Turn your machine into a node on a peer‑to‑peer inference marketplace.
PyrusLLM installs over P2P, updates itself over OTA, and exposes an OpenAI‑compatible endpoint to a swarm of other nodes. Each node announces a signed manifest — what it serves, what it charges, how loaded it is — and every answer that comes back carries provenance: which machine produced it, how fast, and what it cost.
What PyrusLLM is, how the manifest, the swarm and the wallet fit together, and whether you want the CLI or the API.
Over npm or over Pear. First offline prompt, first gateway, first OpenAI‑shaped curl call in under five minutes.
prompt, serve, peers, send,
fetch, files, wallet — every flag, every
footgun called out.
Every route the gateway serves: inference, network, keys, economics, wallet, files, upstream. Grouped, not alphabetised.
An orchestrator splits a requirements doc into tickets, workers write the code against a node's own model, and a CI gate no worker can edit decides.
Point it at your own node — k16 or any machine you run — and fire real requests
from the browser. Safe routes execute live; anything that moves funds or revokes
access only ever generates a curl you run yourself.
Four papers on decentralized LLM inference, read closely and mapped against PyrusLLM's own design: what to copy, what to reject, and what to build first.