infer-lite

Minimum viable architecture: stdlib-only LLM gateway.
No Docker. No LiteLLM. One Python file, one provider catalog, one HTTP server.

---

Quickstart

# 1. Install
git clone https://github.com/bradAGI/infer-lite && cd infer-lite
# or:  npx skills add bradagi/infer-lite

# 2. Add free keys
./infer keys add gemini AIza...
./infer keys add nvidia nvapi-...
./infer keys add openrouter sk-or-...

# 3. Boot the server (foreground; Ctrl-C to stop)
./infer serve

# 4. Verify what works for your account (in another terminal)
./infer probe

# 5. Wire your coding agents (any combination)
./infer wire cc | opencode | pi | openclaw | hermes

# 6. (optional) Make the command run from anywhere
./infer install            # symlinks ~/.local/bin/infer → this script
infer doctor               # works from any directory

`./infer install` detects your shell (bash / zsh / fish on macOS and Linux) and prints the right `~/.zshrc` / `~/.bashrc` / `~/.config/fish/config.fish` line if `~/.local/bin` isn't on `$PATH`. `./infer uninstall` removes the symlink. Pass `--name ` to install under a different command name.

**Permission denied?** If `./infer` errors with "Permission denied" right after clone or skill install, the executable bit didn't survive the transfer. Run `chmod +x infer` once and you're set.

Use it

curl http://localhost:4000/v1/chat/completions