Announcing Let's Talk v0.1.1: Interactive AI Chat for Websites
I'm excited to announce the first release of Let's Talk – an open-source, AI-powered chat widget designed specifically for websites, blogs and documentation sites. You can experience it live right now in the bottom-right corner of this page! or 👉
Contents
✨ What is Let's Talk?
Let's Talk is a plug-and-play chat widget that brings Retrieval-Augmented Generation (RAG), tool use and conversational AI directly to your readers. Built with Svelte, TypeScript, and Tailwind CSS on the frontend, it delivers a modern, responsive, and accessible chat experience. The backend leverages Python, FastAPI, LangChain, and LangGraph, with PostgreSQL, Qdrant, and Redis powering intelligent conversations through vector stores for knowledge retrieval. The entire solution is available for self-hosting with Docker.
Key Features in v0.1.1:
- RAG Enhancements: Improved information retrieval for more accurate answers.
- Svelte Frontend: Fast, lightweight, and easy to integrate.
- ChatWidget Improvements:
- Timezone-aware responses for global users.
- Smooth message streaming and markdown rendering.
- User/AI avatars and clear message separation.
- Animated chat open/close transitions.
- Error handling and loading indicators.
- RSS Feed Tool: Article summary support and flexible configuration.
- Vector Store Workflow: Streamlined build and configuration for custom knowledge bases.
- Docker & CI/CD: Ready for production deployment.
For a full changelog, see the release notes.
🛠️ How It's Integrated Here
On TheDataGuy, the ChatWidget is live and fully functional! Here’s how it works under the hood:
- Svelte Component: The widget is implemented as a Svelte component (
ChatWidget.svelte), leveraging reactive state for messages, loading, and UI transitions. - LangGraph SDK: It connects to a backend using the LangGraph SDK, supporting streaming responses and thread management.
- Timezone Support: The widget detects the user's timezone and includes it in the system prompt, ensuring context-aware answers.
- Markdown Rendering: AI responses are rendered with markdown, supporting code blocks and rich formatting.
- User Experience:
- Animated open/close button
- Responsive design for desktop and mobile
- Clear separation of user and AI messages with avatars
- Loading spinners and error messages for robust UX
- Attribution: A subtle footer credits Let's Talk and links to the open-source repo.
Related: Introducing Let's Talk
If you missed the original announcement and background, check out the Introducing Let's Talk post for the story behind the project, its goals, and early development insights.
🌐 Try It Out & Get Involved
- Live Demo 1: 👉
- Live Demo 2: D365 Stuff Chat - Powering the D365 Stuff Blog
- Source Code: github.com/mafzaal/lets-talk
- Release Notes: v0.1.1
If you’re looking to add conversational AI to your own blog or docs, give Let's Talk a try! Contributions, feedback, and feature requests are welcome.