A collection of examples demonstrating different approaches to building AI chat applications with tool calling capabilities.
chat/ - Basic Chat Application
Simple interactive chat application built with umwelten classes. Demonstrates basic chat functionality with support for multiple model providers (OpenAI, Anthropic, Google, Ollama, OpenRouter, LM Studio) and file attachments.
Key Features:
- Multi-provider support with easy switching
- File attachment capabilities
- Streaming responses
- Command-line interface
weather-umwelten/ - Tool Calling with Umwelten Framework
Weather assistant chatbot using the Umwelten framework for tool orchestration. Shows how to build AI agents with declarative tool composition and automatic tool chaining.
Key Features:
- Declarative tool definitions
- Automatic tool orchestration
- Weather and location lookup tools
- Built-in error handling and retry logic
weather-straight-sdk/ - Direct Vercel AI SDK Integration
Weather assistant built directly with the Vercel AI SDK without framework abstraction. Demonstrates manual tool orchestration and explicit control over tool execution.
Key Features:
- Direct tool control and visibility
- Manual tool chaining
- Explicit error handling
- Minimal dependencies
- Both interactive and non-interactive modes
assistant-ui/ - Next.js Web Interface with Assistant UI
Modern web-based weather assistant built with Next.js and Assistant UI framework. Demonstrates how to integrate umwelten with a full-stack web application using the latest UI components.
Key Features:
- Next.js 15 with App Router
- Assistant UI framework integration
- Modern React components with shadcn/ui
- Streaming chat interface
- Tool calling with visual feedback
- Responsive design
- pnpm + TypeScript: All projects use pnpm for package management and TypeScript for type safety
- tsx execution: Direct TypeScript execution without separate build steps
- Latest libraries: Always use the latest versions of dependencies
- Node.js testing: Prefer
node --testover external testing frameworks