Playground
Here I'm just experimenting with ideas/technologies that seem interesting. Mostly for fun.
llm.log
A local proxy that logs all your LLM API calls: tokens, costs, prompts, responses. Single binary, zero config, runs entirely on your machine.
Awsom
A chat application that runs language models entirely in the browser using WebLLM (+WebGPU) and WASM. No server required, everything runs client-side, keeping your conversations private and enabling offline use. Note: not all browsers are supported.
llm.log
A local proxy that sits between your apps and LLM APIs. It intercepts requests, extracts token usage, calculates costs, and stores everything in SQLite, without changing a single line of code.
• Zero code changes: works via HTTPS_PROXY, picked up by most apps automatically. Supports OpenAI, Anthropic, and OpenRouter.
• Real costs: auto-updated pricing for 780+ models, cache token breakdowns. Claude Code aware, so you can see what you'd pay on API keys vs subscription.
• TUI dashboard with overview, charts, cost breakdown, and request inspector. Export to CSV, JSON, or JSONL for analysis.
• Single binary, pure Go, no dependencies. Minimal overhead: logging is async and never blocks your requests.