Released · v0.1.0
LLM Stream Debugger
A desktop proxy that shows every LLM streaming call live: tokens, reasoning, tool calls and stalls.
Run it between your code and any LLM API. It decodes each streaming response as it arrives (text, reasoning, tool calls assembled from partial JSON, token usage and timing) while passing the bytes through untouched.
- Platforms
- macOS, Windows, Linux
- Version
- 0.1.0
- Built with
- Rust, Tauri 2, tokio, hyper, Next.js, Cloudflare Workers
What it does
- 15 built-in providers, including OpenAI, Anthropic, Gemini, OpenRouter, Groq, Mistral and Ollama
- Time to first byte and token, tokens per second, and a timeline that highlights stalls
- API keys masked in the interface and never written to disk
- A headless command-line mode that shares the same core
How it’s built
- Proxy core on tokio and hyper with SSE and NDJSON framing and a decoder per provider
- CI builds installers for every platform, uploads them to R2 with a checksum manifest, and a Worker serves the latest version
- A 50-second launch video made in Remotion