feat(provider): load catalog request variants
This commit is contained in:
parent
d00d274192
commit
75ab5ceae9
11 changed files with 207 additions and 243 deletions
|
|
@ -69,6 +69,7 @@ export const Model = Schema.Struct({
|
|||
temperature: Schema.Boolean,
|
||||
tool_call: Schema.Boolean,
|
||||
reasoning_options: Schema.optional(Schema.Array(ReasoningOption)),
|
||||
variants: Schema.optional(Schema.Record(Schema.String, Schema.Record(Schema.String, Schema.MutableJson))),
|
||||
interleaved: Schema.optional(
|
||||
Schema.Union([
|
||||
Schema.Literal(true),
|
||||
|
|
@ -109,7 +110,13 @@ export const Model = Schema.Struct({
|
|||
),
|
||||
status: Schema.optional(CatalogModelStatus),
|
||||
provider: Schema.optional(
|
||||
Schema.Struct({ npm: Schema.optional(Schema.String), api: Schema.optional(Schema.String) }),
|
||||
Schema.Struct({
|
||||
npm: Schema.optional(Schema.String),
|
||||
api: Schema.optional(Schema.String),
|
||||
variant: Schema.optional(Schema.String),
|
||||
body: Schema.optional(Schema.Record(Schema.String, Schema.MutableJson)),
|
||||
headers: Schema.optional(Schema.Record(Schema.String, Schema.String)),
|
||||
}),
|
||||
),
|
||||
})
|
||||
export type Model = Schema.Schema.Type<typeof Model>
|
||||
|
|
|
|||
|
|
@ -3,6 +3,7 @@ import type { ModelV2Info } from "@opencode-ai/sdk/v2/types"
|
|||
import { Effect, Stream } from "effect"
|
||||
import { EventV2 } from "../event"
|
||||
import { ModelsDev } from "../models-dev"
|
||||
import { ModelV2 } from "../model"
|
||||
import { ProviderV2 } from "../provider"
|
||||
|
||||
function released(date: string) {
|
||||
|
|
@ -102,7 +103,11 @@ function applyModel(
|
|||
input: [...(model.modalities?.input ?? [])],
|
||||
output: [...(model.modalities?.output ?? [])],
|
||||
}
|
||||
draft.variants = []
|
||||
draft.variants = Object.entries(model.variants ?? {}).map(([id, body]) => ({
|
||||
id: ModelV2.VariantID.make(id),
|
||||
headers: {},
|
||||
body: { ...body },
|
||||
}))
|
||||
draft.time.released = released(model.release_date)
|
||||
draft.cost = input.cost ?? cost(model.cost)
|
||||
draft.status = model.status ?? "active"
|
||||
|
|
@ -112,6 +117,9 @@ function applyModel(
|
|||
input: model.limit.input,
|
||||
output: model.limit.output,
|
||||
}
|
||||
if (model.provider?.variant !== undefined) draft.request.variant = ModelV2.VariantID.make(model.provider.variant)
|
||||
Object.assign(draft.request.headers, model.provider?.headers ?? {})
|
||||
Object.assign(draft.request.body, model.provider?.body ?? {})
|
||||
Object.assign(draft.request.headers, input.request?.headers ?? {})
|
||||
Object.assign(draft.request.body, input.request?.body ?? {})
|
||||
}
|
||||
|
|
|
|||
|
|
@ -48,8 +48,13 @@ describe("ModelsDevPlugin", () => {
|
|||
release_date: "2026-01-01",
|
||||
attachment: false,
|
||||
reasoning: true,
|
||||
reasoning_options: [{ type: "toggle" }],
|
||||
temperature: true,
|
||||
tool_call: true,
|
||||
variants: {
|
||||
none: { thinking: { type: "disabled" } },
|
||||
thinking: { thinking: { type: "adaptive" } },
|
||||
},
|
||||
cost: {
|
||||
input: 2.5,
|
||||
output: 15,
|
||||
|
|
@ -64,6 +69,11 @@ describe("ModelsDevPlugin", () => {
|
|||
context_over_200k: { input: 5, output: 22.5, cache_read: 0.5 },
|
||||
},
|
||||
limit: { context: 1_050_000, input: 922_000, output: 128_000 },
|
||||
provider: {
|
||||
variant: "thinking",
|
||||
body: { thinking: { type: "adaptive" } },
|
||||
headers: { "x-model": "gpt-5.4" },
|
||||
},
|
||||
experimental: {
|
||||
modes: {
|
||||
fast: {
|
||||
|
|
@ -93,18 +103,29 @@ describe("ModelsDevPlugin", () => {
|
|||
const base = yield* catalog.model.get(providerID, ModelV2.ID.make("gpt-5.4"))
|
||||
const fast = yield* catalog.model.get(providerID, ModelV2.ID.make("gpt-5.4-fast"))
|
||||
|
||||
expect(base?.variants).toEqual([])
|
||||
expect(base?.request.body).toEqual({})
|
||||
expect(base?.variants).toEqual([
|
||||
{ id: ModelV2.VariantID.make("none"), headers: {}, body: { thinking: { type: "disabled" } } },
|
||||
{ id: ModelV2.VariantID.make("thinking"), headers: {}, body: { thinking: { type: "adaptive" } } },
|
||||
])
|
||||
expect(base?.request).toEqual({
|
||||
variant: ModelV2.VariantID.make("thinking"),
|
||||
headers: { "x-model": "gpt-5.4" },
|
||||
body: { thinking: { type: "adaptive" } },
|
||||
})
|
||||
expect(fast).toMatchObject({
|
||||
id: "gpt-5.4-fast",
|
||||
providerID: "acme",
|
||||
name: "GPT-5.4 Fast",
|
||||
api: { id: "gpt-5.4" },
|
||||
request: {
|
||||
headers: { "x-mode": "fast" },
|
||||
body: { service_tier: "priority" },
|
||||
variant: ModelV2.VariantID.make("thinking"),
|
||||
headers: { "x-model": "gpt-5.4", "x-mode": "fast" },
|
||||
body: { thinking: { type: "adaptive" }, service_tier: "priority" },
|
||||
},
|
||||
variants: [],
|
||||
variants: [
|
||||
{ id: ModelV2.VariantID.make("none"), headers: {}, body: { thinking: { type: "disabled" } } },
|
||||
{ id: ModelV2.VariantID.make("thinking"), headers: {}, body: { thinking: { type: "adaptive" } } },
|
||||
],
|
||||
})
|
||||
expect(fast?.cost).toEqual([
|
||||
{ input: 5, output: 30, cache: { read: 0.5, write: 0 } },
|
||||
|
|
|
|||
|
|
@ -22,7 +22,7 @@ type Api =
|
|||
}
|
||||
| { readonly type: "native"; readonly url?: string; readonly settings: Record<string, unknown> }
|
||||
|
||||
const model = (api: Api, variants: ModelV2.Info["variants"] = []) =>
|
||||
const model = (api: Api, variants: ModelV2.Info["variants"] = [], variant?: ModelV2.VariantID) =>
|
||||
ModelV2.Info.make({
|
||||
id: ModelV2.ID.make("test-model"),
|
||||
providerID: ProviderV2.ID.make("test-provider"),
|
||||
|
|
@ -32,6 +32,7 @@ const model = (api: Api, variants: ModelV2.Info["variants"] = []) =>
|
|||
request: {
|
||||
headers: { "x-test": "header" },
|
||||
body: { apiKey: "secret", custom_extension: { enabled: true } },
|
||||
...(variant ? { variant } : {}),
|
||||
},
|
||||
variants,
|
||||
time: { released: 0 },
|
||||
|
|
@ -175,6 +176,44 @@ describe("SessionRunnerModel", () => {
|
|||
}),
|
||||
)
|
||||
|
||||
it.effect("uses the catalog default Session variant", () =>
|
||||
Effect.gen(function* () {
|
||||
const catalog = model(
|
||||
{ type: "aisdk", package: "@ai-sdk/anthropic", url: "https://anthropic.example/v1" },
|
||||
[
|
||||
{
|
||||
id: ModelV2.VariantID.make("none"),
|
||||
headers: {},
|
||||
body: { thinking: { type: "disabled" } },
|
||||
},
|
||||
{
|
||||
id: ModelV2.VariantID.make("thinking"),
|
||||
headers: {},
|
||||
body: { thinking: { type: "adaptive" } },
|
||||
},
|
||||
],
|
||||
ModelV2.VariantID.make("thinking"),
|
||||
)
|
||||
const session = SessionV2.Info.make({
|
||||
id: SessionV2.ID.make("ses_default_variant"),
|
||||
projectID: ProjectV2.ID.global,
|
||||
title: "test",
|
||||
model: { id: catalog.id, providerID: catalog.providerID },
|
||||
cost: 0,
|
||||
tokens: { input: 0, output: 0, reasoning: 0, cache: { read: 0, write: 0 } },
|
||||
time: { created: DateTime.makeUnsafe(0), updated: DateTime.makeUnsafe(0) },
|
||||
location: { directory: AbsolutePath.make("/project") },
|
||||
})
|
||||
|
||||
const resolved = yield* SessionRunnerModel.resolve(session, catalog)
|
||||
|
||||
expect(resolved.route.defaults.http?.body).toEqual({
|
||||
custom_extension: { enabled: true },
|
||||
thinking: { type: "adaptive" },
|
||||
})
|
||||
}),
|
||||
)
|
||||
|
||||
it.effect("rejects an explicit unavailable Session variant during model resolution", () =>
|
||||
Effect.gen(function* () {
|
||||
const catalog = model({ type: "aisdk", package: "@ai-sdk/openai", url: "https://openai.example/v1" })
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue