GLLM Web: A Lightweight Web Client Built for Go Language LLM Backends

GLLM Web is a headless web client designed specifically for GLLM (Go Large Language Model) SSE backends. It uses native TypeScript and CSS with zero dependencies on front-end frameworks. Key features include dynamic reasoning display, intelligent agent action tracking, and glassmorphism design. This project balances modern UI experience with minimal technical complexity, making it an ideal choice for GLLM backend integration.

In LLM application development, front-end interfaces face a dilemma: heavy frameworks (React/Vue) bring complex dependencies, while pure HTML/JS lacks modern interaction capabilities. GLLM Web offers a third path—built with native web tech (no frameworks) yet delivering modern UX. Its zero-dependency design makes it perfect for pairing with GLLM SSE backends.

Tech Stack: Vite (build tool), native TypeScript, native DOM API, native CSS. This choice ensures small bundle size, fast loading, and easy maintenance. Protocol Support: Natively handles two stream formats—standard OpenAI stream (via choices field) and GLLM's agent payloads (via type field), enabling both general chat and agent-specific features.

1. Dynamic Reasoning Display: Recognizes start_reasoning/end_reasoning markers, wraps reasoning in foldable components (keeps interface clean). 2. Agent Action Tracking: Generates badges for tool calls, intercepts CLI/MCP resource calls, allows checking execution details. 3. Glassmorphism UI: Dark theme with HSL deep slate colors, responsive flexbox layout, Apple-style staggered loading animations (reduces visual fatigue).

Prerequisites: Node.js/npm (use nvm), running GLLM backend. Steps: 1. Start GLLM backend: cd /path/to/gllm && make build && ./dist/gllm serve -p 8080. 2. Run front-end: npm install && npm run dev (access at http://localhost:5173).

GLLM Web is ideal for: 1. Agent Development Debugging: Real-time observation of tool calls, reasoning, and responses. 2. Lightweight Deployment: Zero dependencies suit edge/embedded systems. 3. Customization: Clear code structure (no framework) allows easy modifications. 4. Learning: A good example for SSE handling, native TS, and modern CSS.

GLLM Web demonstrates how to build a full-featured LLM client with minimal tech stack. It proves 'returning to native' is a viable (even better) option in some cases. For GLLM users, it's a ready-to-use interface; for web devs, it's a sample of balancing complexity and functionality.