Nextjs · Langchain · VectorDB · Gemini · PuppeteerWebBaseLoader

Chat-JPT

A chatbot built using Langchain, VectorDB, and Gemini Model that can answer questions and provide information about me and my work.

GithubVisit
Chat-JPT main preview

Project details

Tools & stack

  • Next.js
  • LangChain
  • VectorDB
  • Gemini
  • Supabase

Problem and solution

problem

Hard to explore a portfolio at scale

Visitors want quick answers about your work, skills, and background without clicking through every page or reading long case studies.

solution

RAG-powered conversational assistant

LangChain and a vector database ground Gemini responses in your actual content, so questions get accurate, contextual answers in natural language.

Process & highlights

01

Knowledge sourcing

Collected portfolio content, project summaries, and key facts so the assistant had a useful and focused response base.

02

Retrieval tuning

Adjusted chunking and vector lookup behavior to keep responses grounded in the right project details.

03

Answer shaping

Guided the model to answer in a concise, portfolio-friendly tone that stays useful without sounding generic.

04

Conversation polish

Refined response states and fallbacks so the chatbot feels intentional even when a question needs a careful answer.

Project output

Conversational portfolio assistant

6 highlights

Chat-based portfolio navigation

LLM-powered answers

Retrieval-grounded responses

Project knowledge base

Safer, more reliable responses

Fast Q&A experience