Delight

Multi-agent AI assistant to boost your productivity. Access 30+ models, automate tasks, enhance writing and speed up work.

Comment

Inspiration

AI is everywhere, but it's scattered. I found myself constantly jumping between ChatGPT for brainstorming, Claude for analysis, Gemini for quick tasks — each in separate tabs with different logins and lost context.

The real spark came from watching my fiancée work as an executive assistant. Her days are packed with scheduling meetings, drafting emails, booking flights, and chasing down details across dozens of browser tabs. I thought: What if she could have an AI assistant for her assistant work — built right into the browser where she already spends her day?

This was ambitious for a solo developer. Building a multi-provider AI system with persistent conversations and automation would normally take months. But with Kiro as my development partner, I had the confidence to attempt something at this scale. Kiro gave me the ability to think big and execute fast.

What Delight Does

Instead of juggling multiple AI platforms, Delight feels like one unified assistant that:

  • Unifies 7+ major AI providers (OpenAI, Anthropic, Google Gemini, Grok, Groq, SambaNova, Ollama) with 25+ models
  • Remembers every conversation across browser sessions — no more starting over
  • Lets you switch AI models mid-chat without losing context
  • Pulls webpage content directly into conversations for smart analysis
  • Automates web tasks using plain English commands
  • Includes 10+ writing tools for explaining, rewriting, summarizing, and tone adjustment

The magic is in the seamless experience. Users don't feel like they're using "different AIs" — it's just one intelligent assistant that happens to have access to the best of every provider.

Kiro's impact here: This wasn't just a prototype. Kiro helped me generate production-grade systems with enterprise-level error handling, performance optimization, and user experience polish that would typically take a team months to build.

How I Built It with Kiro

My development process with Kiro was fundamentally different from traditional coding:

Specs Before Code

As I interacted with Kiro, we created structured specifications in a .kiro/specs/ folder for each major feature. Kiro thrives on clarity — the better I defined what I wanted, the better code it generated.

Incremental Prompting Method

I followed a layered approach:

  1. Foundation: Core interfaces and data structures
  2. Integration: Provider connections and message handling
  3. Advanced features: Automation, optimization, error recovery
  4. Polish: Performance tuning and user experience

The ConversationManager Achievement

The most impressive example is Kiro generating the ConversationManager — a 500+ line TypeScript service that handles:

  • Session persistence across browser restarts
  • Provider switching with context preservation
  • Automatic cleanup and performance optimization
  • Graceful error recovery and fallback systems

This single component would have taken me weeks to design and implement properly. Kiro generated it systematically based on clear specifications.

Agent-Powered Automation

For web task automation, Kiro helped me build a multi-agent system where users can say "extract all email addresses from this page" and the extension actually does it — planning, executing, and handling errors automatically.

Kiro's key benefits:

  • Saved weeks of boilerplate: Error handling, edge cases, and optimization came built-in
  • Let me focus on architecture: I spent time on product vision while Kiro handled implementation details
  • Enabled "vibe coding": I could iterate quickly without losing code quality or structure

Challenges I Ran Into

Provider Inconsistencies

Each AI provider has different message formats, token limits, and streaming protocols. Kiro helped design conversion logic and unified interfaces that made 7+ providers feel like one system.

Performance with Long Histories

Loading conversations with hundreds of messages initially took 2-3 seconds. Kiro proposed and implemented lazy loading, message compression, and intelligent cleanup strategies I wouldn't have thought of.

Graceful Error Handling

AI providers fail sometimes. Instead of showing users error screens, Kiro helped build automatic retry mechanisms and provider fallbacks that keep conversations flowing seamlessly.

Browser Automation Edge Cases

Making natural language commands work reliably across different websites required sophisticated DOM analysis and error recovery. Kiro generated robust solutions for handling these complex scenarios.

Kiro wasn't just generating code — it was solving problems and proposing strategies that elevated the entire project.

Accomplishments I'm Proud Of

  • Published to Chrome Web Store: Delight went from idea to published product in one week
  • The ConversationManager: A complex service that rivals what you'd find in commercial software
  • Seamless multi-provider UX: Users genuinely feel like they're using one AI, not seven
  • Speed to production: Built something that would normally take months in days

The real accomplishment: Kiro enabled me to build at a scale and quality level that would typically require a full development team.

What I Learned

Specs + Incremental Prompting = Huge Productivity

Writing clear specifications before asking Kiro to code produced dramatically better results than broad, vague requests.

AI as a Development Partner

Kiro taught me a new way of building software — more conversational, iterative, and structured. Instead of wrestling with implementation details, I focused on user problems and product vision.

Building for Real Users is Different

Every design choice was about reliability, speed, and smooth experience. Kiro helped ensure the "boring but critical" parts (error handling, performance, cleanup) were built right from the start.

Production-Grade Code from Day One

Traditional development often involves building quick prototypes and then rewriting for production. With Kiro, the first version was already production-ready.

What's Next for Delight

The roadmap for Delight includes:

  • Expanded automation: More complex web interactions powered by natural language
  • Specialized AI tools: Translation, code review, research assistance
  • More providers: Supporting additional AI models and open-source options
  • Community platform: Website with tutorials, templates, and user community

As Delight grows, I see Kiro as a long-term development partner — enabling me to scale features quickly without compromising on quality or user experience.

The ultimate goal remains simple: make Delight the one-stop AI assistant that lives in your browser and actually makes your work easier.

Built With

Share this project:

Updates