Skip to content

Abdullah1053/Takder-Ai

Repository files navigation

Takder‑Ai: Your Intelligent Knowledge Architect

Takder‑Ai is a sophisticated, AI-driven searchable knowledge base designed to transform static documents into dynamic, interactive assets. By leveraging cutting-edge Retrieval-Augmented Generation (RAG), Takder‑Ai allows users to upload PDFs and notes, creating an indexed repository that can be queried in natural language to extract precise, context-aware insights.


🏛️ Architecture Overview

The application follows a robust RAG (Retrieval-Augmented Generation) pipeline to ensure accuracy and groundedness in every response:

  1. Ingestion: Files (PDFs, Markdown, Text) are uploaded and processed.
  2. Preprocessing & Chunking: Documents are split into semantic chunks to preserve context while fitting model constraints.
  3. Embedding: Each chunk is converted into a high-dimensional vector using Google Gemini's embedding models.
  4. Vector Store: Vectors are stored in a specialized database (PostgreSQL with pgvector or similar) for lightning-fast similarity searching.
  5. Retrieval: When a query is asked, the system retrieves the most relevant document fragments based on vector similarity.
  6. Generation: The retrieved context + the original query are fed into Google Gemini Flash to generate a grounded, source-cited response.

💻 Tech Stack Breakdown

Frontend & Framework

  • Next.js (React-based hybrid framework for speed and scalability)
  • Tailwind CSS (Utility-first styling for a polished, modern interface)
  • Lucide React (Clean, consistent iconography)
  • Motion (Smooth, fluid UI transitions and micro-interactions)

AI & Logic

  • LangChain (The orchestration layer for RAG pipelines and LLM workflows)
  • Google Gemini Flash (High-performance large language model for generation)
  • Google GenAI SDK (Native integration for embeddings and multi-modal analysis)

Infrastructure

  • PostgreSQL / pgvector (Relational data + vector operations)
  • Vite/Express (Development environment and server-side logic)

🗓️ The 21-Day Roadmap

Week 01: The Core

  • Architecture Setup: Project initialization with Next.js & PostgreSQL schema design.
  • Ingestion Engine: Implementation of the multi-format file ingestion logic (PDF/Notes).
  • Embedding Pipeline: Integration of Google GenAI for vectorization of document chunks.

Week 02: Intelligence

  • Gemini Integration: Orchestrating LangChain with Google Gemini 1.5 Flash for RAG.
  • Chat Interface UI: Developing the interactive natural language query interface.
  • Vector Refinement: Optimizing similarity search and retrieval accuracy.

Week 03: Stability

  • Security & Access: Implementation of the authentication layer and role-based access.
  • Edge Optimizations: Latency reduction and context window performance tuning.
  • Production Ready: Full-scale testing, audit, and production deployment on Vercel/Cloud Run.

🔑 Environment Variables

To run Takder‑Ai, you need to configure the following keys in your .env file:

# AI & LLM
GEMINI_API_KEY="your_google_gemini_api_key_here"

# Database
DATABASE_URL="postgresql://user:password@localhost:5432/takder_ai"

# Application
APP_URL="http://localhost:3000"
NODE_ENV="development"

🚀 Setup Instructions

1. Prerequisites

Ensure you have Node.js (v18+) and npm/yarn installed.

2. Installation

Clone the repository and install dependencies:

git clone https://github.com/your-username/takder-ai.git
cd takder-ai
npm install

3. Database Setup

Ensure PostgreSQL is running with the pgvector extension enabled.

CREATE EXTENSION IF NOT EXISTS vector;

4. Running the App

Start the development server:

npm run dev

The app will be available at http://localhost:3000.


"Turning data into dialogue."The Takder‑Ai Team

About

Takder‑Ai is a web app that turns your uploaded PDFs and notes into a searchable knowledge base. You can ask questions in natural language and get answers sourced directly from your own documents. Built with Next.js, LangChain, Google Gemini Flash, and PostgreSQL.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors