A single-file, dependency-free gateway with per-key quotas and a tamper-evident usage log. Ollama has no auth — voidllm gives it one.
# 1. grab it — one file, no dependencies curl -O https://voidllm-cli.vercel.app/voidllm.py # 2. mint yourself a key (printed once) python3 voidllm.py keygen me # 3. run the gateway (loopback:8111, in front of your Ollama) python3 voidllm.py serve # 4. use it like any OpenAI endpoint curl http://127.0.0.1:8111/v1/chat/completions \ -H "Authorization: Bearer vllm-..." \ -d '{"model":"qwen2.5-coder:1.5b","messages":[{"role":"user","content":"hi"}]}'
Pure Python standard library. No pip, no Node, no Docker, no database. Copy it and run it.
Every request is appended to an HMAC hash-chained ledger. Edit history and usage --verify catches it.
Scope each key to specific models, paths, a daily quota, and an expiry. Revoke one without touching the rest.
Missing, unknown, expired, out-of-scope, or over-quota → rejected. If config can't load, nothing is served.
Optionally require a second, different key (X-Void-Key-2) on sensitive models or paths — no single key can call them alone.
No SaaS, no telemetry, no account. It runs on your machine, in front of your Ollama, and answers only to keys you mint.
Source-available under PolyForm Noncommercial 1.0.0. Free for personal, hobby, research, education, nonprofit, and government use. Using it in a for-profit business? Buy a commercial license — one time, per company. Launch sale: $10 for 10 days.