llmprof (npx launcher)

pprof for your LLM context - run it with no Python install.

Try the live dashboard in your browser - no install, a real recorded session.

npx llmprof up

This is a thin launcher for the llmprof profiler. It bootstraps uv - a single static binary that provisions its own Python - and runs the real llmprof package. You do not need Python, pip, or a virtualenv.

Usage

npx llmprof up                 # start the profiling proxy on http://localhost:4000
npx llmprof up --port 4100     # any llmprof argument is forwarded
npx llmprof traces             # show recent captured calls

Then point your LLM client's base URL at http://localhost:4000/v1 and open the dashboard at http://localhost:4000. Full docs: https://luthrag.github.io/llmprof.

How it works

1. If uv is already on your PATH, it is used as-is.
2. Otherwise the matching uv static binary is downloaded once (checksum verified) and cached under ~/.cache/llmprof.
3. uv tool run --from llmprof llmprof <args> runs the profiler, provisioning a managed Python the first time if your machine has none.

Environment

| Variable | Purpose | | --- | --- | | LLMPROF_SPEC | Package spec to run (default llmprof==<launcher version>); set to a path or git+https://... URL for a dev build. | | LLMPROF_CACHE_DIR | Where the uv binary is cached (default ~/.cache/llmprof). | | LLMPROF_UV | Path to an existing uv binary to use directly. |

Prefer Python? pipx install llmprof works too - see the docs.

License

MIT (c) Gaurav Luthra