feat: add support for fast modes for claude and gpt models (that support it) (#21706)
This commit is contained in:
@@ -376,9 +376,9 @@ export async function CodexAuthPlugin(input: PluginInput): Promise<Hooks> {
|
||||
"gpt-5.4",
|
||||
"gpt-5.4-mini",
|
||||
])
|
||||
for (const modelId of Object.keys(provider.models)) {
|
||||
for (const [modelId, model] of Object.entries(provider.models)) {
|
||||
if (modelId.includes("codex")) continue
|
||||
if (allowedModels.has(modelId)) continue
|
||||
if (allowedModels.has(model.api.id)) continue
|
||||
delete provider.models[modelId]
|
||||
}
|
||||
|
||||
|
||||
@@ -22,6 +22,27 @@ export namespace ModelsDev {
|
||||
)
|
||||
const ttl = 5 * 60 * 1000
|
||||
|
||||
type JsonValue = string | number | boolean | null | { [key: string]: JsonValue } | JsonValue[]
|
||||
|
||||
const JsonValue: z.ZodType<JsonValue> = z.lazy(() =>
|
||||
z.union([z.string(), z.number(), z.boolean(), z.null(), z.array(JsonValue), z.record(z.string(), JsonValue)]),
|
||||
)
|
||||
|
||||
const Cost = z.object({
|
||||
input: z.number(),
|
||||
output: z.number(),
|
||||
cache_read: z.number().optional(),
|
||||
cache_write: z.number().optional(),
|
||||
context_over_200k: z
|
||||
.object({
|
||||
input: z.number(),
|
||||
output: z.number(),
|
||||
cache_read: z.number().optional(),
|
||||
cache_write: z.number().optional(),
|
||||
})
|
||||
.optional(),
|
||||
})
|
||||
|
||||
export const Model = z.object({
|
||||
id: z.string(),
|
||||
name: z.string(),
|
||||
@@ -41,22 +62,7 @@ export namespace ModelsDev {
|
||||
.strict(),
|
||||
])
|
||||
.optional(),
|
||||
cost: z
|
||||
.object({
|
||||
input: z.number(),
|
||||
output: z.number(),
|
||||
cache_read: z.number().optional(),
|
||||
cache_write: z.number().optional(),
|
||||
context_over_200k: z
|
||||
.object({
|
||||
input: z.number(),
|
||||
output: z.number(),
|
||||
cache_read: z.number().optional(),
|
||||
cache_write: z.number().optional(),
|
||||
})
|
||||
.optional(),
|
||||
})
|
||||
.optional(),
|
||||
cost: Cost.optional(),
|
||||
limit: z.object({
|
||||
context: z.number(),
|
||||
input: z.number().optional(),
|
||||
@@ -68,7 +74,24 @@ export namespace ModelsDev {
|
||||
output: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
|
||||
})
|
||||
.optional(),
|
||||
experimental: z.boolean().optional(),
|
||||
experimental: z
|
||||
.object({
|
||||
modes: z
|
||||
.record(
|
||||
z.string(),
|
||||
z.object({
|
||||
cost: Cost.optional(),
|
||||
provider: z
|
||||
.object({
|
||||
body: z.record(z.string(), JsonValue).optional(),
|
||||
headers: z.record(z.string(), z.string()).optional(),
|
||||
})
|
||||
.optional(),
|
||||
}),
|
||||
)
|
||||
.optional(),
|
||||
})
|
||||
.optional(),
|
||||
status: z.enum(["alpha", "beta", "deprecated"]).optional(),
|
||||
provider: z.object({ npm: z.string().optional(), api: z.string().optional() }).optional(),
|
||||
})
|
||||
|
||||
@@ -926,6 +926,28 @@ export namespace Provider {
|
||||
|
||||
export class Service extends ServiceMap.Service<Service, Interface>()("@opencode/Provider") {}
|
||||
|
||||
function cost(c: ModelsDev.Model["cost"]): Model["cost"] {
|
||||
const result: Model["cost"] = {
|
||||
input: c?.input ?? 0,
|
||||
output: c?.output ?? 0,
|
||||
cache: {
|
||||
read: c?.cache_read ?? 0,
|
||||
write: c?.cache_write ?? 0,
|
||||
},
|
||||
}
|
||||
if (c?.context_over_200k) {
|
||||
result.experimentalOver200K = {
|
||||
cache: {
|
||||
read: c.context_over_200k.cache_read ?? 0,
|
||||
write: c.context_over_200k.cache_write ?? 0,
|
||||
},
|
||||
input: c.context_over_200k.input,
|
||||
output: c.context_over_200k.output,
|
||||
}
|
||||
}
|
||||
return result
|
||||
}
|
||||
|
||||
function fromModelsDevModel(provider: ModelsDev.Provider, model: ModelsDev.Model): Model {
|
||||
const m: Model = {
|
||||
id: ModelID.make(model.id),
|
||||
@@ -940,24 +962,7 @@ export namespace Provider {
|
||||
status: model.status ?? "active",
|
||||
headers: {},
|
||||
options: {},
|
||||
cost: {
|
||||
input: model.cost?.input ?? 0,
|
||||
output: model.cost?.output ?? 0,
|
||||
cache: {
|
||||
read: model.cost?.cache_read ?? 0,
|
||||
write: model.cost?.cache_write ?? 0,
|
||||
},
|
||||
experimentalOver200K: model.cost?.context_over_200k
|
||||
? {
|
||||
cache: {
|
||||
read: model.cost.context_over_200k.cache_read ?? 0,
|
||||
write: model.cost.context_over_200k.cache_write ?? 0,
|
||||
},
|
||||
input: model.cost.context_over_200k.input,
|
||||
output: model.cost.context_over_200k.output,
|
||||
}
|
||||
: undefined,
|
||||
},
|
||||
cost: cost(model.cost),
|
||||
limit: {
|
||||
context: model.limit.context,
|
||||
input: model.limit.input,
|
||||
@@ -994,13 +999,31 @@ export namespace Provider {
|
||||
}
|
||||
|
||||
export function fromModelsDevProvider(provider: ModelsDev.Provider): Info {
|
||||
const models: Record<string, Model> = {}
|
||||
for (const [key, model] of Object.entries(provider.models)) {
|
||||
models[key] = fromModelsDevModel(provider, model)
|
||||
for (const [mode, opts] of Object.entries(model.experimental?.modes ?? {})) {
|
||||
const id = `${model.id}-${mode}`
|
||||
const m = fromModelsDevModel(provider, model)
|
||||
m.id = ModelID.make(id)
|
||||
m.name = `${model.name} ${mode[0].toUpperCase()}${mode.slice(1)}`
|
||||
if (opts.cost) m.cost = mergeDeep(m.cost, cost(opts.cost))
|
||||
// convert body params to camelCase for ai sdk compatibility
|
||||
if (opts.provider?.body)
|
||||
m.options = Object.fromEntries(
|
||||
Object.entries(opts.provider.body).map(([k, v]) => [k.replace(/_([a-z])/g, (_, c) => c.toUpperCase()), v]),
|
||||
)
|
||||
if (opts.provider?.headers) m.headers = opts.provider.headers
|
||||
models[id] = m
|
||||
}
|
||||
}
|
||||
return {
|
||||
id: ProviderID.make(provider.id),
|
||||
source: "custom",
|
||||
name: provider.name,
|
||||
env: provider.env ?? [],
|
||||
options: {},
|
||||
models: mapValues(provider.models, (model) => fromModelsDevModel(provider, model)),
|
||||
models,
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
Reference in New Issue
Block a user