aweshare Community Hub Opens for Beta: 10 Consumer Spots, Unlimited Producers

aweshare works by putting your idle AI capacity behind a standard facade: you run a lightweight producer on your own machine, where the upstream keys stay in secrets.json and never leave; friends point ordinary OpenAI / Anthropic SDKs at a hub, whose requests travel back to your machine over an outbound WebSocket — keys are injected only at call time. The whole stack is self-hostable — aweshare hub serve is one command away. Its only real threshold is that relay server: it needs a public IP, TLS, and someone watching the process. That’s the one link the community hub replaces: that machine already exists.

Starting today, the community hub at https://aweshare.wehuman.top is taking its first batch of testers: 10 consumer spots, first come first served — and no cap on producers, welcome anytime.

To be clear about what this does and doesn’t replace: the hub takes over the “relay” link only. Keys and models still leave from your machine — for others to use your models, your machine must stay online with the producer running. That’s aweshare by design: keys don’t travel.

How to Apply

Just send an email:

To: peng@wehuman.top
Include: who you are; which role you want (producer / consumer); what you plan to share — Ollama / vLLM on an idle GPU, or an OpenAI Chat / OpenAI Responses / Anthropic-compatible API account (base URL + API key counts) — or which kinds of models you’d like to call.

I’ll reply with a one-time invite code (asi_...). Consumer spots go fast — later emails automatically join the waitlist for the next round; producer spots are uncapped.

Once You Have an Invite

Never used aweshare? Ask your agent to install it:

“Read https://github.com/wehuman01/aweshare/blob/main/README.ai.md and follow it to install and configure aweshare.”

Doing it yourself, there are two paths. Producer (share your capacity):

npm install -g aweshare                                        # Node ≥ 22
aweshare producer join --hub https://aweshare.wehuman.top --code asi_...
# Edit ~/.aweshare/config.toml to register backends / offerings; put keys in secrets.json — they never leave this machine
aweshare producer doctor                 # fix the first FAIL, rerun until green
aweshare producer start --background     # your terminal, not your agent's

Consumer (use models others share):

npm install -g aweshare
aweshare consumer join --hub https://aweshare.wehuman.top --code asi_...
# The asc_ token prints once — save it; no Node? A single curl redeems too.
export OPENAI_BASE_URL=https://aweshare.wehuman.top/v1
export OPENAI_API_KEY=asc_...

From there, Claude Code, Codex, OpenCode, anything that speaks OpenAI Chat / Anthropic Messages / Responses points straight at the hub — just use the alias (e.g. peng/qwen2.5.7b) as the model name. For juggling multiple providers day to day, aweswitch manages profiles nicely.

What Testers Get

No server to run. Public entry, TLS, relaying, usage accounting — the hub handles all of it; you just run the producer locally.
Keys stay home. Upstream keys live only in your machine’s secrets.json (mode 600), injected by the local agent at forward time — the hub never sees them.
Leave anytime. Your config and key originals live on your machine throughout; nothing is locked in. Done? aweshare producer stop, ask me to revoke the invite code, and the spot goes to the waitlist.

What I’ll Try to Protect (and What You Should Weigh)

The hub relays traffic but never touches your keys — that’s an architectural guarantee, not a verbal promise. The honest caveats are elsewhere:

For producers: relaying a personal subscription-based API key to third parties may violate that upstream provider’s terms of service. Self-hosted open models (Ollama / vLLM) have no such problem; if unsure, don’t share that backend. Consequences fall on the publisher.
For consumers: everything you send passes through the hub in plaintext — the operator (me) can see prompts and responses flowing through, and uses them for usage accounting; your token is stored on the hub as a hash only. Treat shared models as “operated in my line of sight,” not as a private channel.
For everyone: if you don’t trust a server I operate, don’t use it — run aweshare hub serve on any machine you control. The code is open source, every line.

What Backends Are Welcome

What you want to share	Welcome on the hub?
Local open models (Ollama / vLLM on an idle GPU)	Most welcome — clean, no terms-of-service worries
API accounts (OpenAI Chat / OpenAI Responses / Anthropic-compatible)	Yes — keys stay on your machine
Personal subscription keys	Technically yes, but ToS risk is yours — skip it if unsure

One alias speaks one protocol, and the hub does no translation: callers pick the endpoint matching the alias’s protocol, and a wrong endpoint gets a clear error pointing the way — never silently garbled output.

Compliance and Disclaimer

Some things belong in a terms section, not just a one-line FAQ note:

aweshare is relay software: it cannot and does not judge whether you are allowed to share a given upstream key or subscription — that question is between you and the upstream provider. Being able to call an API yourself does not mean you may re-provide it to third parties.
The two kinds of backends therefore sit in very different positions: self-hosted open models (Ollama / vLLM on your own GPU) share your own hardware and open weights — no upstream account involved, clean; third-party API accounts and personal subscription keys (coding plans included) — read the upstream’s terms first (account rules, subscription and seat limits, forwarding, commercial-use clauses); re-providing to third parties likely violates them. When in doubt, don’t share.
The consequences of sharing (key revocation, account suspension or termination by the upstream) fall on the producer; the hub operator (me) is responsible for operating the hub lawfully and for informing every consumer of the “traffic transits the hub in plaintext” boundary.
The software is provided “as is” under the proprietary license (free to use and self-host, no redistribution), without warranty of any kind; the authors and contributors are not liable for how aweshare is used or for any damage arising from sharing access through it.

You Might Be Wondering

Why 10? The hub’s global admission cap currently applies to consumers — 10 of them; producers are uncapped. If you want to consume, come early; if you want to produce, come anytime.

Is there an availability guarantee? No — the hub is run by me personally. If it ever shuts down, producers lose only the connection itself (config and keys were always local); consumers find another hub or self-host one. If that day comes, I’ll announce it.

Calls return 503. The producer behind that alias is offline; all you can do is wait — check STATUS in consumer list. And a note for producers: around-the-clock availability needs an always-on machine. A laptop that powers off at night won’t serve through the night.

Lost token. A consumer’s asc_ token can’t be recovered (the hub stores hashes only) — say the word and I’ll issue a fresh invite.

Someone’s abusing what I share. Remove the offering from your config and run aweshare producer reload — it delists instantly, no restart, catalog updates right away. For anything worse, contact me; revoking an identity is one command.

Rate limited. Per-model guardrails (maxConcurrencyPerUser, maxConcurrentUsers, daily token budget) are set by each publisher and visible in consumer list; the hub adds its own rate limits. When you hit them, the 429 explains why.

Build It Together

This hub isn’t meant to stay just a relay server. I’m inviting users with ideas to co-create something more useful:

Share your idle compute. The GPU in the corner, the subscription you can’t max out — hook it up and it becomes an offering on the hub. Producer spots are uncapped; idle capacity is always welcome.
Sustainability is being explored. Down the road: ad revenue, or rewarding producers with credits the way compute-sharing communities (like 科研通) do — the compute you contribute earns points, and points may become redeemable down the line. All still ideas — come help define them.
Bring your ideas. Which models are missing, how to set caps fairly, what a credit should be worth — email peng@wehuman.top or open a GitHub issue.

Apply Now

10 consumer spots, first come first served; producers uncapped, welcome anytime.

Email peng@wehuman.top — who you are, whether you want to share or consume, and what backend you’ll bring. Bugs in aweshare itself go to GitHub issues.