This repository showcases a collection of production-ready AI agents and automation workflows built using n8n and Relevance AI. These projects demonstrate the use of multi-agent orchestration, Retrieval-Augmented Generation (RAG), and conditional logic to solve complex business challenges.
An autonomous research agent that transforms a company URL into a high-level sales strategy.
- The Problem: Sales reps spend 30+ minutes researching a lead before a call.
- The Solution: A modular suite consisting of a "Brain" (Copilot) and specialized "Muscle" (Tools).
- Research Company Tool: Scrapes and summarizes company websites via Firecrawl.
- Research Prospect Tool: Ingests LinkedIn data to build a persona profile.
- Report Generator: Synthesizes all data into a strategic "Pre-call Report."
- Tech Stack: Relevance AI, GPT-4o, Firecrawl API, LinkedIn Scraper.
A full-lifecycle Retrieval-Augmented Generation pipeline using vector embeddings.
- Ingestion (Workflow 3): Monitors Microsoft OneDrive ➔ Recursive Character Text Splitting ➔ Vectorization ➔ Storage in Pinecone.
- Retrieval (Workflow 4): A chat-based agent that uses the Pinecone index to provide fact-based answers, preventing model hallucinations.
- Technical Highlight: Implemented Recursive Character Splitting to maintain semantic integrity during the chunking process.
- Tech Stack: n8n, OpenAI Embeddings, Pinecone Vector DB, OneDrive API.
- Function: Automatically classifies incoming emails into categories (e.g., Internships, Scholarships, General Inquiry).
- Logic: Uses an LLM to determine intent, drafts a context-aware response in Gmail, and assigns priority labels.
- Tech Stack: n8n, OpenAI GPT-4o-mini, Gmail API.
- n8n Workflows: Download the
.jsonfiles and import them into your n8n instance (Settings > Import Workflow). - Relevance AI: Upload the
.raifiles into the Relevance AI "Tools" or "Agents" section. - Credentials: You will need to provide your own API keys for OpenAI, Pinecone, and Google/Microsoft OAuth.
- Modularity: I build specialized tools rather than "monolithic prompts" to ensure higher accuracy and easier debugging.
- Agentic Thinking: Moving beyond simple "If/Then" logic into agents that can plan, research, and execute tasks autonomously.
- Security First: All workflows provided here have been sanitized. API keys and personal IDs have been replaced with environment variables.