElevenLabs that can query this database, allowing users to 'talk' to the documents.

Automated PDF Ingestion & Chunking Pipeline.
Vector Storage in Pinecone for Semantic Search.
Voice-to-Voice Interface using ElevenLabs.
High-speed retrieval using Gemini 2.5 Pro.

Agentic Architecture

This system relies on event-driven orchestration handled via n8n and Python microservices.