Repartee is a pay-per-token chat API on our own GPUs. It speaks the request your code already sends: change the base address and the key. No sales call, no contract. Adult conversation is fine; minors and non-consent are refused in code, on the way in and on the way out.
Nothing opens in this browser — this one takes a job and does the work on our own graphics cards in Oregon. The free allowance below is how you try it: 200 units, no card and no payment, on the key you are issued.
Runs on this computer today. It opens to the world when two pieces are in place — see below.
The problem
Hosted model providers close accounts that run adult companionship, and self-hosting means buying GPUs and running them. Repartee is the third option: an API that says yes to adults and no to the things nobody should say yes to.
Same chat-completions request and answer most SDKs already use. Point the base address here and keep your code.
The log holds token counts, never a message or an answer. Your users' conversations are theirs.
Minors, non-consent and the prohibited list are refused in code, both ways. A refusal is free and names its category.
Repartee answers this computer only until it runs on GPU capacity whose terms allow adult content and has a public https address. It refuses to open without both.
How it works
One click. 200,000 tokens free.
Point your SDK at Repartee and give it the key.
Packs in bitcoin, added to your key as soon as the payment confirms.
What it will never do
A developer can rewrite a system prompt. Nobody can rewrite the floor.
Pricing
A unit is 1,000 tokens, prompt plus answer, counted by the model server itself and rounded up per request.
200,000 tokens to build against.
10 million tokens for a launched app.
100 million tokens for an app with real users.
Paid in bitcoin through our own checkout; the units are added to your key as soon as the payment confirms. Your licence key is emailed to you within two working days of payment.
Checkout is our own bitcoin server, not a card processor. The units you buy are added to your key: buy from the Buy more box inside the program and they arrive on their own once the payment confirms, or buy here and paste the invoice number from your receipt into that box.
Questions
The chat-completions request and answer, including token usage. Streaming is not offered yet; send stream false.
Anything involving a minor, non-consent, or the rest of the line's prohibited list, in any message or in the answer. It is conservative: it refuses some innocent sentences, says which word did it, and charges nothing.
No conversation is ever written down. The log is token counts, model, time and finish reason.
The ones GET /v1/models lists. Today that is one mid-size model; more as the GPUs allow.
A key. One line of text, issued to you and shown exactly once — only its hash is kept here, so it cannot be looked up later and we could not tell you what yours is. Every key arrives with a free allowance already on it, which is the tier at the top of the pricing above, and the work starts the first time you send a job with that key on it.
Each job is metered in the unit named in the pricing, and a unit that fails — a file that will not read, a row the model held back — is not charged and is named in the job report rather than silently skipped. When an allowance runs out the next job is refused before any work starts, with the number it used printed on the refusal.
More units are bought in bitcoin. The checkout is a bitcoin server we run ourselves. Buy from the Buy more box inside the program and the units are added to your key on their own as soon as the payment confirms; buy from the plan buttons above and paste the invoice number from your receipt into that same box. Each payment adds its units once, to one key.
Make a key, change your SDK's base address, and send the conversation you already have.