NIMO

A Local-First, Website-Aware Assistant Built Into the Portfolio

NIMO is an interactive assistant designed specifically for this portfolio. Instead of acting like a generic chatbot, it understands the site itself — projects, case studies, navigation, developer concepts, current-page context, and user intent.

NIMO Assistant Interface and Local Intent Architecture
Design & Development
Interactive Assistant
100% Local-First
Trilingual NLU
HTML5 / CSS / ES Modules
Input Focus Isolated
01. Overview

Site-Aware Intelligence.

Problem

Traditional portfolios force visitors to manually navigate menus, scroll past project cards, and read long case studies to find specific details. Recruiters and visitors frequently seek immediate answers: "Which projects involve UI?", "What is ToolVerse?", "What am I looking at?", "Summarize this project", "Who built this website?", or "What does HTML mean?".

Solution

NIMO was designed as a grounded, website-aware assistant embedded directly into the portfolio. Rather than relying on heavy third-party AI APIs or generic chatbot widgets, NIMO resolves natural-language navigation, project lookups, current-page context, developer concepts, and multilingual queries locally using a structured intent engine.

Target Audience

Recruiters, hiring managers, and developers exploring the portfolio who want instant answers, fast section navigation, case study summaries, or developer concept explanations without manual searching.

Key Objectives

Provide zero-latency answers, track current page and project context automatically, support trilingual English, Hindi, and Hinglish queries, enforce a strict no-hallucination rule, and maintain full keyboard focus isolation with Arcade OS.

02. Design Philosophy

Embedded by design.
Intelligent by context.

NIMO is intentionally built to feel like an organic part of the portfolio experience rather than an embedded third-party widget. Its intelligence is grounded directly in real project metadata, page state, and structured developer knowledge.

Native Visual System

Utilizes the portfolio's native CSS tokens, glassmorphism layers, dark/light theme switching, and responsive layout rules.

Grounded Knowledge Source

All answers draw from verified portfolio metadata, structured case study context, site configurations, and built-in developer concepts.

Contextual Action Triggers

Every response offers dynamic action buttons, enabling users to jump to sections, open live apps, or ask follow-up questions in one click.

No-Hallucination Reliability

NIMO never fabricates credentials or project claims. If a query falls outside its structured knowledge base, it gracefully provides a polite redirect.

03. Architecture

Local First. Zero Server Latency.

NIMO operates entirely inside the client's browser using modular Vanilla JavaScript (ES Modules). This architectural choice eliminates external API subscriptions, key leakage risks, and network latency, providing sub-millisecond query evaluation.

When a user submits a message, NIMO passes the text through a 6-stage local execution pipeline: language detection, text normalization, entity extraction, context resolution, intent scoring, and response localization.

HTML5
CSS3 (Variables)
Vanilla ES Modules
Intent NLU Engine
Context Store
Trilingual Dictionary
04. Intent NLU Engine

Scalable Intent Architecture.

Rather than relying on brittle hardcoded if/else conditions, NIMO implements a scalable intent-resolution engine. Wording variations, shorthand, typos, and phrases map into canonical intent categories.

  • Creator Identity: "Who built this?", "Who made this site?", "Whose portfolio is this?" $\rightarrow$ Creator intent.
  • NIMO Identity: "What is NIMO?", "Who is NIMO?", "Explain NIMO" $\rightarrow$ NIMO definition.
  • AI Honesty: "Are you an AI?", "Are you a chatbot?" $\rightarrow$ Technical honesty response.
  • Contextual Pronouns: "Tell me about ToolVerse" $\rightarrow$ "What tech does it use?" $\rightarrow$ Resolves "it" to ToolVerse.
  • Typo Tolerance: Handles common typos (e.g. htlm $\rightarrow$ html, nintedno $\rightarrow$ nintendo).
  • Vocative Stripping: Strips conversational names (e.g. "How are you NIMO?" $\rightarrow$ "How are you").
05. Context Awareness

Current-Page & Case Study Awareness

NIMO dynamically detects the user's active location across Home, Projects, About, Contact, Arcade OS, and individual project case studies.

  • "What am I viewing?": Identifies the exact active section or case study page.
  • "Summarize this project": Summarizes the current case study without requiring the project title.
  • "What tech was used?": Lists documented technologies for the currently active project.
  • "What problem does it solve?": Extracts core challenges from case study metadata.
06. Multilingual Engine

Trilingual Conversation Layer

NIMO supports English, Devanagari Hindi, and Hinglish (Roman Hindi) through a single canonical intent architecture.

  • English: "Take me to Projects" $\rightarrow$ Navigates to Projects section in English.
  • Hinglish: "Mujhe projects par le chalo" $\rightarrow$ Resolves navigation intent & replies in Hinglish.
  • Hindi (Devanagari): "मुझे प्रोजेक्ट्स पर ले चलो" $\rightarrow$ Transliterates & replies in modern Devanagari Hindi.
  • Style Mirroring: Automatically mirrors user's language preference while retaining technical English terms.
07. Engineering Challenges

Key Technical Challenges & Solutions

WASD Input Conflict

Problem: Arcade OS registered global WASD keyboard handlers for retro game movement. When typing inside NIMO, WASD keys triggered game actions.

Solution: Updated keydown handlers with input-focus isolation checking editable targets (input, textarea, contenteditable) to bypass game controls during active typing.

Multilingual Engine

Problem: Supporting English, Hindi, and Hinglish without duplicating business logic or codebase size.

Solution: Built a 3-layer language pipeline: automatic style detection $\rightarrow$ Devanagari transliteration & phrase normalization $\rightarrow$ canonical intent resolver $\rightarrow$ response output localizer.

Contextual Pronoun Memory

Problem: Handling follow-up queries like "Tell me about ToolVerse" followed by "Summarize it" or "Open it".

Solution: Engineered a ConversationalContext store tracking last-viewed project entities and section states to resolve pronouns without manual repetition.

08. Feature Summary

Capability Matrix

  • Navigation: Natural-language portfolio section routing.
  • Project Discovery: Category, tech stack, and alias filtering.
  • Case-Study Context: Current-page awareness and instant summaries.
  • Conversational Memory: Follow-up pronoun resolution.
  • Developer Glossary: Built-in 75+ web development definitions.
  • Trilingual Support: English, Hindi (Devanagari), and Hinglish (Roman).
  • 100% Local-First: Zero external AI APIs, zero server latency.
  • Input Safety: Keyboard focus isolation with Arcade OS.
09. Outcomes

Grounded & Reliable

NIMO evolved from a simple phrase matcher into a complete portfolio interaction layer capable of natural navigation, contextual case study assistance, trilingual queries, and developer education — while remaining 100% local-first and tightly integrated with the portfolio's visual system.

Automated test suites confirm 100% pass rates across 58 unit tests and 25 multilingual intent evaluation scenarios.

10. Future Possibilities

Roadmap

Future iterations may include an optional secure server-side LLM fallback for out-of-scope domain questions, Web Speech API integration for hands-free voice commands, and expanded recruiter-mode evaluation features.