mirror of
https://github.com/hanzoai/ai.git
synced 2026-11-20 14:13:23 +00:00
prompt_tokens meant two things. The family and relay paths recorded the part of the prompt read fresh, the table path and the providers the whole prompt, so a cached enso-auto row read cache_read > prompt. No path recorded a cache write: Anthropic's cache_creation_input_tokens was folded into the fresh read and billed at the input rate, though a write is priced above it, and OpenRouter's cache_write_tokens was not read at all. cache_write_tokens was 0 on every row. A record's prompt is now every input token, with the cache reads and writes parts of it, whatever shape the upstream reported: OpenAI's cached_tokens inside prompt_tokens, OpenRouter's cache_write_tokens beside them, Anthropic's two counts beside input_tokens. Pricing bills the input rate on the part read fresh and each cache part at its own rate: a family SKU's cache_read and cache_write (a resold model's input_cache_read and input_cache_write plus the vendor's fee), else the input rate. The Anthropic provider and the OpenAI adapter over Anthropic now read both cache counts, and the providers' settles pass them.
63 lines
2.7 KiB
Go
63 lines
2.7 KiB
Go
// Copyright 2026 Hanzo AI Inc. All Rights Reserved.
|
||
//
|
||
// Licensed under the Apache License, Version 2.0 (the "License");
|
||
// you may not use this file except in compliance with the License.
|
||
// You may obtain a copy of the License at
|
||
//
|
||
// http://www.apache.org/licenses/LICENSE-2.0
|
||
//
|
||
// Unless required by applicable law or agreed to in writing, software
|
||
// distributed under the License is distributed on an "AS IS" BASIS,
|
||
// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
||
// See the License for the specific language governing permissions and
|
||
// limitations under the License.
|
||
|
||
package controllers
|
||
|
||
import (
|
||
"testing"
|
||
|
||
"github.com/hanzoai/decimal"
|
||
)
|
||
|
||
// A stated cost of zero or less never credits the caller: a variable SKU bills its
|
||
// ceiling.
|
||
func TestResaleIgnoresNonPositiveCost(t *testing.T) {
|
||
tier := zenTier{MaxCtx: 1 << 20, In: decimal.New(3, 0), Out: decimal.New(15, 0)}
|
||
m := zenModel{Margin: decimal.New(12, 1), Variable: true, Base: tier, Tiers: []zenTier{tier}}
|
||
ceiling := m.bill(nil, tokens{fresh: 1000, cached: 0, completion: 100})
|
||
if ceiling <= 0 {
|
||
t.Fatalf("ceiling = %d, want a positive charge", ceiling)
|
||
}
|
||
for _, c := range []int64{0, -600_000_000} {
|
||
c := c
|
||
if nano := m.bill(&c, tokens{fresh: 1000, cached: 0, completion: 100}); nano != ceiling {
|
||
t.Fatalf("cost %d: nano=%d, want the ceiling %d", c, nano, ceiling)
|
||
}
|
||
}
|
||
}
|
||
|
||
// A SKU we resell bills its tokens at retail, and never less than what the vendor
|
||
// says the call cost us times the margin: a cache write or a search the token rates
|
||
// do not price is still paid for. A SKU we do not resell bills its tokens alone.
|
||
func TestAResoldCallNeverBillsBelowItsStatedCost(t *testing.T) {
|
||
tier := zenTier{MaxCtx: 1 << 20, In: decimal.New(12, 2), Out: decimal.New(6, 1)} // $0.12 and $0.60 per MTok
|
||
m := zenModel{Margin: decimal.New(12, 1), Base: tier, Tiers: []zenTier{tier}}
|
||
retail := m.bill(nil, tokens{fresh: 1000, cached: 0, completion: 100}) // 1000 × 0.12e-6 + 100 × 0.6e-6 = $0.00018
|
||
if retail != 180_000 {
|
||
t.Fatalf("retail = %d nano, want 180000", retail)
|
||
}
|
||
under := int64(100_000) // the vendor charged $0.0001: retail covers it
|
||
if got := m.bill(&under, tokens{fresh: 1000, cached: 0, completion: 100}); got != retail {
|
||
t.Errorf("a cost retail covers billed %d, want retail %d", got, retail)
|
||
}
|
||
over := int64(200_000) // a cache write made it $0.0002: billed at 1.2 × that
|
||
if got := m.bill(&over, tokens{fresh: 1000, cached: 0, completion: 100}); got != 240_000 {
|
||
t.Errorf("a cost above retail billed %d, want 240000", got)
|
||
}
|
||
own := zenModel{Base: tier, Tiers: []zenTier{tier}}
|
||
if got := own.bill(&over, tokens{fresh: 1000, cached: 0, completion: 100}); got != retail {
|
||
t.Errorf("a SKU we do not resell billed %d, want its tokens %d", got, retail)
|
||
}
|
||
}
|