huge optimization for token usage with anthropic

This commit is contained in:
Dax Raad
2025-06-20 22:43:04 -04:00
parent 59f0004d34
commit 1684042fb6
2 changed files with 35 additions and 21 deletions

View File

@@ -1,24 +1,25 @@
import type { CoreMessage } from "ai"
import type { CoreMessage, LanguageModelV1Prompt } from "ai"
import { unique } from "remeda"
export namespace ProviderTransform {
export function message(
msg: CoreMessage,
index: number,
msgs: LanguageModelV1Prompt,
providerID: string,
modelID: string,
) {
if (
(providerID === "anthropic" || modelID.includes("anthropic")) &&
index < 4
) {
msg.providerOptions = {
...msg.providerOptions,
anthropic: {
cacheControl: { type: "ephemeral" },
},
if (providerID === "anthropic" || modelID.includes("anthropic")) {
const system = msgs.filter((msg) => msg.role === "system").slice(0, 2)
const final = msgs.filter((msg) => msg.role !== "system").slice(-2)
for (const msg of unique([...system, ...final])) {
msg.providerMetadata = {
...msg.providerMetadata,
anthropic: {
cacheControl: { type: "ephemeral" },
},
}
}
}
return msg
return msgs
}
}