Chris Han

GlyphStream

2026 - present

glyphstream

openai-api-bridge

GlyphStream is a lightweight web interface for AI inference - chatting with LLMs and image/video generation. It supports inference with any OpenAI-compatible endpoint, for example local services like llama-server as well as remote api providers like OpenRouter.

GlyphStream supports comparison workflows, with one prompt sent to multiple text, image, or video models, with results presented side by side for comparison. It has many image and video generation features, such as image-to-image, prompt enhancement, aspect ratio controls, and gallery features to manage generated images/videos. It supports custom models with system prompts as well as character-style models with avatars and texting-style responses.

Glyphstream supports a variety of tooling to enhance AI workflows, such as web searching, a Python sandbox, MCP servers, and agent skills. It supports passkey and OAuth authentication, isolated user accounts, and a privacy lock for PWA installs.

GlyphStream is written in Svelte with SvelteKit.

openai-api-bridge is a companion project which acts as an API gateway for one or more backend AI services, presenting them in an OpenAI-compatible format for GlyphStream or other OpenAI-compatible applications. It supports ComfyUI, Venice.ai, OpenRouter, ImageRouter, fal.ai, and OpenAI passthrough.

openai-api-bridge is written in Python.