feat: unwrap provider namespaces to flat exports + barrel (#22760)
This commit is contained in:
Kit Langton
@@ -6,7 +6,7 @@ import { generateObject, streamObject, type ModelMessage } from "ai"
|
||||
import { Instance } from "../project/instance"
|
||||
import { Truncate } from "../tool"
|
||||
import { Auth } from "../auth"
|
||||
import { ProviderTransform } from "../provider/transform"
|
||||
import { ProviderTransform } from "../provider"
|
||||
|
||||
import PROMPT_GENERATE from "./generate.txt"
|
||||
import PROMPT_COMPACTION from "./prompt/compaction.txt"
|
||||
|
||||
@@ -18,7 +18,7 @@ import type {
|
||||
} from "@octokit/webhooks-types"
|
||||
import { UI } from "../ui"
|
||||
import { cmd } from "./cmd"
|
||||
import { ModelsDev } from "../../provider/models"
|
||||
import { ModelsDev } from "../../provider"
|
||||
import { Instance } from "@/project/instance"
|
||||
import { bootstrap } from "../bootstrap"
|
||||
import { SessionShare } from "@/share"
|
||||
|
||||
@@ -2,7 +2,7 @@ import type { Argv } from "yargs"
|
||||
import { Instance } from "../../project/instance"
|
||||
import { Provider } from "../../provider"
|
||||
import { ProviderID } from "../../provider/schema"
|
||||
import { ModelsDev } from "../../provider/models"
|
||||
import { ModelsDev } from "../../provider"
|
||||
import { cmd } from "./cmd"
|
||||
import { UI } from "../ui"
|
||||
import { EOL } from "os"
|
||||
|
||||
@@ -3,7 +3,7 @@ import { AppRuntime } from "../../effect/app-runtime"
|
||||
import { cmd } from "./cmd"
|
||||
import * as prompts from "@clack/prompts"
|
||||
import { UI } from "../ui"
|
||||
import { ModelsDev } from "../../provider/models"
|
||||
import { ModelsDev } from "../../provider"
|
||||
import { map, pipe, sortBy, values } from "remeda"
|
||||
import path from "path"
|
||||
import os from "os"
|
||||
|
||||
@@ -16,7 +16,7 @@ import { Storage } from "@/storage"
|
||||
import { Snapshot } from "@/snapshot"
|
||||
import { Plugin } from "@/plugin"
|
||||
import { Provider } from "@/provider"
|
||||
import { ProviderAuth } from "@/provider/auth"
|
||||
import { ProviderAuth } from "@/provider"
|
||||
import { Agent } from "@/agent/agent"
|
||||
import { Skill } from "@/skill"
|
||||
import { Discovery } from "@/skill/discovery"
|
||||
|
||||
@@ -9,219 +9,217 @@ import { ProviderID } from "./schema"
|
||||
import { Array as Arr, Effect, Layer, Record, Result, Context, Schema } from "effect"
|
||||
import z from "zod"
|
||||
|
||||
export namespace ProviderAuth {
|
||||
const When = Schema.Struct({
|
||||
key: Schema.String,
|
||||
op: Schema.Literals(["eq", "neq"]),
|
||||
value: Schema.String,
|
||||
})
|
||||
const When = Schema.Struct({
|
||||
key: Schema.String,
|
||||
op: Schema.Literals(["eq", "neq"]),
|
||||
value: Schema.String,
|
||||
})
|
||||
|
||||
const TextPrompt = Schema.Struct({
|
||||
type: Schema.Literal("text"),
|
||||
key: Schema.String,
|
||||
message: Schema.String,
|
||||
placeholder: Schema.optional(Schema.String),
|
||||
when: Schema.optional(When),
|
||||
})
|
||||
const TextPrompt = Schema.Struct({
|
||||
type: Schema.Literal("text"),
|
||||
key: Schema.String,
|
||||
message: Schema.String,
|
||||
placeholder: Schema.optional(Schema.String),
|
||||
when: Schema.optional(When),
|
||||
})
|
||||
|
||||
const SelectOption = Schema.Struct({
|
||||
label: Schema.String,
|
||||
value: Schema.String,
|
||||
hint: Schema.optional(Schema.String),
|
||||
})
|
||||
const SelectOption = Schema.Struct({
|
||||
label: Schema.String,
|
||||
value: Schema.String,
|
||||
hint: Schema.optional(Schema.String),
|
||||
})
|
||||
|
||||
const SelectPrompt = Schema.Struct({
|
||||
type: Schema.Literal("select"),
|
||||
key: Schema.String,
|
||||
message: Schema.String,
|
||||
options: Schema.Array(SelectOption),
|
||||
when: Schema.optional(When),
|
||||
})
|
||||
const SelectPrompt = Schema.Struct({
|
||||
type: Schema.Literal("select"),
|
||||
key: Schema.String,
|
||||
message: Schema.String,
|
||||
options: Schema.Array(SelectOption),
|
||||
when: Schema.optional(When),
|
||||
})
|
||||
|
||||
const Prompt = Schema.Union([TextPrompt, SelectPrompt])
|
||||
const Prompt = Schema.Union([TextPrompt, SelectPrompt])
|
||||
|
||||
export class Method extends Schema.Class<Method>("ProviderAuthMethod")({
|
||||
type: Schema.Literals(["oauth", "api"]),
|
||||
label: Schema.String,
|
||||
prompts: Schema.optional(Schema.Array(Prompt)),
|
||||
}) {
|
||||
static readonly zod = zod(this)
|
||||
}
|
||||
export class Method extends Schema.Class<Method>("ProviderAuthMethod")({
|
||||
type: Schema.Literals(["oauth", "api"]),
|
||||
label: Schema.String,
|
||||
prompts: Schema.optional(Schema.Array(Prompt)),
|
||||
}) {
|
||||
static readonly zod = zod(this)
|
||||
}
|
||||
|
||||
export const Methods = Schema.Record(Schema.String, Schema.Array(Method)).pipe(withStatics((s) => ({ zod: zod(s) })))
|
||||
export type Methods = typeof Methods.Type
|
||||
export const Methods = Schema.Record(Schema.String, Schema.Array(Method)).pipe(withStatics((s) => ({ zod: zod(s) })))
|
||||
export type Methods = typeof Methods.Type
|
||||
|
||||
export class Authorization extends Schema.Class<Authorization>("ProviderAuthAuthorization")({
|
||||
url: Schema.String,
|
||||
method: Schema.Literals(["auto", "code"]),
|
||||
instructions: Schema.String,
|
||||
}) {
|
||||
static readonly zod = zod(this)
|
||||
}
|
||||
export class Authorization extends Schema.Class<Authorization>("ProviderAuthAuthorization")({
|
||||
url: Schema.String,
|
||||
method: Schema.Literals(["auto", "code"]),
|
||||
instructions: Schema.String,
|
||||
}) {
|
||||
static readonly zod = zod(this)
|
||||
}
|
||||
|
||||
export const OauthMissing = NamedError.create("ProviderAuthOauthMissing", z.object({ providerID: ProviderID.zod }))
|
||||
export const OauthMissing = NamedError.create("ProviderAuthOauthMissing", z.object({ providerID: ProviderID.zod }))
|
||||
|
||||
export const OauthCodeMissing = NamedError.create(
|
||||
"ProviderAuthOauthCodeMissing",
|
||||
z.object({ providerID: ProviderID.zod }),
|
||||
)
|
||||
export const OauthCodeMissing = NamedError.create(
|
||||
"ProviderAuthOauthCodeMissing",
|
||||
z.object({ providerID: ProviderID.zod }),
|
||||
)
|
||||
|
||||
export const OauthCallbackFailed = NamedError.create("ProviderAuthOauthCallbackFailed", z.object({}))
|
||||
export const OauthCallbackFailed = NamedError.create("ProviderAuthOauthCallbackFailed", z.object({}))
|
||||
|
||||
export const ValidationFailed = NamedError.create(
|
||||
"ProviderAuthValidationFailed",
|
||||
z.object({
|
||||
field: z.string(),
|
||||
message: z.string(),
|
||||
}),
|
||||
)
|
||||
export const ValidationFailed = NamedError.create(
|
||||
"ProviderAuthValidationFailed",
|
||||
z.object({
|
||||
field: z.string(),
|
||||
message: z.string(),
|
||||
}),
|
||||
)
|
||||
|
||||
export type Error =
|
||||
| Auth.AuthError
|
||||
| InstanceType<typeof OauthMissing>
|
||||
| InstanceType<typeof OauthCodeMissing>
|
||||
| InstanceType<typeof OauthCallbackFailed>
|
||||
| InstanceType<typeof ValidationFailed>
|
||||
export type Error =
|
||||
| Auth.AuthError
|
||||
| InstanceType<typeof OauthMissing>
|
||||
| InstanceType<typeof OauthCodeMissing>
|
||||
| InstanceType<typeof OauthCallbackFailed>
|
||||
| InstanceType<typeof ValidationFailed>
|
||||
|
||||
type Hook = NonNullable<Hooks["auth"]>
|
||||
type Hook = NonNullable<Hooks["auth"]>
|
||||
|
||||
export interface Interface {
|
||||
readonly methods: () => Effect.Effect<Methods>
|
||||
readonly authorize: (input: {
|
||||
export interface Interface {
|
||||
readonly methods: () => Effect.Effect<Methods>
|
||||
readonly authorize: (input: {
|
||||
providerID: ProviderID
|
||||
method: number
|
||||
inputs?: Record<string, string>
|
||||
}) => Effect.Effect<Authorization | undefined, Error>
|
||||
readonly callback: (input: { providerID: ProviderID; method: number; code?: string }) => Effect.Effect<void, Error>
|
||||
}
|
||||
|
||||
interface State {
|
||||
hooks: Record<ProviderID, Hook>
|
||||
pending: Map<ProviderID, AuthOAuthResult>
|
||||
}
|
||||
|
||||
export class Service extends Context.Service<Service, Interface>()("@opencode/ProviderAuth") {}
|
||||
|
||||
export const layer: Layer.Layer<Service, never, Auth.Service | Plugin.Service> = Layer.effect(
|
||||
Service,
|
||||
Effect.gen(function* () {
|
||||
const auth = yield* Auth.Service
|
||||
const plugin = yield* Plugin.Service
|
||||
const state = yield* InstanceState.make<State>(
|
||||
Effect.fn("ProviderAuth.state")(function* () {
|
||||
const plugins = yield* plugin.list()
|
||||
return {
|
||||
hooks: Record.fromEntries(
|
||||
Arr.filterMap(plugins, (x) =>
|
||||
x.auth?.provider !== undefined
|
||||
? Result.succeed([ProviderID.make(x.auth.provider), x.auth] as const)
|
||||
: Result.failVoid,
|
||||
),
|
||||
),
|
||||
pending: new Map<ProviderID, AuthOAuthResult>(),
|
||||
}
|
||||
}),
|
||||
)
|
||||
|
||||
const decode = Schema.decodeUnknownSync(Methods)
|
||||
const methods = Effect.fn("ProviderAuth.methods")(function* () {
|
||||
const hooks = (yield* InstanceState.get(state)).hooks
|
||||
return decode(
|
||||
Record.map(hooks, (item) =>
|
||||
item.methods.map((method) => ({
|
||||
type: method.type,
|
||||
label: method.label,
|
||||
prompts: method.prompts?.map((prompt) => {
|
||||
if (prompt.type === "select") {
|
||||
return {
|
||||
type: "select" as const,
|
||||
key: prompt.key,
|
||||
message: prompt.message,
|
||||
options: prompt.options,
|
||||
when: prompt.when,
|
||||
}
|
||||
}
|
||||
return {
|
||||
type: "text" as const,
|
||||
key: prompt.key,
|
||||
message: prompt.message,
|
||||
placeholder: prompt.placeholder,
|
||||
when: prompt.when,
|
||||
}
|
||||
}),
|
||||
})),
|
||||
),
|
||||
)
|
||||
})
|
||||
|
||||
const authorize = Effect.fn("ProviderAuth.authorize")(function* (input: {
|
||||
providerID: ProviderID
|
||||
method: number
|
||||
inputs?: Record<string, string>
|
||||
}) => Effect.Effect<Authorization | undefined, Error>
|
||||
readonly callback: (input: { providerID: ProviderID; method: number; code?: string }) => Effect.Effect<void, Error>
|
||||
}
|
||||
}) {
|
||||
const { hooks, pending } = yield* InstanceState.get(state)
|
||||
const method = hooks[input.providerID].methods[input.method]
|
||||
if (method.type !== "oauth") return
|
||||
|
||||
interface State {
|
||||
hooks: Record<ProviderID, Hook>
|
||||
pending: Map<ProviderID, AuthOAuthResult>
|
||||
}
|
||||
|
||||
export class Service extends Context.Service<Service, Interface>()("@opencode/ProviderAuth") {}
|
||||
|
||||
export const layer: Layer.Layer<Service, never, Auth.Service | Plugin.Service> = Layer.effect(
|
||||
Service,
|
||||
Effect.gen(function* () {
|
||||
const auth = yield* Auth.Service
|
||||
const plugin = yield* Plugin.Service
|
||||
const state = yield* InstanceState.make<State>(
|
||||
Effect.fn("ProviderAuth.state")(function* () {
|
||||
const plugins = yield* plugin.list()
|
||||
return {
|
||||
hooks: Record.fromEntries(
|
||||
Arr.filterMap(plugins, (x) =>
|
||||
x.auth?.provider !== undefined
|
||||
? Result.succeed([ProviderID.make(x.auth.provider), x.auth] as const)
|
||||
: Result.failVoid,
|
||||
),
|
||||
),
|
||||
pending: new Map<ProviderID, AuthOAuthResult>(),
|
||||
if (method.prompts && input.inputs) {
|
||||
for (const prompt of method.prompts) {
|
||||
if (prompt.type === "text" && prompt.validate && input.inputs[prompt.key] !== undefined) {
|
||||
const error = prompt.validate(input.inputs[prompt.key])
|
||||
if (error) return yield* Effect.fail(new ValidationFailed({ field: prompt.key, message: error }))
|
||||
}
|
||||
}),
|
||||
}
|
||||
}
|
||||
|
||||
const result = yield* Effect.promise(() => method.authorize(input.inputs))
|
||||
pending.set(input.providerID, result)
|
||||
return {
|
||||
url: result.url,
|
||||
method: result.method,
|
||||
instructions: result.instructions,
|
||||
}
|
||||
})
|
||||
|
||||
const callback = Effect.fn("ProviderAuth.callback")(function* (input: {
|
||||
providerID: ProviderID
|
||||
method: number
|
||||
code?: string
|
||||
}) {
|
||||
const pending = (yield* InstanceState.get(state)).pending
|
||||
const match = pending.get(input.providerID)
|
||||
if (!match) return yield* Effect.fail(new OauthMissing({ providerID: input.providerID }))
|
||||
if (match.method === "code" && !input.code) {
|
||||
return yield* Effect.fail(new OauthCodeMissing({ providerID: input.providerID }))
|
||||
}
|
||||
|
||||
const result = yield* Effect.promise(() =>
|
||||
match.method === "code" ? match.callback(input.code!) : match.callback(),
|
||||
)
|
||||
if (!result || result.type !== "success") return yield* Effect.fail(new OauthCallbackFailed({}))
|
||||
|
||||
const decode = Schema.decodeUnknownSync(Methods)
|
||||
const methods = Effect.fn("ProviderAuth.methods")(function* () {
|
||||
const hooks = (yield* InstanceState.get(state)).hooks
|
||||
return decode(
|
||||
Record.map(hooks, (item) =>
|
||||
item.methods.map((method) => ({
|
||||
type: method.type,
|
||||
label: method.label,
|
||||
prompts: method.prompts?.map((prompt) => {
|
||||
if (prompt.type === "select") {
|
||||
return {
|
||||
type: "select" as const,
|
||||
key: prompt.key,
|
||||
message: prompt.message,
|
||||
options: prompt.options,
|
||||
when: prompt.when,
|
||||
}
|
||||
}
|
||||
return {
|
||||
type: "text" as const,
|
||||
key: prompt.key,
|
||||
message: prompt.message,
|
||||
placeholder: prompt.placeholder,
|
||||
when: prompt.when,
|
||||
}
|
||||
}),
|
||||
})),
|
||||
),
|
||||
)
|
||||
})
|
||||
if ("key" in result) {
|
||||
yield* auth.set(input.providerID, {
|
||||
type: "api",
|
||||
key: result.key,
|
||||
})
|
||||
}
|
||||
|
||||
const authorize = Effect.fn("ProviderAuth.authorize")(function* (input: {
|
||||
providerID: ProviderID
|
||||
method: number
|
||||
inputs?: Record<string, string>
|
||||
}) {
|
||||
const { hooks, pending } = yield* InstanceState.get(state)
|
||||
const method = hooks[input.providerID].methods[input.method]
|
||||
if (method.type !== "oauth") return
|
||||
if ("refresh" in result) {
|
||||
const { type: _, provider: __, refresh, access, expires, ...extra } = result
|
||||
yield* auth.set(input.providerID, {
|
||||
type: "oauth",
|
||||
access,
|
||||
refresh,
|
||||
expires,
|
||||
...extra,
|
||||
})
|
||||
}
|
||||
})
|
||||
|
||||
if (method.prompts && input.inputs) {
|
||||
for (const prompt of method.prompts) {
|
||||
if (prompt.type === "text" && prompt.validate && input.inputs[prompt.key] !== undefined) {
|
||||
const error = prompt.validate(input.inputs[prompt.key])
|
||||
if (error) return yield* Effect.fail(new ValidationFailed({ field: prompt.key, message: error }))
|
||||
}
|
||||
}
|
||||
}
|
||||
return Service.of({ methods, authorize, callback })
|
||||
}),
|
||||
)
|
||||
|
||||
const result = yield* Effect.promise(() => method.authorize(input.inputs))
|
||||
pending.set(input.providerID, result)
|
||||
return {
|
||||
url: result.url,
|
||||
method: result.method,
|
||||
instructions: result.instructions,
|
||||
}
|
||||
})
|
||||
|
||||
const callback = Effect.fn("ProviderAuth.callback")(function* (input: {
|
||||
providerID: ProviderID
|
||||
method: number
|
||||
code?: string
|
||||
}) {
|
||||
const pending = (yield* InstanceState.get(state)).pending
|
||||
const match = pending.get(input.providerID)
|
||||
if (!match) return yield* Effect.fail(new OauthMissing({ providerID: input.providerID }))
|
||||
if (match.method === "code" && !input.code) {
|
||||
return yield* Effect.fail(new OauthCodeMissing({ providerID: input.providerID }))
|
||||
}
|
||||
|
||||
const result = yield* Effect.promise(() =>
|
||||
match.method === "code" ? match.callback(input.code!) : match.callback(),
|
||||
)
|
||||
if (!result || result.type !== "success") return yield* Effect.fail(new OauthCallbackFailed({}))
|
||||
|
||||
if ("key" in result) {
|
||||
yield* auth.set(input.providerID, {
|
||||
type: "api",
|
||||
key: result.key,
|
||||
})
|
||||
}
|
||||
|
||||
if ("refresh" in result) {
|
||||
const { type: _, provider: __, refresh, access, expires, ...extra } = result
|
||||
yield* auth.set(input.providerID, {
|
||||
type: "oauth",
|
||||
access,
|
||||
refresh,
|
||||
expires,
|
||||
...extra,
|
||||
})
|
||||
}
|
||||
})
|
||||
|
||||
return Service.of({ methods, authorize, callback })
|
||||
}),
|
||||
)
|
||||
|
||||
export const defaultLayer = Layer.suspend(() =>
|
||||
layer.pipe(Layer.provide(Auth.defaultLayer), Layer.provide(Plugin.defaultLayer)),
|
||||
)
|
||||
}
|
||||
export const defaultLayer = Layer.suspend(() =>
|
||||
layer.pipe(Layer.provide(Auth.defaultLayer), Layer.provide(Plugin.defaultLayer)),
|
||||
)
|
||||
|
||||
@@ -3,195 +3,193 @@ import { STATUS_CODES } from "http"
|
||||
import { iife } from "@/util/iife"
|
||||
import type { ProviderID } from "./schema"
|
||||
|
||||
export namespace ProviderError {
|
||||
// Adapted from overflow detection patterns in:
|
||||
// https://github.com/badlogic/pi-mono/blob/main/packages/ai/src/utils/overflow.ts
|
||||
const OVERFLOW_PATTERNS = [
|
||||
/prompt is too long/i, // Anthropic
|
||||
/input is too long for requested model/i, // Amazon Bedrock
|
||||
/exceeds the context window/i, // OpenAI (Completions + Responses API message text)
|
||||
/input token count.*exceeds the maximum/i, // Google (Gemini)
|
||||
/maximum prompt length is \d+/i, // xAI (Grok)
|
||||
/reduce the length of the messages/i, // Groq
|
||||
/maximum context length is \d+ tokens/i, // OpenRouter, DeepSeek, vLLM
|
||||
/exceeds the limit of \d+/i, // GitHub Copilot
|
||||
/exceeds the available context size/i, // llama.cpp server
|
||||
/greater than the context length/i, // LM Studio
|
||||
/context window exceeds limit/i, // MiniMax
|
||||
/exceeded model token limit/i, // Kimi For Coding, Moonshot
|
||||
/context[_ ]length[_ ]exceeded/i, // Generic fallback
|
||||
/request entity too large/i, // HTTP 413
|
||||
/context length is only \d+ tokens/i, // vLLM
|
||||
/input length.*exceeds.*context length/i, // vLLM
|
||||
/prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error
|
||||
/too large for model with \d+ maximum context length/i, // Mistral
|
||||
/model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text
|
||||
]
|
||||
// Adapted from overflow detection patterns in:
|
||||
// https://github.com/badlogic/pi-mono/blob/main/packages/ai/src/utils/overflow.ts
|
||||
const OVERFLOW_PATTERNS = [
|
||||
/prompt is too long/i, // Anthropic
|
||||
/input is too long for requested model/i, // Amazon Bedrock
|
||||
/exceeds the context window/i, // OpenAI (Completions + Responses API message text)
|
||||
/input token count.*exceeds the maximum/i, // Google (Gemini)
|
||||
/maximum prompt length is \d+/i, // xAI (Grok)
|
||||
/reduce the length of the messages/i, // Groq
|
||||
/maximum context length is \d+ tokens/i, // OpenRouter, DeepSeek, vLLM
|
||||
/exceeds the limit of \d+/i, // GitHub Copilot
|
||||
/exceeds the available context size/i, // llama.cpp server
|
||||
/greater than the context length/i, // LM Studio
|
||||
/context window exceeds limit/i, // MiniMax
|
||||
/exceeded model token limit/i, // Kimi For Coding, Moonshot
|
||||
/context[_ ]length[_ ]exceeded/i, // Generic fallback
|
||||
/request entity too large/i, // HTTP 413
|
||||
/context length is only \d+ tokens/i, // vLLM
|
||||
/input length.*exceeds.*context length/i, // vLLM
|
||||
/prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error
|
||||
/too large for model with \d+ maximum context length/i, // Mistral
|
||||
/model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text
|
||||
]
|
||||
|
||||
function isOpenAiErrorRetryable(e: APICallError) {
|
||||
const status = e.statusCode
|
||||
if (!status) return e.isRetryable
|
||||
// openai sometimes returns 404 for models that are actually available
|
||||
return status === 404 || e.isRetryable
|
||||
}
|
||||
function isOpenAiErrorRetryable(e: APICallError) {
|
||||
const status = e.statusCode
|
||||
if (!status) return e.isRetryable
|
||||
// openai sometimes returns 404 for models that are actually available
|
||||
return status === 404 || e.isRetryable
|
||||
}
|
||||
|
||||
// Providers not reliably handled in this function:
|
||||
// - z.ai: can accept overflow silently (needs token-count/context-window checks)
|
||||
function isOverflow(message: string) {
|
||||
if (OVERFLOW_PATTERNS.some((p) => p.test(message))) return true
|
||||
// Providers not reliably handled in this function:
|
||||
// - z.ai: can accept overflow silently (needs token-count/context-window checks)
|
||||
function isOverflow(message: string) {
|
||||
if (OVERFLOW_PATTERNS.some((p) => p.test(message))) return true
|
||||
|
||||
// Providers/status patterns handled outside of regex list:
|
||||
// - Cerebras: often returns "400 (no body)" / "413 (no body)"
|
||||
// - Mistral: often returns "400 (no body)" / "413 (no body)"
|
||||
return /^4(00|13)\s*(status code)?\s*\(no body\)/i.test(message)
|
||||
}
|
||||
// Providers/status patterns handled outside of regex list:
|
||||
// - Cerebras: often returns "400 (no body)" / "413 (no body)"
|
||||
// - Mistral: often returns "400 (no body)" / "413 (no body)"
|
||||
return /^4(00|13)\s*(status code)?\s*\(no body\)/i.test(message)
|
||||
}
|
||||
|
||||
function message(providerID: ProviderID, e: APICallError) {
|
||||
return iife(() => {
|
||||
const msg = e.message
|
||||
if (msg === "") {
|
||||
if (e.responseBody) return e.responseBody
|
||||
if (e.statusCode) {
|
||||
const err = STATUS_CODES[e.statusCode]
|
||||
if (err) return err
|
||||
}
|
||||
return "Unknown error"
|
||||
}
|
||||
|
||||
if (!e.responseBody || (e.statusCode && msg !== STATUS_CODES[e.statusCode])) {
|
||||
return msg
|
||||
}
|
||||
|
||||
try {
|
||||
const body = JSON.parse(e.responseBody)
|
||||
// try to extract common error message fields
|
||||
const errMsg = body.message || body.error || body.error?.message
|
||||
if (errMsg && typeof errMsg === "string") {
|
||||
return `${msg}: ${errMsg}`
|
||||
}
|
||||
} catch {}
|
||||
|
||||
// If responseBody is HTML (e.g. from a gateway or proxy error page),
|
||||
// provide a human-readable message instead of dumping raw markup
|
||||
if (/^\s*<!doctype|^\s*<html/i.test(e.responseBody)) {
|
||||
if (e.statusCode === 401) {
|
||||
return "Unauthorized: request was blocked by a gateway or proxy. Your authentication token may be missing or expired — try running `opencode auth login <your provider URL>` to re-authenticate."
|
||||
}
|
||||
if (e.statusCode === 403) {
|
||||
return "Forbidden: request was blocked by a gateway or proxy. You may not have permission to access this resource — check your account and provider settings."
|
||||
}
|
||||
return msg
|
||||
}
|
||||
|
||||
return `${msg}: ${e.responseBody}`
|
||||
}).trim()
|
||||
}
|
||||
|
||||
function json(input: unknown) {
|
||||
if (typeof input === "string") {
|
||||
try {
|
||||
const result = JSON.parse(input)
|
||||
if (result && typeof result === "object") return result
|
||||
return undefined
|
||||
} catch {
|
||||
return undefined
|
||||
function message(providerID: ProviderID, e: APICallError) {
|
||||
return iife(() => {
|
||||
const msg = e.message
|
||||
if (msg === "") {
|
||||
if (e.responseBody) return e.responseBody
|
||||
if (e.statusCode) {
|
||||
const err = STATUS_CODES[e.statusCode]
|
||||
if (err) return err
|
||||
}
|
||||
return "Unknown error"
|
||||
}
|
||||
if (typeof input === "object" && input !== null) {
|
||||
return input
|
||||
|
||||
if (!e.responseBody || (e.statusCode && msg !== STATUS_CODES[e.statusCode])) {
|
||||
return msg
|
||||
}
|
||||
return undefined
|
||||
}
|
||||
|
||||
export type ParsedStreamError =
|
||||
| {
|
||||
type: "context_overflow"
|
||||
message: string
|
||||
responseBody: string
|
||||
try {
|
||||
const body = JSON.parse(e.responseBody)
|
||||
// try to extract common error message fields
|
||||
const errMsg = body.message || body.error || body.error?.message
|
||||
if (errMsg && typeof errMsg === "string") {
|
||||
return `${msg}: ${errMsg}`
|
||||
}
|
||||
| {
|
||||
type: "api_error"
|
||||
message: string
|
||||
isRetryable: false
|
||||
responseBody: string
|
||||
} catch {}
|
||||
|
||||
// If responseBody is HTML (e.g. from a gateway or proxy error page),
|
||||
// provide a human-readable message instead of dumping raw markup
|
||||
if (/^\s*<!doctype|^\s*<html/i.test(e.responseBody)) {
|
||||
if (e.statusCode === 401) {
|
||||
return "Unauthorized: request was blocked by a gateway or proxy. Your authentication token may be missing or expired — try running `opencode auth login <your provider URL>` to re-authenticate."
|
||||
}
|
||||
if (e.statusCode === 403) {
|
||||
return "Forbidden: request was blocked by a gateway or proxy. You may not have permission to access this resource — check your account and provider settings."
|
||||
}
|
||||
return msg
|
||||
}
|
||||
|
||||
export function parseStreamError(input: unknown): ParsedStreamError | undefined {
|
||||
const body = json(input)
|
||||
if (!body) return
|
||||
return `${msg}: ${e.responseBody}`
|
||||
}).trim()
|
||||
}
|
||||
|
||||
const responseBody = JSON.stringify(body)
|
||||
if (body.type !== "error") return
|
||||
|
||||
switch (body?.error?.code) {
|
||||
case "context_length_exceeded":
|
||||
return {
|
||||
type: "context_overflow",
|
||||
message: "Input exceeds context window of this model",
|
||||
responseBody,
|
||||
}
|
||||
case "insufficient_quota":
|
||||
return {
|
||||
type: "api_error",
|
||||
message: "Quota exceeded. Check your plan and billing details.",
|
||||
isRetryable: false,
|
||||
responseBody,
|
||||
}
|
||||
case "usage_not_included":
|
||||
return {
|
||||
type: "api_error",
|
||||
message: "To use Codex with your ChatGPT plan, upgrade to Plus: https://chatgpt.com/explore/plus.",
|
||||
isRetryable: false,
|
||||
responseBody,
|
||||
}
|
||||
case "invalid_prompt":
|
||||
return {
|
||||
type: "api_error",
|
||||
message: typeof body?.error?.message === "string" ? body?.error?.message : "Invalid prompt.",
|
||||
isRetryable: false,
|
||||
responseBody,
|
||||
}
|
||||
function json(input: unknown) {
|
||||
if (typeof input === "string") {
|
||||
try {
|
||||
const result = JSON.parse(input)
|
||||
if (result && typeof result === "object") return result
|
||||
return undefined
|
||||
} catch {
|
||||
return undefined
|
||||
}
|
||||
}
|
||||
if (typeof input === "object" && input !== null) {
|
||||
return input
|
||||
}
|
||||
return undefined
|
||||
}
|
||||
|
||||
export type ParsedAPICallError =
|
||||
| {
|
||||
type: "context_overflow"
|
||||
message: string
|
||||
responseBody?: string
|
||||
}
|
||||
| {
|
||||
type: "api_error"
|
||||
message: string
|
||||
statusCode?: number
|
||||
isRetryable: boolean
|
||||
responseHeaders?: Record<string, string>
|
||||
responseBody?: string
|
||||
metadata?: Record<string, string>
|
||||
}
|
||||
export type ParsedStreamError =
|
||||
| {
|
||||
type: "context_overflow"
|
||||
message: string
|
||||
responseBody: string
|
||||
}
|
||||
| {
|
||||
type: "api_error"
|
||||
message: string
|
||||
isRetryable: false
|
||||
responseBody: string
|
||||
}
|
||||
|
||||
export function parseAPICallError(input: { providerID: ProviderID; error: APICallError }): ParsedAPICallError {
|
||||
const m = message(input.providerID, input.error)
|
||||
const body = json(input.error.responseBody)
|
||||
if (isOverflow(m) || input.error.statusCode === 413 || body?.error?.code === "context_length_exceeded") {
|
||||
export function parseStreamError(input: unknown): ParsedStreamError | undefined {
|
||||
const body = json(input)
|
||||
if (!body) return
|
||||
|
||||
const responseBody = JSON.stringify(body)
|
||||
if (body.type !== "error") return
|
||||
|
||||
switch (body?.error?.code) {
|
||||
case "context_length_exceeded":
|
||||
return {
|
||||
type: "context_overflow",
|
||||
message: m,
|
||||
responseBody: input.error.responseBody,
|
||||
message: "Input exceeds context window of this model",
|
||||
responseBody,
|
||||
}
|
||||
case "insufficient_quota":
|
||||
return {
|
||||
type: "api_error",
|
||||
message: "Quota exceeded. Check your plan and billing details.",
|
||||
isRetryable: false,
|
||||
responseBody,
|
||||
}
|
||||
case "usage_not_included":
|
||||
return {
|
||||
type: "api_error",
|
||||
message: "To use Codex with your ChatGPT plan, upgrade to Plus: https://chatgpt.com/explore/plus.",
|
||||
isRetryable: false,
|
||||
responseBody,
|
||||
}
|
||||
case "invalid_prompt":
|
||||
return {
|
||||
type: "api_error",
|
||||
message: typeof body?.error?.message === "string" ? body?.error?.message : "Invalid prompt.",
|
||||
isRetryable: false,
|
||||
responseBody,
|
||||
}
|
||||
}
|
||||
|
||||
const metadata = input.error.url ? { url: input.error.url } : undefined
|
||||
return {
|
||||
type: "api_error",
|
||||
message: m,
|
||||
statusCode: input.error.statusCode,
|
||||
isRetryable: input.providerID.startsWith("openai")
|
||||
? isOpenAiErrorRetryable(input.error)
|
||||
: input.error.isRetryable,
|
||||
responseHeaders: input.error.responseHeaders,
|
||||
responseBody: input.error.responseBody,
|
||||
metadata,
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
export type ParsedAPICallError =
|
||||
| {
|
||||
type: "context_overflow"
|
||||
message: string
|
||||
responseBody?: string
|
||||
}
|
||||
| {
|
||||
type: "api_error"
|
||||
message: string
|
||||
statusCode?: number
|
||||
isRetryable: boolean
|
||||
responseHeaders?: Record<string, string>
|
||||
responseBody?: string
|
||||
metadata?: Record<string, string>
|
||||
}
|
||||
|
||||
export function parseAPICallError(input: { providerID: ProviderID; error: APICallError }): ParsedAPICallError {
|
||||
const m = message(input.providerID, input.error)
|
||||
const body = json(input.error.responseBody)
|
||||
if (isOverflow(m) || input.error.statusCode === 413 || body?.error?.code === "context_length_exceeded") {
|
||||
return {
|
||||
type: "context_overflow",
|
||||
message: m,
|
||||
responseBody: input.error.responseBody,
|
||||
}
|
||||
}
|
||||
|
||||
const metadata = input.error.url ? { url: input.error.url } : undefined
|
||||
return {
|
||||
type: "api_error",
|
||||
message: m,
|
||||
statusCode: input.error.statusCode,
|
||||
isRetryable: input.providerID.startsWith("openai")
|
||||
? isOpenAiErrorRetryable(input.error)
|
||||
: input.error.isRetryable,
|
||||
responseHeaders: input.error.responseHeaders,
|
||||
responseBody: input.error.responseBody,
|
||||
metadata,
|
||||
}
|
||||
}
|
||||
|
||||
@@ -1 +1,5 @@
|
||||
export * as Provider from "./provider"
|
||||
export * as ProviderAuth from "./auth"
|
||||
export * as ProviderError from "./error"
|
||||
export * as ModelsDev from "./models"
|
||||
export * as ProviderTransform from "./transform"
|
||||
|
||||
@@ -13,169 +13,167 @@ import { Hash } from "@opencode-ai/shared/util/hash"
|
||||
// Falls back to undefined in dev mode when snapshot doesn't exist
|
||||
/* @ts-ignore */
|
||||
|
||||
export namespace ModelsDev {
|
||||
const log = Log.create({ service: "models.dev" })
|
||||
const source = url()
|
||||
const filepath = path.join(
|
||||
Global.Path.cache,
|
||||
source === "https://models.dev" ? "models.json" : `models-${Hash.fast(source)}.json`,
|
||||
)
|
||||
const ttl = 5 * 60 * 1000
|
||||
const log = Log.create({ service: "models.dev" })
|
||||
const source = url()
|
||||
const filepath = path.join(
|
||||
Global.Path.cache,
|
||||
source === "https://models.dev" ? "models.json" : `models-${Hash.fast(source)}.json`,
|
||||
)
|
||||
const ttl = 5 * 60 * 1000
|
||||
|
||||
type JsonValue = string | number | boolean | null | { [key: string]: JsonValue } | JsonValue[]
|
||||
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 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(),
|
||||
family: z.string().optional(),
|
||||
release_date: z.string(),
|
||||
attachment: z.boolean(),
|
||||
reasoning: z.boolean(),
|
||||
temperature: z.boolean(),
|
||||
tool_call: z.boolean(),
|
||||
interleaved: z
|
||||
.union([
|
||||
z.literal(true),
|
||||
z
|
||||
.object({
|
||||
field: z.enum(["reasoning_content", "reasoning_details"]),
|
||||
})
|
||||
.strict(),
|
||||
])
|
||||
.optional(),
|
||||
cost: Cost.optional(),
|
||||
limit: z.object({
|
||||
context: z.number(),
|
||||
input: z.number().optional(),
|
||||
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(),
|
||||
}),
|
||||
modalities: z
|
||||
.object({
|
||||
input: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
|
||||
output: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
|
||||
})
|
||||
.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(),
|
||||
})
|
||||
export type Model = z.infer<typeof Model>
|
||||
|
||||
export const Provider = z.object({
|
||||
api: z.string().optional(),
|
||||
name: z.string(),
|
||||
env: z.array(z.string()),
|
||||
id: z.string(),
|
||||
npm: z.string().optional(),
|
||||
models: z.record(z.string(), Model),
|
||||
})
|
||||
|
||||
export type Provider = z.infer<typeof Provider>
|
||||
|
||||
function url() {
|
||||
return Flag.OPENCODE_MODELS_URL || "https://models.dev"
|
||||
}
|
||||
|
||||
function fresh() {
|
||||
return Date.now() - Number(Filesystem.stat(filepath)?.mtimeMs ?? 0) < ttl
|
||||
}
|
||||
|
||||
function skip(force: boolean) {
|
||||
return !force && fresh()
|
||||
}
|
||||
|
||||
const fetchApi = async () => {
|
||||
const result = await fetch(`${url()}/api.json`, {
|
||||
headers: { "User-Agent": Installation.USER_AGENT },
|
||||
signal: AbortSignal.timeout(10000),
|
||||
cache_read: z.number().optional(),
|
||||
cache_write: z.number().optional(),
|
||||
})
|
||||
return { ok: result.ok, text: await result.text() }
|
||||
}
|
||||
.optional(),
|
||||
})
|
||||
|
||||
export const Data = lazy(async () => {
|
||||
export const Model = z.object({
|
||||
id: z.string(),
|
||||
name: z.string(),
|
||||
family: z.string().optional(),
|
||||
release_date: z.string(),
|
||||
attachment: z.boolean(),
|
||||
reasoning: z.boolean(),
|
||||
temperature: z.boolean(),
|
||||
tool_call: z.boolean(),
|
||||
interleaved: z
|
||||
.union([
|
||||
z.literal(true),
|
||||
z
|
||||
.object({
|
||||
field: z.enum(["reasoning_content", "reasoning_details"]),
|
||||
})
|
||||
.strict(),
|
||||
])
|
||||
.optional(),
|
||||
cost: Cost.optional(),
|
||||
limit: z.object({
|
||||
context: z.number(),
|
||||
input: z.number().optional(),
|
||||
output: z.number(),
|
||||
}),
|
||||
modalities: z
|
||||
.object({
|
||||
input: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
|
||||
output: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
|
||||
})
|
||||
.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(),
|
||||
})
|
||||
export type Model = z.infer<typeof Model>
|
||||
|
||||
export const Provider = z.object({
|
||||
api: z.string().optional(),
|
||||
name: z.string(),
|
||||
env: z.array(z.string()),
|
||||
id: z.string(),
|
||||
npm: z.string().optional(),
|
||||
models: z.record(z.string(), Model),
|
||||
})
|
||||
|
||||
export type Provider = z.infer<typeof Provider>
|
||||
|
||||
function url() {
|
||||
return Flag.OPENCODE_MODELS_URL || "https://models.dev"
|
||||
}
|
||||
|
||||
function fresh() {
|
||||
return Date.now() - Number(Filesystem.stat(filepath)?.mtimeMs ?? 0) < ttl
|
||||
}
|
||||
|
||||
function skip(force: boolean) {
|
||||
return !force && fresh()
|
||||
}
|
||||
|
||||
const fetchApi = async () => {
|
||||
const result = await fetch(`${url()}/api.json`, {
|
||||
headers: { "User-Agent": Installation.USER_AGENT },
|
||||
signal: AbortSignal.timeout(10000),
|
||||
})
|
||||
return { ok: result.ok, text: await result.text() }
|
||||
}
|
||||
|
||||
export const Data = lazy(async () => {
|
||||
const result = await Filesystem.readJson(Flag.OPENCODE_MODELS_PATH ?? filepath).catch(() => {})
|
||||
if (result) return result
|
||||
// @ts-ignore
|
||||
const snapshot = await import("./models-snapshot.js")
|
||||
.then((m) => m.snapshot as Record<string, unknown>)
|
||||
.catch(() => undefined)
|
||||
if (snapshot) return snapshot
|
||||
if (Flag.OPENCODE_DISABLE_MODELS_FETCH) return {}
|
||||
return Flock.withLock(`models-dev:${filepath}`, async () => {
|
||||
const result = await Filesystem.readJson(Flag.OPENCODE_MODELS_PATH ?? filepath).catch(() => {})
|
||||
if (result) return result
|
||||
// @ts-ignore
|
||||
const snapshot = await import("./models-snapshot.js")
|
||||
.then((m) => m.snapshot as Record<string, unknown>)
|
||||
.catch(() => undefined)
|
||||
if (snapshot) return snapshot
|
||||
if (Flag.OPENCODE_DISABLE_MODELS_FETCH) return {}
|
||||
return Flock.withLock(`models-dev:${filepath}`, async () => {
|
||||
const result = await Filesystem.readJson(Flag.OPENCODE_MODELS_PATH ?? filepath).catch(() => {})
|
||||
if (result) return result
|
||||
const result2 = await fetchApi()
|
||||
if (result2.ok) {
|
||||
await Filesystem.write(filepath, result2.text).catch((e) => {
|
||||
log.error("Failed to write models cache", { error: e })
|
||||
})
|
||||
}
|
||||
return JSON.parse(result2.text)
|
||||
const result2 = await fetchApi()
|
||||
if (result2.ok) {
|
||||
await Filesystem.write(filepath, result2.text).catch((e) => {
|
||||
log.error("Failed to write models cache", { error: e })
|
||||
})
|
||||
}
|
||||
return JSON.parse(result2.text)
|
||||
})
|
||||
})
|
||||
|
||||
export async function get() {
|
||||
const result = await Data()
|
||||
return result as Record<string, Provider>
|
||||
}
|
||||
|
||||
export async function refresh(force = false) {
|
||||
if (skip(force)) return Data.reset()
|
||||
await Flock.withLock(`models-dev:${filepath}`, async () => {
|
||||
if (skip(force)) return Data.reset()
|
||||
const result = await fetchApi()
|
||||
if (!result.ok) return
|
||||
await Filesystem.write(filepath, result.text)
|
||||
Data.reset()
|
||||
}).catch((e) => {
|
||||
log.error("Failed to fetch models.dev", {
|
||||
error: e,
|
||||
})
|
||||
})
|
||||
|
||||
export async function get() {
|
||||
const result = await Data()
|
||||
return result as Record<string, Provider>
|
||||
}
|
||||
|
||||
export async function refresh(force = false) {
|
||||
if (skip(force)) return ModelsDev.Data.reset()
|
||||
await Flock.withLock(`models-dev:${filepath}`, async () => {
|
||||
if (skip(force)) return ModelsDev.Data.reset()
|
||||
const result = await fetchApi()
|
||||
if (!result.ok) return
|
||||
await Filesystem.write(filepath, result.text)
|
||||
ModelsDev.Data.reset()
|
||||
}).catch((e) => {
|
||||
log.error("Failed to fetch models.dev", {
|
||||
error: e,
|
||||
})
|
||||
})
|
||||
}
|
||||
}
|
||||
|
||||
if (!Flag.OPENCODE_DISABLE_MODELS_FETCH && !process.argv.includes("--get-yargs-completions")) {
|
||||
void ModelsDev.refresh()
|
||||
void refresh()
|
||||
setInterval(
|
||||
async () => {
|
||||
await ModelsDev.refresh()
|
||||
await refresh()
|
||||
},
|
||||
60 * 1000 * 60,
|
||||
).unref()
|
||||
|
||||
@@ -10,7 +10,7 @@ import { Hash } from "@opencode-ai/shared/util/hash"
|
||||
import { Plugin } from "../plugin"
|
||||
import { NamedError } from "@opencode-ai/shared/util/error"
|
||||
import { type LanguageModelV3 } from "@ai-sdk/provider"
|
||||
import { ModelsDev } from "./models"
|
||||
import * as ModelsDev from "./models"
|
||||
import { Auth } from "../auth"
|
||||
import { Env } from "../env"
|
||||
import { Instance } from "../project/instance"
|
||||
@@ -55,7 +55,7 @@ import {
|
||||
} from "gitlab-ai-provider"
|
||||
import { fromNodeProviderChain } from "@aws-sdk/credential-providers"
|
||||
import { GoogleAuth } from "google-auth-library"
|
||||
import { ProviderTransform } from "./transform"
|
||||
import * as ProviderTransform from "./transform"
|
||||
import { Installation } from "../installation"
|
||||
import { ModelID, ProviderID } from "./schema"
|
||||
|
||||
|
||||
@@ -3,7 +3,7 @@ import { mergeDeep, unique } from "remeda"
|
||||
import type { JSONSchema7 } from "@ai-sdk/provider"
|
||||
import type { JSONSchema } from "zod/v4/core"
|
||||
import type * as Provider from "./provider"
|
||||
import type { ModelsDev } from "./models"
|
||||
import type * as ModelsDev from "./models"
|
||||
import { iife } from "@/util/iife"
|
||||
import { Flag } from "@/flag/flag"
|
||||
|
||||
@@ -17,570 +17,420 @@ function mimeToModality(mime: string): Modality | undefined {
|
||||
return undefined
|
||||
}
|
||||
|
||||
export namespace ProviderTransform {
|
||||
export const OUTPUT_TOKEN_MAX = Flag.OPENCODE_EXPERIMENTAL_OUTPUT_TOKEN_MAX || 32_000
|
||||
export const OUTPUT_TOKEN_MAX = Flag.OPENCODE_EXPERIMENTAL_OUTPUT_TOKEN_MAX || 32_000
|
||||
|
||||
// Maps npm package to the key the AI SDK expects for providerOptions
|
||||
function sdkKey(npm: string): string | undefined {
|
||||
switch (npm) {
|
||||
case "@ai-sdk/github-copilot":
|
||||
return "copilot"
|
||||
case "@ai-sdk/azure":
|
||||
return "azure"
|
||||
case "@ai-sdk/openai":
|
||||
return "openai"
|
||||
case "@ai-sdk/amazon-bedrock":
|
||||
return "bedrock"
|
||||
case "@ai-sdk/anthropic":
|
||||
case "@ai-sdk/google-vertex/anthropic":
|
||||
return "anthropic"
|
||||
case "@ai-sdk/google-vertex":
|
||||
return "vertex"
|
||||
case "@ai-sdk/google":
|
||||
return "google"
|
||||
case "@ai-sdk/gateway":
|
||||
return "gateway"
|
||||
case "@openrouter/ai-sdk-provider":
|
||||
return "openrouter"
|
||||
}
|
||||
return undefined
|
||||
// Maps npm package to the key the AI SDK expects for providerOptions
|
||||
function sdkKey(npm: string): string | undefined {
|
||||
switch (npm) {
|
||||
case "@ai-sdk/github-copilot":
|
||||
return "copilot"
|
||||
case "@ai-sdk/azure":
|
||||
return "azure"
|
||||
case "@ai-sdk/openai":
|
||||
return "openai"
|
||||
case "@ai-sdk/amazon-bedrock":
|
||||
return "bedrock"
|
||||
case "@ai-sdk/anthropic":
|
||||
case "@ai-sdk/google-vertex/anthropic":
|
||||
return "anthropic"
|
||||
case "@ai-sdk/google-vertex":
|
||||
return "vertex"
|
||||
case "@ai-sdk/google":
|
||||
return "google"
|
||||
case "@ai-sdk/gateway":
|
||||
return "gateway"
|
||||
case "@openrouter/ai-sdk-provider":
|
||||
return "openrouter"
|
||||
}
|
||||
return undefined
|
||||
}
|
||||
|
||||
function normalizeMessages(
|
||||
msgs: ModelMessage[],
|
||||
model: Provider.Model,
|
||||
_options: Record<string, unknown>,
|
||||
): ModelMessage[] {
|
||||
// Anthropic rejects messages with empty content - filter out empty string messages
|
||||
// and remove empty text/reasoning parts from array content
|
||||
if (model.api.npm === "@ai-sdk/anthropic" || model.api.npm === "@ai-sdk/amazon-bedrock") {
|
||||
msgs = msgs
|
||||
.map((msg) => {
|
||||
if (typeof msg.content === "string") {
|
||||
if (msg.content === "") return undefined
|
||||
return msg
|
||||
}
|
||||
if (!Array.isArray(msg.content)) return msg
|
||||
const filtered = msg.content.filter((part) => {
|
||||
if (part.type === "text" || part.type === "reasoning") {
|
||||
return part.text !== ""
|
||||
}
|
||||
return true
|
||||
})
|
||||
if (filtered.length === 0) return undefined
|
||||
return { ...msg, content: filtered }
|
||||
})
|
||||
.filter((msg): msg is ModelMessage => msg !== undefined && msg.content !== "")
|
||||
}
|
||||
|
||||
function normalizeMessages(
|
||||
msgs: ModelMessage[],
|
||||
model: Provider.Model,
|
||||
_options: Record<string, unknown>,
|
||||
): ModelMessage[] {
|
||||
// Anthropic rejects messages with empty content - filter out empty string messages
|
||||
// and remove empty text/reasoning parts from array content
|
||||
if (model.api.npm === "@ai-sdk/anthropic" || model.api.npm === "@ai-sdk/amazon-bedrock") {
|
||||
msgs = msgs
|
||||
.map((msg) => {
|
||||
if (typeof msg.content === "string") {
|
||||
if (msg.content === "") return undefined
|
||||
return msg
|
||||
}
|
||||
if (!Array.isArray(msg.content)) return msg
|
||||
const filtered = msg.content.filter((part) => {
|
||||
if (part.type === "text" || part.type === "reasoning") {
|
||||
return part.text !== ""
|
||||
}
|
||||
return true
|
||||
})
|
||||
if (filtered.length === 0) return undefined
|
||||
return { ...msg, content: filtered }
|
||||
})
|
||||
.filter((msg): msg is ModelMessage => msg !== undefined && msg.content !== "")
|
||||
}
|
||||
|
||||
if (model.api.id.includes("claude")) {
|
||||
const scrub = (id: string) => id.replace(/[^a-zA-Z0-9_-]/g, "_")
|
||||
msgs = msgs.map((msg) => {
|
||||
if (msg.role === "assistant" && Array.isArray(msg.content)) {
|
||||
return {
|
||||
...msg,
|
||||
content: msg.content.map((part) => {
|
||||
if (part.type === "tool-call" || part.type === "tool-result") {
|
||||
return { ...part, toolCallId: scrub(part.toolCallId) }
|
||||
}
|
||||
return part
|
||||
}),
|
||||
}
|
||||
}
|
||||
if (msg.role === "tool" && Array.isArray(msg.content)) {
|
||||
return {
|
||||
...msg,
|
||||
content: msg.content.map((part) => {
|
||||
if (part.type === "tool-result") {
|
||||
return { ...part, toolCallId: scrub(part.toolCallId) }
|
||||
}
|
||||
return part
|
||||
}),
|
||||
}
|
||||
}
|
||||
return msg
|
||||
})
|
||||
}
|
||||
if (["@ai-sdk/anthropic", "@ai-sdk/google-vertex/anthropic"].includes(model.api.npm)) {
|
||||
// Anthropic rejects assistant turns where tool_use blocks are followed by non-tool
|
||||
// content, e.g. [tool_use, tool_use, text], with:
|
||||
// `tool_use` ids were found without `tool_result` blocks immediately after...
|
||||
//
|
||||
// Reorder that invalid shape into [text] + [tool_use, tool_use]. Consecutive
|
||||
// assistant messages are later merged by the provider/SDK, so preserving the
|
||||
// original [tool_use...] then [text] order still produces the invalid payload.
|
||||
//
|
||||
// The root cause appears to be somewhere upstream where the stream is originally
|
||||
// processed. We were unable to locate an exact narrower reproduction elsewhere,
|
||||
// so we keep this transform in place for the time being.
|
||||
msgs = msgs.flatMap((msg) => {
|
||||
if (msg.role !== "assistant" || !Array.isArray(msg.content)) return [msg]
|
||||
|
||||
const parts = msg.content
|
||||
const first = parts.findIndex((part) => part.type === "tool-call")
|
||||
if (first === -1) return [msg]
|
||||
if (!parts.slice(first).some((part) => part.type !== "tool-call")) return [msg]
|
||||
return [
|
||||
{ ...msg, content: parts.filter((part) => part.type !== "tool-call") },
|
||||
{ ...msg, content: parts.filter((part) => part.type === "tool-call") },
|
||||
]
|
||||
})
|
||||
}
|
||||
if (
|
||||
model.providerID === "mistral" ||
|
||||
model.api.id.toLowerCase().includes("mistral") ||
|
||||
model.api.id.toLocaleLowerCase().includes("devstral")
|
||||
) {
|
||||
const scrub = (id: string) => {
|
||||
return id
|
||||
.replace(/[^a-zA-Z0-9]/g, "") // Remove non-alphanumeric characters
|
||||
.substring(0, 9) // Take first 9 characters
|
||||
.padEnd(9, "0") // Pad with zeros if less than 9 characters
|
||||
}
|
||||
const result: ModelMessage[] = []
|
||||
for (let i = 0; i < msgs.length; i++) {
|
||||
const msg = msgs[i]
|
||||
const nextMsg = msgs[i + 1]
|
||||
|
||||
if (msg.role === "assistant" && Array.isArray(msg.content)) {
|
||||
msg.content = msg.content.map((part) => {
|
||||
if (model.api.id.includes("claude")) {
|
||||
const scrub = (id: string) => id.replace(/[^a-zA-Z0-9_-]/g, "_")
|
||||
msgs = msgs.map((msg) => {
|
||||
if (msg.role === "assistant" && Array.isArray(msg.content)) {
|
||||
return {
|
||||
...msg,
|
||||
content: msg.content.map((part) => {
|
||||
if (part.type === "tool-call" || part.type === "tool-result") {
|
||||
return { ...part, toolCallId: scrub(part.toolCallId) }
|
||||
}
|
||||
return part
|
||||
})
|
||||
}),
|
||||
}
|
||||
if (msg.role === "tool" && Array.isArray(msg.content)) {
|
||||
msg.content = msg.content.map((part) => {
|
||||
}
|
||||
if (msg.role === "tool" && Array.isArray(msg.content)) {
|
||||
return {
|
||||
...msg,
|
||||
content: msg.content.map((part) => {
|
||||
if (part.type === "tool-result") {
|
||||
return { ...part, toolCallId: scrub(part.toolCallId) }
|
||||
}
|
||||
return part
|
||||
})
|
||||
}
|
||||
result.push(msg)
|
||||
|
||||
// Fix message sequence: tool messages cannot be followed by user messages
|
||||
if (msg.role === "tool" && nextMsg?.role === "user") {
|
||||
result.push({
|
||||
role: "assistant",
|
||||
content: [
|
||||
{
|
||||
type: "text",
|
||||
text: "Done.",
|
||||
},
|
||||
],
|
||||
})
|
||||
}),
|
||||
}
|
||||
}
|
||||
return result
|
||||
return msg
|
||||
})
|
||||
}
|
||||
if (["@ai-sdk/anthropic", "@ai-sdk/google-vertex/anthropic"].includes(model.api.npm)) {
|
||||
// Anthropic rejects assistant turns where tool_use blocks are followed by non-tool
|
||||
// content, e.g. [tool_use, tool_use, text], with:
|
||||
// `tool_use` ids were found without `tool_result` blocks immediately after...
|
||||
//
|
||||
// Reorder that invalid shape into [text] + [tool_use, tool_use]. Consecutive
|
||||
// assistant messages are later merged by the provider/SDK, so preserving the
|
||||
// original [tool_use...] then [text] order still produces the invalid payload.
|
||||
//
|
||||
// The root cause appears to be somewhere upstream where the stream is originally
|
||||
// processed. We were unable to locate an exact narrower reproduction elsewhere,
|
||||
// so we keep this transform in place for the time being.
|
||||
msgs = msgs.flatMap((msg) => {
|
||||
if (msg.role !== "assistant" || !Array.isArray(msg.content)) return [msg]
|
||||
|
||||
const parts = msg.content
|
||||
const first = parts.findIndex((part) => part.type === "tool-call")
|
||||
if (first === -1) return [msg]
|
||||
if (!parts.slice(first).some((part) => part.type !== "tool-call")) return [msg]
|
||||
return [
|
||||
{ ...msg, content: parts.filter((part) => part.type !== "tool-call") },
|
||||
{ ...msg, content: parts.filter((part) => part.type === "tool-call") },
|
||||
]
|
||||
})
|
||||
}
|
||||
if (
|
||||
model.providerID === "mistral" ||
|
||||
model.api.id.toLowerCase().includes("mistral") ||
|
||||
model.api.id.toLocaleLowerCase().includes("devstral")
|
||||
) {
|
||||
const scrub = (id: string) => {
|
||||
return id
|
||||
.replace(/[^a-zA-Z0-9]/g, "") // Remove non-alphanumeric characters
|
||||
.substring(0, 9) // Take first 9 characters
|
||||
.padEnd(9, "0") // Pad with zeros if less than 9 characters
|
||||
}
|
||||
const result: ModelMessage[] = []
|
||||
for (let i = 0; i < msgs.length; i++) {
|
||||
const msg = msgs[i]
|
||||
const nextMsg = msgs[i + 1]
|
||||
|
||||
if (typeof model.capabilities.interleaved === "object" && model.capabilities.interleaved.field) {
|
||||
const field = model.capabilities.interleaved.field
|
||||
return msgs.map((msg) => {
|
||||
if (msg.role === "assistant" && Array.isArray(msg.content)) {
|
||||
const reasoningParts = msg.content.filter((part: any) => part.type === "reasoning")
|
||||
const reasoningText = reasoningParts.map((part: any) => part.text).join("")
|
||||
|
||||
// Filter out reasoning parts from content
|
||||
const filteredContent = msg.content.filter((part: any) => part.type !== "reasoning")
|
||||
|
||||
// Include reasoning_content | reasoning_details directly on the message for all assistant messages
|
||||
if (reasoningText) {
|
||||
return {
|
||||
...msg,
|
||||
content: filteredContent,
|
||||
providerOptions: {
|
||||
...msg.providerOptions,
|
||||
openaiCompatible: {
|
||||
...(msg.providerOptions as any)?.openaiCompatible,
|
||||
[field]: reasoningText,
|
||||
},
|
||||
},
|
||||
}
|
||||
if (msg.role === "assistant" && Array.isArray(msg.content)) {
|
||||
msg.content = msg.content.map((part) => {
|
||||
if (part.type === "tool-call" || part.type === "tool-result") {
|
||||
return { ...part, toolCallId: scrub(part.toolCallId) }
|
||||
}
|
||||
return part
|
||||
})
|
||||
}
|
||||
if (msg.role === "tool" && Array.isArray(msg.content)) {
|
||||
msg.content = msg.content.map((part) => {
|
||||
if (part.type === "tool-result") {
|
||||
return { ...part, toolCallId: scrub(part.toolCallId) }
|
||||
}
|
||||
return part
|
||||
})
|
||||
}
|
||||
result.push(msg)
|
||||
|
||||
// Fix message sequence: tool messages cannot be followed by user messages
|
||||
if (msg.role === "tool" && nextMsg?.role === "user") {
|
||||
result.push({
|
||||
role: "assistant",
|
||||
content: [
|
||||
{
|
||||
type: "text",
|
||||
text: "Done.",
|
||||
},
|
||||
],
|
||||
})
|
||||
}
|
||||
}
|
||||
return result
|
||||
}
|
||||
|
||||
if (typeof model.capabilities.interleaved === "object" && model.capabilities.interleaved.field) {
|
||||
const field = model.capabilities.interleaved.field
|
||||
return msgs.map((msg) => {
|
||||
if (msg.role === "assistant" && Array.isArray(msg.content)) {
|
||||
const reasoningParts = msg.content.filter((part: any) => part.type === "reasoning")
|
||||
const reasoningText = reasoningParts.map((part: any) => part.text).join("")
|
||||
|
||||
// Filter out reasoning parts from content
|
||||
const filteredContent = msg.content.filter((part: any) => part.type !== "reasoning")
|
||||
|
||||
// Include reasoning_content | reasoning_details directly on the message for all assistant messages
|
||||
if (reasoningText) {
|
||||
return {
|
||||
...msg,
|
||||
content: filteredContent,
|
||||
providerOptions: {
|
||||
...msg.providerOptions,
|
||||
openaiCompatible: {
|
||||
...(msg.providerOptions as any)?.openaiCompatible,
|
||||
[field]: reasoningText,
|
||||
},
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
return msg
|
||||
})
|
||||
}
|
||||
|
||||
return msgs
|
||||
}
|
||||
|
||||
function applyCaching(msgs: ModelMessage[], model: Provider.Model): ModelMessage[] {
|
||||
const system = msgs.filter((msg) => msg.role === "system").slice(0, 2)
|
||||
const final = msgs.filter((msg) => msg.role !== "system").slice(-2)
|
||||
|
||||
const providerOptions = {
|
||||
anthropic: {
|
||||
cacheControl: { type: "ephemeral" },
|
||||
},
|
||||
openrouter: {
|
||||
cacheControl: { type: "ephemeral" },
|
||||
},
|
||||
bedrock: {
|
||||
cachePoint: { type: "default" },
|
||||
},
|
||||
openaiCompatible: {
|
||||
cache_control: { type: "ephemeral" },
|
||||
},
|
||||
copilot: {
|
||||
copilot_cache_control: { type: "ephemeral" },
|
||||
},
|
||||
alibaba: {
|
||||
cacheControl: { type: "ephemeral" },
|
||||
},
|
||||
}
|
||||
|
||||
for (const msg of unique([...system, ...final])) {
|
||||
const useMessageLevelOptions =
|
||||
model.providerID === "anthropic" ||
|
||||
model.providerID.includes("bedrock") ||
|
||||
model.api.npm === "@ai-sdk/amazon-bedrock"
|
||||
const shouldUseContentOptions = !useMessageLevelOptions && Array.isArray(msg.content) && msg.content.length > 0
|
||||
|
||||
if (shouldUseContentOptions) {
|
||||
const lastContent = msg.content[msg.content.length - 1]
|
||||
if (
|
||||
lastContent &&
|
||||
typeof lastContent === "object" &&
|
||||
lastContent.type !== "tool-approval-request" &&
|
||||
lastContent.type !== "tool-approval-response"
|
||||
) {
|
||||
lastContent.providerOptions = mergeDeep(lastContent.providerOptions ?? {}, providerOptions)
|
||||
continue
|
||||
return {
|
||||
...msg,
|
||||
content: filteredContent,
|
||||
}
|
||||
}
|
||||
|
||||
msg.providerOptions = mergeDeep(msg.providerOptions ?? {}, providerOptions)
|
||||
}
|
||||
|
||||
return msgs
|
||||
}
|
||||
|
||||
function unsupportedParts(msgs: ModelMessage[], model: Provider.Model): ModelMessage[] {
|
||||
return msgs.map((msg) => {
|
||||
if (msg.role !== "user" || !Array.isArray(msg.content)) return msg
|
||||
|
||||
const filtered = msg.content.map((part) => {
|
||||
if (part.type !== "file" && part.type !== "image") return part
|
||||
|
||||
// Check for empty base64 image data
|
||||
if (part.type === "image") {
|
||||
const imageStr = String(part.image)
|
||||
if (imageStr.startsWith("data:")) {
|
||||
const match = imageStr.match(/^data:([^;]+);base64,(.*)$/)
|
||||
if (match && (!match[2] || match[2].length === 0)) {
|
||||
return {
|
||||
type: "text" as const,
|
||||
text: "ERROR: Image file is empty or corrupted. Please provide a valid image.",
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
const mime = part.type === "image" ? String(part.image).split(";")[0].replace("data:", "") : part.mediaType
|
||||
const filename = part.type === "file" ? part.filename : undefined
|
||||
const modality = mimeToModality(mime)
|
||||
if (!modality) return part
|
||||
if (model.capabilities.input[modality]) return part
|
||||
|
||||
const name = filename ? `"${filename}"` : modality
|
||||
return {
|
||||
type: "text" as const,
|
||||
text: `ERROR: Cannot read ${name} (this model does not support ${modality} input). Inform the user.`,
|
||||
}
|
||||
})
|
||||
|
||||
return { ...msg, content: filtered }
|
||||
return msg
|
||||
})
|
||||
}
|
||||
|
||||
export function message(msgs: ModelMessage[], model: Provider.Model, options: Record<string, unknown>) {
|
||||
msgs = unsupportedParts(msgs, model)
|
||||
msgs = normalizeMessages(msgs, model, options)
|
||||
if (
|
||||
(model.providerID === "anthropic" ||
|
||||
model.providerID === "google-vertex-anthropic" ||
|
||||
model.api.id.includes("anthropic") ||
|
||||
model.api.id.includes("claude") ||
|
||||
model.id.includes("anthropic") ||
|
||||
model.id.includes("claude") ||
|
||||
model.api.npm === "@ai-sdk/anthropic" ||
|
||||
model.api.npm === "@ai-sdk/alibaba") &&
|
||||
model.api.npm !== "@ai-sdk/gateway"
|
||||
) {
|
||||
msgs = applyCaching(msgs, model)
|
||||
}
|
||||
return msgs
|
||||
}
|
||||
|
||||
// Remap providerOptions keys from stored providerID to expected SDK key
|
||||
const key = sdkKey(model.api.npm)
|
||||
if (key && key !== model.providerID) {
|
||||
const remap = (opts: Record<string, any> | undefined) => {
|
||||
if (!opts) return opts
|
||||
if (!(model.providerID in opts)) return opts
|
||||
const result = { ...opts }
|
||||
result[key] = result[model.providerID]
|
||||
delete result[model.providerID]
|
||||
return result
|
||||
}
|
||||
function applyCaching(msgs: ModelMessage[], model: Provider.Model): ModelMessage[] {
|
||||
const system = msgs.filter((msg) => msg.role === "system").slice(0, 2)
|
||||
const final = msgs.filter((msg) => msg.role !== "system").slice(-2)
|
||||
|
||||
msgs = msgs.map((msg) => {
|
||||
if (!Array.isArray(msg.content)) return { ...msg, providerOptions: remap(msg.providerOptions) }
|
||||
return {
|
||||
...msg,
|
||||
providerOptions: remap(msg.providerOptions),
|
||||
content: msg.content.map((part) => {
|
||||
if (part.type === "tool-approval-request" || part.type === "tool-approval-response") {
|
||||
return { ...part }
|
||||
}
|
||||
return { ...part, providerOptions: remap(part.providerOptions) }
|
||||
}),
|
||||
} as typeof msg
|
||||
})
|
||||
}
|
||||
|
||||
return msgs
|
||||
const providerOptions = {
|
||||
anthropic: {
|
||||
cacheControl: { type: "ephemeral" },
|
||||
},
|
||||
openrouter: {
|
||||
cacheControl: { type: "ephemeral" },
|
||||
},
|
||||
bedrock: {
|
||||
cachePoint: { type: "default" },
|
||||
},
|
||||
openaiCompatible: {
|
||||
cache_control: { type: "ephemeral" },
|
||||
},
|
||||
copilot: {
|
||||
copilot_cache_control: { type: "ephemeral" },
|
||||
},
|
||||
alibaba: {
|
||||
cacheControl: { type: "ephemeral" },
|
||||
},
|
||||
}
|
||||
|
||||
export function temperature(model: Provider.Model) {
|
||||
const id = model.id.toLowerCase()
|
||||
if (id.includes("qwen")) return 0.55
|
||||
if (id.includes("claude")) return undefined
|
||||
if (id.includes("gemini")) return 1.0
|
||||
if (id.includes("glm-4.6")) return 1.0
|
||||
if (id.includes("glm-4.7")) return 1.0
|
||||
if (id.includes("minimax-m2")) return 1.0
|
||||
if (id.includes("kimi-k2")) {
|
||||
// kimi-k2-thinking & kimi-k2.5 && kimi-k2p5 && kimi-k2-5
|
||||
if (["thinking", "k2.", "k2p", "k2-5"].some((s) => id.includes(s))) {
|
||||
return 1.0
|
||||
}
|
||||
return 0.6
|
||||
}
|
||||
return undefined
|
||||
}
|
||||
for (const msg of unique([...system, ...final])) {
|
||||
const useMessageLevelOptions =
|
||||
model.providerID === "anthropic" ||
|
||||
model.providerID.includes("bedrock") ||
|
||||
model.api.npm === "@ai-sdk/amazon-bedrock"
|
||||
const shouldUseContentOptions = !useMessageLevelOptions && Array.isArray(msg.content) && msg.content.length > 0
|
||||
|
||||
export function topP(model: Provider.Model) {
|
||||
const id = model.id.toLowerCase()
|
||||
if (id.includes("qwen")) return 1
|
||||
if (["minimax-m2", "gemini", "kimi-k2.5", "kimi-k2p5", "kimi-k2-5"].some((s) => id.includes(s))) {
|
||||
return 0.95
|
||||
}
|
||||
return undefined
|
||||
}
|
||||
|
||||
export function topK(model: Provider.Model) {
|
||||
const id = model.id.toLowerCase()
|
||||
if (id.includes("minimax-m2")) {
|
||||
if (["m2.", "m25", "m21"].some((s) => id.includes(s))) return 40
|
||||
return 20
|
||||
}
|
||||
if (id.includes("gemini")) return 64
|
||||
return undefined
|
||||
}
|
||||
|
||||
const WIDELY_SUPPORTED_EFFORTS = ["low", "medium", "high"]
|
||||
const OPENAI_EFFORTS = ["none", "minimal", ...WIDELY_SUPPORTED_EFFORTS, "xhigh"]
|
||||
|
||||
export function variants(model: Provider.Model): Record<string, Record<string, any>> {
|
||||
if (!model.capabilities.reasoning) return {}
|
||||
|
||||
const id = model.id.toLowerCase()
|
||||
const isAnthropicAdaptive = ["opus-4-6", "opus-4.6", "sonnet-4-6", "sonnet-4.6"].some((v) =>
|
||||
model.api.id.includes(v),
|
||||
)
|
||||
const adaptiveEfforts = ["low", "medium", "high", "max"]
|
||||
if (
|
||||
id.includes("deepseek") ||
|
||||
id.includes("minimax") ||
|
||||
id.includes("glm") ||
|
||||
id.includes("mistral") ||
|
||||
id.includes("kimi") ||
|
||||
id.includes("k2p5") ||
|
||||
id.includes("qwen") ||
|
||||
id.includes("big-pickle")
|
||||
)
|
||||
return {}
|
||||
|
||||
// see: https://docs.x.ai/docs/guides/reasoning#control-how-hard-the-model-thinks
|
||||
if (id.includes("grok") && id.includes("grok-3-mini")) {
|
||||
if (model.api.npm === "@openrouter/ai-sdk-provider") {
|
||||
return {
|
||||
low: { reasoning: { effort: "low" } },
|
||||
high: { reasoning: { effort: "high" } },
|
||||
}
|
||||
}
|
||||
return {
|
||||
low: { reasoningEffort: "low" },
|
||||
high: { reasoningEffort: "high" },
|
||||
if (shouldUseContentOptions) {
|
||||
const lastContent = msg.content[msg.content.length - 1]
|
||||
if (
|
||||
lastContent &&
|
||||
typeof lastContent === "object" &&
|
||||
lastContent.type !== "tool-approval-request" &&
|
||||
lastContent.type !== "tool-approval-response"
|
||||
) {
|
||||
lastContent.providerOptions = mergeDeep(lastContent.providerOptions ?? {}, providerOptions)
|
||||
continue
|
||||
}
|
||||
}
|
||||
if (id.includes("grok")) return {}
|
||||
|
||||
switch (model.api.npm) {
|
||||
case "@openrouter/ai-sdk-provider":
|
||||
if (!model.id.includes("gpt") && !model.id.includes("gemini-3") && !model.id.includes("claude")) return {}
|
||||
return Object.fromEntries(OPENAI_EFFORTS.map((effort) => [effort, { reasoning: { effort } }]))
|
||||
msg.providerOptions = mergeDeep(msg.providerOptions ?? {}, providerOptions)
|
||||
}
|
||||
|
||||
case "@ai-sdk/gateway":
|
||||
if (model.id.includes("anthropic")) {
|
||||
if (isAnthropicAdaptive) {
|
||||
return Object.fromEntries(
|
||||
adaptiveEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
thinking: {
|
||||
type: "adaptive",
|
||||
},
|
||||
effort,
|
||||
},
|
||||
]),
|
||||
)
|
||||
}
|
||||
return {
|
||||
high: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: 16000,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: 31999,
|
||||
},
|
||||
},
|
||||
}
|
||||
}
|
||||
if (model.id.includes("google")) {
|
||||
if (id.includes("2.5")) {
|
||||
return msgs
|
||||
}
|
||||
|
||||
function unsupportedParts(msgs: ModelMessage[], model: Provider.Model): ModelMessage[] {
|
||||
return msgs.map((msg) => {
|
||||
if (msg.role !== "user" || !Array.isArray(msg.content)) return msg
|
||||
|
||||
const filtered = msg.content.map((part) => {
|
||||
if (part.type !== "file" && part.type !== "image") return part
|
||||
|
||||
// Check for empty base64 image data
|
||||
if (part.type === "image") {
|
||||
const imageStr = String(part.image)
|
||||
if (imageStr.startsWith("data:")) {
|
||||
const match = imageStr.match(/^data:([^;]+);base64,(.*)$/)
|
||||
if (match && (!match[2] || match[2].length === 0)) {
|
||||
return {
|
||||
high: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 16000,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 24576,
|
||||
},
|
||||
},
|
||||
type: "text" as const,
|
||||
text: "ERROR: Image file is empty or corrupted. Please provide a valid image.",
|
||||
}
|
||||
}
|
||||
return Object.fromEntries(
|
||||
["low", "high"].map((effort) => [
|
||||
effort,
|
||||
{
|
||||
includeThoughts: true,
|
||||
thinkingLevel: effort,
|
||||
},
|
||||
]),
|
||||
)
|
||||
}
|
||||
return Object.fromEntries(OPENAI_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
}
|
||||
|
||||
case "@ai-sdk/github-copilot":
|
||||
if (model.id.includes("gemini")) {
|
||||
// currently github copilot only returns thinking
|
||||
return {}
|
||||
}
|
||||
if (model.id.includes("claude")) {
|
||||
return Object.fromEntries(WIDELY_SUPPORTED_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
}
|
||||
const copilotEfforts = iife(() => {
|
||||
if (id.includes("5.1-codex-max") || id.includes("5.2") || id.includes("5.3"))
|
||||
return [...WIDELY_SUPPORTED_EFFORTS, "xhigh"]
|
||||
const arr = [...WIDELY_SUPPORTED_EFFORTS]
|
||||
if (id.includes("gpt-5") && model.release_date >= "2025-12-04") arr.push("xhigh")
|
||||
return arr
|
||||
})
|
||||
return Object.fromEntries(
|
||||
copilotEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
reasoningSummary: "auto",
|
||||
include: ["reasoning.encrypted_content"],
|
||||
},
|
||||
]),
|
||||
)
|
||||
const mime = part.type === "image" ? String(part.image).split(";")[0].replace("data:", "") : part.mediaType
|
||||
const filename = part.type === "file" ? part.filename : undefined
|
||||
const modality = mimeToModality(mime)
|
||||
if (!modality) return part
|
||||
if (model.capabilities.input[modality]) return part
|
||||
|
||||
case "@ai-sdk/cerebras":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/cerebras
|
||||
case "@ai-sdk/togetherai":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/togetherai
|
||||
case "@ai-sdk/xai":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/xai
|
||||
case "@ai-sdk/deepinfra":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/deepinfra
|
||||
case "venice-ai-sdk-provider":
|
||||
// https://docs.venice.ai/overview/guides/reasoning-models#reasoning-effort
|
||||
case "@ai-sdk/openai-compatible":
|
||||
return Object.fromEntries(WIDELY_SUPPORTED_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
const name = filename ? `"${filename}"` : modality
|
||||
return {
|
||||
type: "text" as const,
|
||||
text: `ERROR: Cannot read ${name} (this model does not support ${modality} input). Inform the user.`,
|
||||
}
|
||||
})
|
||||
|
||||
case "@ai-sdk/azure":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/azure
|
||||
if (id === "o1-mini") return {}
|
||||
const azureEfforts = ["low", "medium", "high"]
|
||||
if (id.includes("gpt-5-") || id === "gpt-5") {
|
||||
azureEfforts.unshift("minimal")
|
||||
}
|
||||
return Object.fromEntries(
|
||||
azureEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
reasoningSummary: "auto",
|
||||
include: ["reasoning.encrypted_content"],
|
||||
},
|
||||
]),
|
||||
)
|
||||
case "@ai-sdk/openai":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/openai
|
||||
if (id === "gpt-5-pro") return {}
|
||||
const openaiEfforts = iife(() => {
|
||||
if (id.includes("codex")) {
|
||||
if (id.includes("5.2") || id.includes("5.3")) return [...WIDELY_SUPPORTED_EFFORTS, "xhigh"]
|
||||
return WIDELY_SUPPORTED_EFFORTS
|
||||
return { ...msg, content: filtered }
|
||||
})
|
||||
}
|
||||
|
||||
export function message(msgs: ModelMessage[], model: Provider.Model, options: Record<string, unknown>) {
|
||||
msgs = unsupportedParts(msgs, model)
|
||||
msgs = normalizeMessages(msgs, model, options)
|
||||
if (
|
||||
(model.providerID === "anthropic" ||
|
||||
model.providerID === "google-vertex-anthropic" ||
|
||||
model.api.id.includes("anthropic") ||
|
||||
model.api.id.includes("claude") ||
|
||||
model.id.includes("anthropic") ||
|
||||
model.id.includes("claude") ||
|
||||
model.api.npm === "@ai-sdk/anthropic" ||
|
||||
model.api.npm === "@ai-sdk/alibaba") &&
|
||||
model.api.npm !== "@ai-sdk/gateway"
|
||||
) {
|
||||
msgs = applyCaching(msgs, model)
|
||||
}
|
||||
|
||||
// Remap providerOptions keys from stored providerID to expected SDK key
|
||||
const key = sdkKey(model.api.npm)
|
||||
if (key && key !== model.providerID) {
|
||||
const remap = (opts: Record<string, any> | undefined) => {
|
||||
if (!opts) return opts
|
||||
if (!(model.providerID in opts)) return opts
|
||||
const result = { ...opts }
|
||||
result[key] = result[model.providerID]
|
||||
delete result[model.providerID]
|
||||
return result
|
||||
}
|
||||
|
||||
msgs = msgs.map((msg) => {
|
||||
if (!Array.isArray(msg.content)) return { ...msg, providerOptions: remap(msg.providerOptions) }
|
||||
return {
|
||||
...msg,
|
||||
providerOptions: remap(msg.providerOptions),
|
||||
content: msg.content.map((part) => {
|
||||
if (part.type === "tool-approval-request" || part.type === "tool-approval-response") {
|
||||
return { ...part }
|
||||
}
|
||||
const arr = [...WIDELY_SUPPORTED_EFFORTS]
|
||||
if (id.includes("gpt-5-") || id === "gpt-5") {
|
||||
arr.unshift("minimal")
|
||||
}
|
||||
if (model.release_date >= "2025-11-13") {
|
||||
arr.unshift("none")
|
||||
}
|
||||
if (model.release_date >= "2025-12-04") {
|
||||
arr.push("xhigh")
|
||||
}
|
||||
return arr
|
||||
})
|
||||
return Object.fromEntries(
|
||||
openaiEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
reasoningSummary: "auto",
|
||||
include: ["reasoning.encrypted_content"],
|
||||
},
|
||||
]),
|
||||
)
|
||||
return { ...part, providerOptions: remap(part.providerOptions) }
|
||||
}),
|
||||
} as typeof msg
|
||||
})
|
||||
}
|
||||
|
||||
case "@ai-sdk/anthropic":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/anthropic
|
||||
case "@ai-sdk/google-vertex/anthropic":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/google-vertex#anthropic-provider
|
||||
return msgs
|
||||
}
|
||||
|
||||
export function temperature(model: Provider.Model) {
|
||||
const id = model.id.toLowerCase()
|
||||
if (id.includes("qwen")) return 0.55
|
||||
if (id.includes("claude")) return undefined
|
||||
if (id.includes("gemini")) return 1.0
|
||||
if (id.includes("glm-4.6")) return 1.0
|
||||
if (id.includes("glm-4.7")) return 1.0
|
||||
if (id.includes("minimax-m2")) return 1.0
|
||||
if (id.includes("kimi-k2")) {
|
||||
// kimi-k2-thinking & kimi-k2.5 && kimi-k2p5 && kimi-k2-5
|
||||
if (["thinking", "k2.", "k2p", "k2-5"].some((s) => id.includes(s))) {
|
||||
return 1.0
|
||||
}
|
||||
return 0.6
|
||||
}
|
||||
return undefined
|
||||
}
|
||||
|
||||
export function topP(model: Provider.Model) {
|
||||
const id = model.id.toLowerCase()
|
||||
if (id.includes("qwen")) return 1
|
||||
if (["minimax-m2", "gemini", "kimi-k2.5", "kimi-k2p5", "kimi-k2-5"].some((s) => id.includes(s))) {
|
||||
return 0.95
|
||||
}
|
||||
return undefined
|
||||
}
|
||||
|
||||
export function topK(model: Provider.Model) {
|
||||
const id = model.id.toLowerCase()
|
||||
if (id.includes("minimax-m2")) {
|
||||
if (["m2.", "m25", "m21"].some((s) => id.includes(s))) return 40
|
||||
return 20
|
||||
}
|
||||
if (id.includes("gemini")) return 64
|
||||
return undefined
|
||||
}
|
||||
|
||||
const WIDELY_SUPPORTED_EFFORTS = ["low", "medium", "high"]
|
||||
const OPENAI_EFFORTS = ["none", "minimal", ...WIDELY_SUPPORTED_EFFORTS, "xhigh"]
|
||||
|
||||
export function variants(model: Provider.Model): Record<string, Record<string, any>> {
|
||||
if (!model.capabilities.reasoning) return {}
|
||||
|
||||
const id = model.id.toLowerCase()
|
||||
const isAnthropicAdaptive = ["opus-4-6", "opus-4.6", "sonnet-4-6", "sonnet-4.6"].some((v) =>
|
||||
model.api.id.includes(v),
|
||||
)
|
||||
const adaptiveEfforts = ["low", "medium", "high", "max"]
|
||||
if (
|
||||
id.includes("deepseek") ||
|
||||
id.includes("minimax") ||
|
||||
id.includes("glm") ||
|
||||
id.includes("mistral") ||
|
||||
id.includes("kimi") ||
|
||||
id.includes("k2p5") ||
|
||||
id.includes("qwen") ||
|
||||
id.includes("big-pickle")
|
||||
)
|
||||
return {}
|
||||
|
||||
// see: https://docs.x.ai/docs/guides/reasoning#control-how-hard-the-model-thinks
|
||||
if (id.includes("grok") && id.includes("grok-3-mini")) {
|
||||
if (model.api.npm === "@openrouter/ai-sdk-provider") {
|
||||
return {
|
||||
low: { reasoning: { effort: "low" } },
|
||||
high: { reasoning: { effort: "high" } },
|
||||
}
|
||||
}
|
||||
return {
|
||||
low: { reasoningEffort: "low" },
|
||||
high: { reasoningEffort: "high" },
|
||||
}
|
||||
}
|
||||
if (id.includes("grok")) return {}
|
||||
|
||||
switch (model.api.npm) {
|
||||
case "@openrouter/ai-sdk-provider":
|
||||
if (!model.id.includes("gpt") && !model.id.includes("gemini-3") && !model.id.includes("claude")) return {}
|
||||
return Object.fromEntries(OPENAI_EFFORTS.map((effort) => [effort, { reasoning: { effort } }]))
|
||||
|
||||
case "@ai-sdk/gateway":
|
||||
if (model.id.includes("anthropic")) {
|
||||
if (isAnthropicAdaptive) {
|
||||
return Object.fromEntries(
|
||||
adaptiveEfforts.map((effort) => [
|
||||
@@ -594,72 +444,22 @@ export namespace ProviderTransform {
|
||||
]),
|
||||
)
|
||||
}
|
||||
|
||||
return {
|
||||
high: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: Math.min(16_000, Math.floor(model.limit.output / 2 - 1)),
|
||||
budgetTokens: 16000,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: Math.min(31_999, model.limit.output - 1),
|
||||
budgetTokens: 31999,
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
case "@ai-sdk/amazon-bedrock":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/amazon-bedrock
|
||||
if (isAnthropicAdaptive) {
|
||||
return Object.fromEntries(
|
||||
adaptiveEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningConfig: {
|
||||
type: "adaptive",
|
||||
maxReasoningEffort: effort,
|
||||
},
|
||||
},
|
||||
]),
|
||||
)
|
||||
}
|
||||
// For Anthropic models on Bedrock, use reasoningConfig with budgetTokens
|
||||
if (model.api.id.includes("anthropic")) {
|
||||
return {
|
||||
high: {
|
||||
reasoningConfig: {
|
||||
type: "enabled",
|
||||
budgetTokens: 16000,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
reasoningConfig: {
|
||||
type: "enabled",
|
||||
budgetTokens: 31999,
|
||||
},
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
// For Amazon Nova models, use reasoningConfig with maxReasoningEffort
|
||||
return Object.fromEntries(
|
||||
WIDELY_SUPPORTED_EFFORTS.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningConfig: {
|
||||
type: "enabled",
|
||||
maxReasoningEffort: effort,
|
||||
},
|
||||
},
|
||||
]),
|
||||
)
|
||||
|
||||
case "@ai-sdk/google-vertex":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/google-vertex
|
||||
case "@ai-sdk/google":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/google-generative-ai
|
||||
}
|
||||
if (model.id.includes("google")) {
|
||||
if (id.includes("2.5")) {
|
||||
return {
|
||||
high: {
|
||||
@@ -676,417 +476,615 @@ export namespace ProviderTransform {
|
||||
},
|
||||
}
|
||||
}
|
||||
let levels = ["low", "high"]
|
||||
if (id.includes("3.1")) {
|
||||
levels = ["low", "medium", "high"]
|
||||
}
|
||||
|
||||
return Object.fromEntries(
|
||||
levels.map((effort) => [
|
||||
["low", "high"].map((effort) => [
|
||||
effort,
|
||||
{
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingLevel: effort,
|
||||
includeThoughts: true,
|
||||
thinkingLevel: effort,
|
||||
},
|
||||
]),
|
||||
)
|
||||
}
|
||||
return Object.fromEntries(OPENAI_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
|
||||
case "@ai-sdk/github-copilot":
|
||||
if (model.id.includes("gemini")) {
|
||||
// currently github copilot only returns thinking
|
||||
return {}
|
||||
}
|
||||
if (model.id.includes("claude")) {
|
||||
return Object.fromEntries(WIDELY_SUPPORTED_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
}
|
||||
const copilotEfforts = iife(() => {
|
||||
if (id.includes("5.1-codex-max") || id.includes("5.2") || id.includes("5.3"))
|
||||
return [...WIDELY_SUPPORTED_EFFORTS, "xhigh"]
|
||||
const arr = [...WIDELY_SUPPORTED_EFFORTS]
|
||||
if (id.includes("gpt-5") && model.release_date >= "2025-12-04") arr.push("xhigh")
|
||||
return arr
|
||||
})
|
||||
return Object.fromEntries(
|
||||
copilotEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
reasoningSummary: "auto",
|
||||
include: ["reasoning.encrypted_content"],
|
||||
},
|
||||
]),
|
||||
)
|
||||
|
||||
case "@ai-sdk/cerebras":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/cerebras
|
||||
case "@ai-sdk/togetherai":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/togetherai
|
||||
case "@ai-sdk/xai":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/xai
|
||||
case "@ai-sdk/deepinfra":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/deepinfra
|
||||
case "venice-ai-sdk-provider":
|
||||
// https://docs.venice.ai/overview/guides/reasoning-models#reasoning-effort
|
||||
case "@ai-sdk/openai-compatible":
|
||||
return Object.fromEntries(WIDELY_SUPPORTED_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
|
||||
case "@ai-sdk/azure":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/azure
|
||||
if (id === "o1-mini") return {}
|
||||
const azureEfforts = ["low", "medium", "high"]
|
||||
if (id.includes("gpt-5-") || id === "gpt-5") {
|
||||
azureEfforts.unshift("minimal")
|
||||
}
|
||||
return Object.fromEntries(
|
||||
azureEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
reasoningSummary: "auto",
|
||||
include: ["reasoning.encrypted_content"],
|
||||
},
|
||||
]),
|
||||
)
|
||||
case "@ai-sdk/openai":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/openai
|
||||
if (id === "gpt-5-pro") return {}
|
||||
const openaiEfforts = iife(() => {
|
||||
if (id.includes("codex")) {
|
||||
if (id.includes("5.2") || id.includes("5.3")) return [...WIDELY_SUPPORTED_EFFORTS, "xhigh"]
|
||||
return WIDELY_SUPPORTED_EFFORTS
|
||||
}
|
||||
const arr = [...WIDELY_SUPPORTED_EFFORTS]
|
||||
if (id.includes("gpt-5-") || id === "gpt-5") {
|
||||
arr.unshift("minimal")
|
||||
}
|
||||
if (model.release_date >= "2025-11-13") {
|
||||
arr.unshift("none")
|
||||
}
|
||||
if (model.release_date >= "2025-12-04") {
|
||||
arr.push("xhigh")
|
||||
}
|
||||
return arr
|
||||
})
|
||||
return Object.fromEntries(
|
||||
openaiEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
reasoningSummary: "auto",
|
||||
include: ["reasoning.encrypted_content"],
|
||||
},
|
||||
]),
|
||||
)
|
||||
|
||||
case "@ai-sdk/anthropic":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/anthropic
|
||||
case "@ai-sdk/google-vertex/anthropic":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/google-vertex#anthropic-provider
|
||||
|
||||
if (isAnthropicAdaptive) {
|
||||
return Object.fromEntries(
|
||||
adaptiveEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
thinking: {
|
||||
type: "adaptive",
|
||||
},
|
||||
effort,
|
||||
},
|
||||
]),
|
||||
)
|
||||
}
|
||||
|
||||
return {
|
||||
high: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: Math.min(16_000, Math.floor(model.limit.output / 2 - 1)),
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: Math.min(31_999, model.limit.output - 1),
|
||||
},
|
||||
},
|
||||
}
|
||||
|
||||
case "@ai-sdk/amazon-bedrock":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/amazon-bedrock
|
||||
if (isAnthropicAdaptive) {
|
||||
return Object.fromEntries(
|
||||
adaptiveEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningConfig: {
|
||||
type: "adaptive",
|
||||
maxReasoningEffort: effort,
|
||||
},
|
||||
},
|
||||
]),
|
||||
)
|
||||
|
||||
case "@ai-sdk/mistral":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/mistral
|
||||
return {}
|
||||
|
||||
case "@ai-sdk/cohere":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/cohere
|
||||
return {}
|
||||
|
||||
case "@ai-sdk/groq":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/groq
|
||||
const groqEffort = ["none", ...WIDELY_SUPPORTED_EFFORTS]
|
||||
return Object.fromEntries(
|
||||
groqEffort.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
}
|
||||
// For Anthropic models on Bedrock, use reasoningConfig with budgetTokens
|
||||
if (model.api.id.includes("anthropic")) {
|
||||
return {
|
||||
high: {
|
||||
reasoningConfig: {
|
||||
type: "enabled",
|
||||
budgetTokens: 16000,
|
||||
},
|
||||
]),
|
||||
)
|
||||
},
|
||||
max: {
|
||||
reasoningConfig: {
|
||||
type: "enabled",
|
||||
budgetTokens: 31999,
|
||||
},
|
||||
},
|
||||
}
|
||||
}
|
||||
|
||||
case "@ai-sdk/perplexity":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/perplexity
|
||||
return {}
|
||||
// For Amazon Nova models, use reasoningConfig with maxReasoningEffort
|
||||
return Object.fromEntries(
|
||||
WIDELY_SUPPORTED_EFFORTS.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningConfig: {
|
||||
type: "enabled",
|
||||
maxReasoningEffort: effort,
|
||||
},
|
||||
},
|
||||
]),
|
||||
)
|
||||
|
||||
case "@jerome-benoit/sap-ai-provider-v2":
|
||||
if (model.api.id.includes("anthropic")) {
|
||||
if (isAnthropicAdaptive) {
|
||||
return Object.fromEntries(
|
||||
adaptiveEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
thinking: {
|
||||
type: "adaptive",
|
||||
},
|
||||
effort,
|
||||
case "@ai-sdk/google-vertex":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/google-vertex
|
||||
case "@ai-sdk/google":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/google-generative-ai
|
||||
if (id.includes("2.5")) {
|
||||
return {
|
||||
high: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 16000,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 24576,
|
||||
},
|
||||
},
|
||||
}
|
||||
}
|
||||
let levels = ["low", "high"]
|
||||
if (id.includes("3.1")) {
|
||||
levels = ["low", "medium", "high"]
|
||||
}
|
||||
|
||||
return Object.fromEntries(
|
||||
levels.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingLevel: effort,
|
||||
},
|
||||
},
|
||||
]),
|
||||
)
|
||||
|
||||
case "@ai-sdk/mistral":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/mistral
|
||||
return {}
|
||||
|
||||
case "@ai-sdk/cohere":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/cohere
|
||||
return {}
|
||||
|
||||
case "@ai-sdk/groq":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/groq
|
||||
const groqEffort = ["none", ...WIDELY_SUPPORTED_EFFORTS]
|
||||
return Object.fromEntries(
|
||||
groqEffort.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
reasoningEffort: effort,
|
||||
},
|
||||
]),
|
||||
)
|
||||
|
||||
case "@ai-sdk/perplexity":
|
||||
// https://v5.ai-sdk.dev/providers/ai-sdk-providers/perplexity
|
||||
return {}
|
||||
|
||||
case "@jerome-benoit/sap-ai-provider-v2":
|
||||
if (model.api.id.includes("anthropic")) {
|
||||
if (isAnthropicAdaptive) {
|
||||
return Object.fromEntries(
|
||||
adaptiveEfforts.map((effort) => [
|
||||
effort,
|
||||
{
|
||||
thinking: {
|
||||
type: "adaptive",
|
||||
},
|
||||
]),
|
||||
)
|
||||
}
|
||||
return {
|
||||
high: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: 16000,
|
||||
effort,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: 31999,
|
||||
},
|
||||
},
|
||||
}
|
||||
]),
|
||||
)
|
||||
}
|
||||
if (model.api.id.includes("gemini") && id.includes("2.5")) {
|
||||
return {
|
||||
high: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 16000,
|
||||
},
|
||||
return {
|
||||
high: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: 16000,
|
||||
},
|
||||
max: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 24576,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinking: {
|
||||
type: "enabled",
|
||||
budgetTokens: 31999,
|
||||
},
|
||||
}
|
||||
},
|
||||
}
|
||||
if (model.api.id.includes("gpt") || /\bo[1-9]/.test(model.api.id)) {
|
||||
return Object.fromEntries(WIDELY_SUPPORTED_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
}
|
||||
if (model.api.id.includes("gemini") && id.includes("2.5")) {
|
||||
return {
|
||||
high: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 16000,
|
||||
},
|
||||
},
|
||||
max: {
|
||||
thinkingConfig: {
|
||||
includeThoughts: true,
|
||||
thinkingBudget: 24576,
|
||||
},
|
||||
},
|
||||
}
|
||||
return {}
|
||||
}
|
||||
return {}
|
||||
}
|
||||
if (model.api.id.includes("gpt") || /\bo[1-9]/.test(model.api.id)) {
|
||||
return Object.fromEntries(WIDELY_SUPPORTED_EFFORTS.map((effort) => [effort, { reasoningEffort: effort }]))
|
||||
}
|
||||
return {}
|
||||
}
|
||||
return {}
|
||||
}
|
||||
|
||||
export function options(input: {
|
||||
model: Provider.Model
|
||||
sessionID: string
|
||||
providerOptions?: Record<string, any>
|
||||
}): Record<string, any> {
|
||||
const result: Record<string, any> = {}
|
||||
|
||||
// openai and providers using openai package should set store to false by default.
|
||||
if (
|
||||
input.model.providerID === "openai" ||
|
||||
input.model.api.npm === "@ai-sdk/openai" ||
|
||||
input.model.api.npm === "@ai-sdk/github-copilot"
|
||||
) {
|
||||
result["store"] = false
|
||||
}
|
||||
|
||||
export function options(input: {
|
||||
model: Provider.Model
|
||||
sessionID: string
|
||||
providerOptions?: Record<string, any>
|
||||
}): Record<string, any> {
|
||||
const result: Record<string, any> = {}
|
||||
|
||||
// openai and providers using openai package should set store to false by default.
|
||||
if (
|
||||
input.model.providerID === "openai" ||
|
||||
input.model.api.npm === "@ai-sdk/openai" ||
|
||||
input.model.api.npm === "@ai-sdk/github-copilot"
|
||||
) {
|
||||
result["store"] = false
|
||||
if (input.model.api.npm === "@openrouter/ai-sdk-provider") {
|
||||
result["usage"] = {
|
||||
include: true,
|
||||
}
|
||||
if (input.model.api.id.includes("gemini-3")) {
|
||||
result["reasoning"] = { effort: "high" }
|
||||
}
|
||||
}
|
||||
|
||||
if (input.model.api.npm === "@openrouter/ai-sdk-provider") {
|
||||
result["usage"] = {
|
||||
include: true,
|
||||
if (
|
||||
input.model.providerID === "baseten" ||
|
||||
(input.model.providerID === "opencode" && ["kimi-k2-thinking", "glm-4.6"].includes(input.model.api.id))
|
||||
) {
|
||||
result["chat_template_args"] = { enable_thinking: true }
|
||||
}
|
||||
|
||||
if (
|
||||
["zai", "zhipuai"].some((id) => input.model.providerID.includes(id)) &&
|
||||
input.model.api.npm === "@ai-sdk/openai-compatible"
|
||||
) {
|
||||
result["thinking"] = {
|
||||
type: "enabled",
|
||||
clear_thinking: false,
|
||||
}
|
||||
}
|
||||
|
||||
if (input.model.providerID === "openai" || input.providerOptions?.setCacheKey) {
|
||||
result["promptCacheKey"] = input.sessionID
|
||||
}
|
||||
|
||||
if (input.model.api.npm === "@ai-sdk/google" || input.model.api.npm === "@ai-sdk/google-vertex") {
|
||||
if (input.model.capabilities.reasoning) {
|
||||
result["thinkingConfig"] = {
|
||||
includeThoughts: true,
|
||||
}
|
||||
if (input.model.api.id.includes("gemini-3")) {
|
||||
result["reasoning"] = { effort: "high" }
|
||||
result["thinkingConfig"]["thinkingLevel"] = "high"
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
if (
|
||||
input.model.providerID === "baseten" ||
|
||||
(input.model.providerID === "opencode" && ["kimi-k2-thinking", "glm-4.6"].includes(input.model.api.id))
|
||||
) {
|
||||
result["chat_template_args"] = { enable_thinking: true }
|
||||
// Enable thinking by default for kimi-k2.5/k2p5 models using anthropic SDK
|
||||
const modelId = input.model.api.id.toLowerCase()
|
||||
if (
|
||||
(input.model.api.npm === "@ai-sdk/anthropic" || input.model.api.npm === "@ai-sdk/google-vertex/anthropic") &&
|
||||
(modelId.includes("k2p5") || modelId.includes("kimi-k2.5") || modelId.includes("kimi-k2p5"))
|
||||
) {
|
||||
result["thinking"] = {
|
||||
type: "enabled",
|
||||
budgetTokens: Math.min(16_000, Math.floor(input.model.limit.output / 2 - 1)),
|
||||
}
|
||||
}
|
||||
|
||||
if (
|
||||
["zai", "zhipuai"].some((id) => input.model.providerID.includes(id)) &&
|
||||
input.model.api.npm === "@ai-sdk/openai-compatible"
|
||||
) {
|
||||
result["thinking"] = {
|
||||
type: "enabled",
|
||||
clear_thinking: false,
|
||||
}
|
||||
}
|
||||
// Enable thinking for reasoning models on alibaba-cn (DashScope).
|
||||
// DashScope's OpenAI-compatible API requires `enable_thinking: true` in the request body
|
||||
// to return reasoning_content. Without it, models like kimi-k2.5, qwen-plus, qwen3, qwq,
|
||||
// deepseek-r1, etc. never output thinking/reasoning tokens.
|
||||
// Note: kimi-k2-thinking is excluded as it returns reasoning_content by default.
|
||||
if (
|
||||
input.model.providerID === "alibaba-cn" &&
|
||||
input.model.capabilities.reasoning &&
|
||||
input.model.api.npm === "@ai-sdk/openai-compatible" &&
|
||||
!modelId.includes("kimi-k2-thinking")
|
||||
) {
|
||||
result["enable_thinking"] = true
|
||||
}
|
||||
|
||||
if (input.model.providerID === "openai" || input.providerOptions?.setCacheKey) {
|
||||
result["promptCacheKey"] = input.sessionID
|
||||
}
|
||||
|
||||
if (input.model.api.npm === "@ai-sdk/google" || input.model.api.npm === "@ai-sdk/google-vertex") {
|
||||
if (input.model.capabilities.reasoning) {
|
||||
result["thinkingConfig"] = {
|
||||
includeThoughts: true,
|
||||
}
|
||||
if (input.model.api.id.includes("gemini-3")) {
|
||||
result["thinkingConfig"]["thinkingLevel"] = "high"
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Enable thinking by default for kimi-k2.5/k2p5 models using anthropic SDK
|
||||
const modelId = input.model.api.id.toLowerCase()
|
||||
if (
|
||||
(input.model.api.npm === "@ai-sdk/anthropic" || input.model.api.npm === "@ai-sdk/google-vertex/anthropic") &&
|
||||
(modelId.includes("k2p5") || modelId.includes("kimi-k2.5") || modelId.includes("kimi-k2p5"))
|
||||
) {
|
||||
result["thinking"] = {
|
||||
type: "enabled",
|
||||
budgetTokens: Math.min(16_000, Math.floor(input.model.limit.output / 2 - 1)),
|
||||
}
|
||||
}
|
||||
|
||||
// Enable thinking for reasoning models on alibaba-cn (DashScope).
|
||||
// DashScope's OpenAI-compatible API requires `enable_thinking: true` in the request body
|
||||
// to return reasoning_content. Without it, models like kimi-k2.5, qwen-plus, qwen3, qwq,
|
||||
// deepseek-r1, etc. never output thinking/reasoning tokens.
|
||||
// Note: kimi-k2-thinking is excluded as it returns reasoning_content by default.
|
||||
if (
|
||||
input.model.providerID === "alibaba-cn" &&
|
||||
input.model.capabilities.reasoning &&
|
||||
input.model.api.npm === "@ai-sdk/openai-compatible" &&
|
||||
!modelId.includes("kimi-k2-thinking")
|
||||
) {
|
||||
result["enable_thinking"] = true
|
||||
}
|
||||
|
||||
if (input.model.api.id.includes("gpt-5") && !input.model.api.id.includes("gpt-5-chat")) {
|
||||
if (!input.model.api.id.includes("gpt-5-pro")) {
|
||||
result["reasoningEffort"] = "medium"
|
||||
// Only inject reasoningSummary for providers that support it natively.
|
||||
// @ai-sdk/openai-compatible proxies (e.g. LiteLLM) do not understand this
|
||||
// parameter and return "Unknown parameter: 'reasoningSummary'".
|
||||
if (
|
||||
input.model.api.npm === "@ai-sdk/openai" ||
|
||||
input.model.api.npm === "@ai-sdk/azure" ||
|
||||
input.model.api.npm === "@ai-sdk/github-copilot"
|
||||
) {
|
||||
result["reasoningSummary"] = "auto"
|
||||
}
|
||||
}
|
||||
|
||||
// Only set textVerbosity for non-chat gpt-5.x models
|
||||
// Chat models (e.g. gpt-5.2-chat-latest) only support "medium" verbosity
|
||||
if (input.model.api.id.includes("gpt-5") && !input.model.api.id.includes("gpt-5-chat")) {
|
||||
if (!input.model.api.id.includes("gpt-5-pro")) {
|
||||
result["reasoningEffort"] = "medium"
|
||||
// Only inject reasoningSummary for providers that support it natively.
|
||||
// @ai-sdk/openai-compatible proxies (e.g. LiteLLM) do not understand this
|
||||
// parameter and return "Unknown parameter: 'reasoningSummary'".
|
||||
if (
|
||||
input.model.api.id.includes("gpt-5.") &&
|
||||
!input.model.api.id.includes("codex") &&
|
||||
!input.model.api.id.includes("-chat") &&
|
||||
input.model.providerID !== "azure"
|
||||
input.model.api.npm === "@ai-sdk/openai" ||
|
||||
input.model.api.npm === "@ai-sdk/azure" ||
|
||||
input.model.api.npm === "@ai-sdk/github-copilot"
|
||||
) {
|
||||
result["textVerbosity"] = "low"
|
||||
}
|
||||
|
||||
if (input.model.providerID.startsWith("opencode")) {
|
||||
result["promptCacheKey"] = input.sessionID
|
||||
result["include"] = ["reasoning.encrypted_content"]
|
||||
result["reasoningSummary"] = "auto"
|
||||
}
|
||||
}
|
||||
|
||||
if (input.model.providerID === "venice") {
|
||||
result["promptCacheKey"] = input.sessionID
|
||||
// Only set textVerbosity for non-chat gpt-5.x models
|
||||
// Chat models (e.g. gpt-5.2-chat-latest) only support "medium" verbosity
|
||||
if (
|
||||
input.model.api.id.includes("gpt-5.") &&
|
||||
!input.model.api.id.includes("codex") &&
|
||||
!input.model.api.id.includes("-chat") &&
|
||||
input.model.providerID !== "azure"
|
||||
) {
|
||||
result["textVerbosity"] = "low"
|
||||
}
|
||||
|
||||
if (input.model.providerID === "openrouter") {
|
||||
result["prompt_cache_key"] = input.sessionID
|
||||
if (input.model.providerID.startsWith("opencode")) {
|
||||
result["promptCacheKey"] = input.sessionID
|
||||
result["include"] = ["reasoning.encrypted_content"]
|
||||
result["reasoningSummary"] = "auto"
|
||||
}
|
||||
if (input.model.api.npm === "@ai-sdk/gateway") {
|
||||
result["gateway"] = {
|
||||
caching: "auto",
|
||||
}
|
||||
|
||||
if (input.model.providerID === "venice") {
|
||||
result["promptCacheKey"] = input.sessionID
|
||||
}
|
||||
|
||||
if (input.model.providerID === "openrouter") {
|
||||
result["prompt_cache_key"] = input.sessionID
|
||||
}
|
||||
if (input.model.api.npm === "@ai-sdk/gateway") {
|
||||
result["gateway"] = {
|
||||
caching: "auto",
|
||||
}
|
||||
}
|
||||
|
||||
return result
|
||||
}
|
||||
|
||||
export function smallOptions(model: Provider.Model) {
|
||||
if (
|
||||
model.providerID === "openai" ||
|
||||
model.api.npm === "@ai-sdk/openai" ||
|
||||
model.api.npm === "@ai-sdk/github-copilot"
|
||||
) {
|
||||
if (model.api.id.includes("gpt-5")) {
|
||||
if (model.api.id.includes("5.")) {
|
||||
return { store: false, reasoningEffort: "low" }
|
||||
}
|
||||
return { store: false, reasoningEffort: "minimal" }
|
||||
}
|
||||
return { store: false }
|
||||
}
|
||||
if (model.providerID === "google") {
|
||||
// gemini-3 uses thinkingLevel, gemini-2.5 uses thinkingBudget
|
||||
if (model.api.id.includes("gemini-3")) {
|
||||
return { thinkingConfig: { thinkingLevel: "minimal" } }
|
||||
}
|
||||
return { thinkingConfig: { thinkingBudget: 0 } }
|
||||
}
|
||||
if (model.providerID === "openrouter") {
|
||||
if (model.api.id.includes("google")) {
|
||||
return { reasoning: { enabled: false } }
|
||||
}
|
||||
return { reasoningEffort: "minimal" }
|
||||
}
|
||||
|
||||
if (model.providerID === "venice") {
|
||||
return { veniceParameters: { disableThinking: true } }
|
||||
}
|
||||
|
||||
return {}
|
||||
}
|
||||
|
||||
// Maps model ID prefix to provider slug used in providerOptions.
|
||||
// Example: "amazon/nova-2-lite" → "bedrock"
|
||||
const SLUG_OVERRIDES: Record<string, string> = {
|
||||
amazon: "bedrock",
|
||||
}
|
||||
|
||||
export function providerOptions(model: Provider.Model, options: { [x: string]: any }) {
|
||||
if (model.api.npm === "@ai-sdk/gateway") {
|
||||
// Gateway providerOptions are split across two namespaces:
|
||||
// - `gateway`: gateway-native routing/caching controls (order, only, byok, etc.)
|
||||
// - `<upstream slug>`: provider-specific model options (anthropic/openai/...)
|
||||
// We keep `gateway` as-is and route every other top-level option under the
|
||||
// model-derived upstream slug.
|
||||
const i = model.api.id.indexOf("/")
|
||||
const rawSlug = i > 0 ? model.api.id.slice(0, i) : undefined
|
||||
const slug = rawSlug ? (SLUG_OVERRIDES[rawSlug] ?? rawSlug) : undefined
|
||||
const gateway = options.gateway
|
||||
const rest = Object.fromEntries(Object.entries(options).filter(([k]) => k !== "gateway"))
|
||||
const has = Object.keys(rest).length > 0
|
||||
|
||||
const result: Record<string, any> = {}
|
||||
if (gateway !== undefined) result.gateway = gateway
|
||||
|
||||
if (has) {
|
||||
if (slug) {
|
||||
// Route model-specific options under the provider slug
|
||||
result[slug] = rest
|
||||
} else if (gateway && typeof gateway === "object" && !Array.isArray(gateway)) {
|
||||
result.gateway = { ...gateway, ...rest }
|
||||
} else {
|
||||
result.gateway = rest
|
||||
}
|
||||
}
|
||||
|
||||
return result
|
||||
}
|
||||
|
||||
export function smallOptions(model: Provider.Model) {
|
||||
if (
|
||||
model.providerID === "openai" ||
|
||||
model.api.npm === "@ai-sdk/openai" ||
|
||||
model.api.npm === "@ai-sdk/github-copilot"
|
||||
) {
|
||||
if (model.api.id.includes("gpt-5")) {
|
||||
if (model.api.id.includes("5.")) {
|
||||
return { store: false, reasoningEffort: "low" }
|
||||
const key = sdkKey(model.api.npm) ?? model.providerID
|
||||
// @ai-sdk/azure delegates to OpenAIChatLanguageModel which reads from
|
||||
// providerOptions["openai"], but OpenAIResponsesLanguageModel checks
|
||||
// "azure" first. Pass both so model options work on either code path.
|
||||
if (model.api.npm === "@ai-sdk/azure") {
|
||||
return { openai: options, azure: options }
|
||||
}
|
||||
return { [key]: options }
|
||||
}
|
||||
|
||||
export function maxOutputTokens(model: Provider.Model): number {
|
||||
return Math.min(model.limit.output, OUTPUT_TOKEN_MAX) || OUTPUT_TOKEN_MAX
|
||||
}
|
||||
|
||||
export function schema(model: Provider.Model, schema: JSONSchema.BaseSchema | JSONSchema7): JSONSchema7 {
|
||||
/*
|
||||
if (["openai", "azure"].includes(providerID)) {
|
||||
if (schema.type === "object" && schema.properties) {
|
||||
for (const [key, value] of Object.entries(schema.properties)) {
|
||||
if (schema.required?.includes(key)) continue
|
||||
schema.properties[key] = {
|
||||
anyOf: [
|
||||
value as JSONSchema.JSONSchema,
|
||||
{
|
||||
type: "null",
|
||||
},
|
||||
],
|
||||
}
|
||||
return { store: false, reasoningEffort: "minimal" }
|
||||
}
|
||||
return { store: false }
|
||||
}
|
||||
if (model.providerID === "google") {
|
||||
// gemini-3 uses thinkingLevel, gemini-2.5 uses thinkingBudget
|
||||
if (model.api.id.includes("gemini-3")) {
|
||||
return { thinkingConfig: { thinkingLevel: "minimal" } }
|
||||
}
|
||||
return { thinkingConfig: { thinkingBudget: 0 } }
|
||||
}
|
||||
if (model.providerID === "openrouter") {
|
||||
if (model.api.id.includes("google")) {
|
||||
return { reasoning: { enabled: false } }
|
||||
}
|
||||
return { reasoningEffort: "minimal" }
|
||||
}
|
||||
|
||||
if (model.providerID === "venice") {
|
||||
return { veniceParameters: { disableThinking: true } }
|
||||
}
|
||||
|
||||
return {}
|
||||
}
|
||||
*/
|
||||
|
||||
// Maps model ID prefix to provider slug used in providerOptions.
|
||||
// Example: "amazon/nova-2-lite" → "bedrock"
|
||||
const SLUG_OVERRIDES: Record<string, string> = {
|
||||
amazon: "bedrock",
|
||||
}
|
||||
// Convert integer enums to string enums for Google/Gemini
|
||||
if (model.providerID === "google" || model.api.id.includes("gemini")) {
|
||||
const isPlainObject = (node: unknown): node is Record<string, any> =>
|
||||
typeof node === "object" && node !== null && !Array.isArray(node)
|
||||
const hasCombiner = (node: unknown) =>
|
||||
isPlainObject(node) && (Array.isArray(node.anyOf) || Array.isArray(node.oneOf) || Array.isArray(node.allOf))
|
||||
const hasSchemaIntent = (node: unknown) => {
|
||||
if (!isPlainObject(node)) return false
|
||||
if (hasCombiner(node)) return true
|
||||
return [
|
||||
"type",
|
||||
"properties",
|
||||
"items",
|
||||
"prefixItems",
|
||||
"enum",
|
||||
"const",
|
||||
"$ref",
|
||||
"additionalProperties",
|
||||
"patternProperties",
|
||||
"required",
|
||||
"not",
|
||||
"if",
|
||||
"then",
|
||||
"else",
|
||||
].some((key) => key in node)
|
||||
}
|
||||
|
||||
export function providerOptions(model: Provider.Model, options: { [x: string]: any }) {
|
||||
if (model.api.npm === "@ai-sdk/gateway") {
|
||||
// Gateway providerOptions are split across two namespaces:
|
||||
// - `gateway`: gateway-native routing/caching controls (order, only, byok, etc.)
|
||||
// - `<upstream slug>`: provider-specific model options (anthropic/openai/...)
|
||||
// We keep `gateway` as-is and route every other top-level option under the
|
||||
// model-derived upstream slug.
|
||||
const i = model.api.id.indexOf("/")
|
||||
const rawSlug = i > 0 ? model.api.id.slice(0, i) : undefined
|
||||
const slug = rawSlug ? (SLUG_OVERRIDES[rawSlug] ?? rawSlug) : undefined
|
||||
const gateway = options.gateway
|
||||
const rest = Object.fromEntries(Object.entries(options).filter(([k]) => k !== "gateway"))
|
||||
const has = Object.keys(rest).length > 0
|
||||
const sanitizeGemini = (obj: any): any => {
|
||||
if (obj === null || typeof obj !== "object") {
|
||||
return obj
|
||||
}
|
||||
|
||||
const result: Record<string, any> = {}
|
||||
if (gateway !== undefined) result.gateway = gateway
|
||||
if (Array.isArray(obj)) {
|
||||
return obj.map(sanitizeGemini)
|
||||
}
|
||||
|
||||
if (has) {
|
||||
if (slug) {
|
||||
// Route model-specific options under the provider slug
|
||||
result[slug] = rest
|
||||
} else if (gateway && typeof gateway === "object" && !Array.isArray(gateway)) {
|
||||
result.gateway = { ...gateway, ...rest }
|
||||
const result: any = {}
|
||||
for (const [key, value] of Object.entries(obj)) {
|
||||
if (key === "enum" && Array.isArray(value)) {
|
||||
// Convert all enum values to strings
|
||||
result[key] = value.map((v) => String(v))
|
||||
// If we have integer type with enum, change type to string
|
||||
if (result.type === "integer" || result.type === "number") {
|
||||
result.type = "string"
|
||||
}
|
||||
} else if (typeof value === "object" && value !== null) {
|
||||
result[key] = sanitizeGemini(value)
|
||||
} else {
|
||||
result.gateway = rest
|
||||
result[key] = value
|
||||
}
|
||||
}
|
||||
|
||||
// Filter required array to only include fields that exist in properties
|
||||
if (result.type === "object" && result.properties && Array.isArray(result.required)) {
|
||||
result.required = result.required.filter((field: any) => field in result.properties)
|
||||
}
|
||||
|
||||
if (result.type === "array" && !hasCombiner(result)) {
|
||||
if (result.items == null) {
|
||||
result.items = {}
|
||||
}
|
||||
// Ensure items has a type only when it's still schema-empty.
|
||||
if (isPlainObject(result.items) && !hasSchemaIntent(result.items)) {
|
||||
result.items.type = "string"
|
||||
}
|
||||
}
|
||||
|
||||
// Remove properties/required from non-object types (Gemini rejects these)
|
||||
if (result.type && result.type !== "object" && !hasCombiner(result)) {
|
||||
delete result.properties
|
||||
delete result.required
|
||||
}
|
||||
|
||||
return result
|
||||
}
|
||||
|
||||
const key = sdkKey(model.api.npm) ?? model.providerID
|
||||
// @ai-sdk/azure delegates to OpenAIChatLanguageModel which reads from
|
||||
// providerOptions["openai"], but OpenAIResponsesLanguageModel checks
|
||||
// "azure" first. Pass both so model options work on either code path.
|
||||
if (model.api.npm === "@ai-sdk/azure") {
|
||||
return { openai: options, azure: options }
|
||||
}
|
||||
return { [key]: options }
|
||||
schema = sanitizeGemini(schema)
|
||||
}
|
||||
|
||||
export function maxOutputTokens(model: Provider.Model): number {
|
||||
return Math.min(model.limit.output, OUTPUT_TOKEN_MAX) || OUTPUT_TOKEN_MAX
|
||||
}
|
||||
|
||||
export function schema(model: Provider.Model, schema: JSONSchema.BaseSchema | JSONSchema7): JSONSchema7 {
|
||||
/*
|
||||
if (["openai", "azure"].includes(providerID)) {
|
||||
if (schema.type === "object" && schema.properties) {
|
||||
for (const [key, value] of Object.entries(schema.properties)) {
|
||||
if (schema.required?.includes(key)) continue
|
||||
schema.properties[key] = {
|
||||
anyOf: [
|
||||
value as JSONSchema.JSONSchema,
|
||||
{
|
||||
type: "null",
|
||||
},
|
||||
],
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
*/
|
||||
|
||||
// Convert integer enums to string enums for Google/Gemini
|
||||
if (model.providerID === "google" || model.api.id.includes("gemini")) {
|
||||
const isPlainObject = (node: unknown): node is Record<string, any> =>
|
||||
typeof node === "object" && node !== null && !Array.isArray(node)
|
||||
const hasCombiner = (node: unknown) =>
|
||||
isPlainObject(node) && (Array.isArray(node.anyOf) || Array.isArray(node.oneOf) || Array.isArray(node.allOf))
|
||||
const hasSchemaIntent = (node: unknown) => {
|
||||
if (!isPlainObject(node)) return false
|
||||
if (hasCombiner(node)) return true
|
||||
return [
|
||||
"type",
|
||||
"properties",
|
||||
"items",
|
||||
"prefixItems",
|
||||
"enum",
|
||||
"const",
|
||||
"$ref",
|
||||
"additionalProperties",
|
||||
"patternProperties",
|
||||
"required",
|
||||
"not",
|
||||
"if",
|
||||
"then",
|
||||
"else",
|
||||
].some((key) => key in node)
|
||||
}
|
||||
|
||||
const sanitizeGemini = (obj: any): any => {
|
||||
if (obj === null || typeof obj !== "object") {
|
||||
return obj
|
||||
}
|
||||
|
||||
if (Array.isArray(obj)) {
|
||||
return obj.map(sanitizeGemini)
|
||||
}
|
||||
|
||||
const result: any = {}
|
||||
for (const [key, value] of Object.entries(obj)) {
|
||||
if (key === "enum" && Array.isArray(value)) {
|
||||
// Convert all enum values to strings
|
||||
result[key] = value.map((v) => String(v))
|
||||
// If we have integer type with enum, change type to string
|
||||
if (result.type === "integer" || result.type === "number") {
|
||||
result.type = "string"
|
||||
}
|
||||
} else if (typeof value === "object" && value !== null) {
|
||||
result[key] = sanitizeGemini(value)
|
||||
} else {
|
||||
result[key] = value
|
||||
}
|
||||
}
|
||||
|
||||
// Filter required array to only include fields that exist in properties
|
||||
if (result.type === "object" && result.properties && Array.isArray(result.required)) {
|
||||
result.required = result.required.filter((field: any) => field in result.properties)
|
||||
}
|
||||
|
||||
if (result.type === "array" && !hasCombiner(result)) {
|
||||
if (result.items == null) {
|
||||
result.items = {}
|
||||
}
|
||||
// Ensure items has a type only when it's still schema-empty.
|
||||
if (isPlainObject(result.items) && !hasSchemaIntent(result.items)) {
|
||||
result.items.type = "string"
|
||||
}
|
||||
}
|
||||
|
||||
// Remove properties/required from non-object types (Gemini rejects these)
|
||||
if (result.type && result.type !== "object" && !hasCombiner(result)) {
|
||||
delete result.properties
|
||||
delete result.required
|
||||
}
|
||||
|
||||
return result
|
||||
}
|
||||
|
||||
schema = sanitizeGemini(schema)
|
||||
}
|
||||
|
||||
return schema as JSONSchema7
|
||||
}
|
||||
return schema as JSONSchema7
|
||||
}
|
||||
|
||||
@@ -1,4 +1,4 @@
|
||||
import { ProviderAuth } from "@/provider/auth"
|
||||
import { ProviderAuth } from "@/provider"
|
||||
import { Effect, Layer } from "effect"
|
||||
import { HttpApi, HttpApiBuilder, HttpApiEndpoint, HttpApiGroup, OpenApi } from "effect/unstable/httpapi"
|
||||
|
||||
|
||||
@@ -3,8 +3,8 @@ import { describeRoute, validator, resolver } from "hono-openapi"
|
||||
import z from "zod"
|
||||
import { Config } from "../../config"
|
||||
import { Provider } from "../../provider"
|
||||
import { ModelsDev } from "../../provider/models"
|
||||
import { ProviderAuth } from "../../provider/auth"
|
||||
import { ModelsDev } from "../../provider"
|
||||
import { ProviderAuth } from "../../provider"
|
||||
import { ProviderID } from "../../provider/schema"
|
||||
import { AppRuntime } from "../../effect/app-runtime"
|
||||
import { mapValues } from "remeda"
|
||||
|
||||
@@ -5,7 +5,7 @@ import * as Stream from "effect/Stream"
|
||||
import { streamText, wrapLanguageModel, type ModelMessage, type Tool, tool, jsonSchema } from "ai"
|
||||
import { mergeDeep, pipe } from "remeda"
|
||||
import { GitLabWorkflowLanguageModel } from "gitlab-ai-provider"
|
||||
import { ProviderTransform } from "@/provider/transform"
|
||||
import { ProviderTransform } from "@/provider"
|
||||
import { Config } from "@/config"
|
||||
import { Instance } from "@/project/instance"
|
||||
import type { Agent } from "@/agent/agent"
|
||||
|
||||
@@ -8,7 +8,7 @@ import { Snapshot } from "@/snapshot"
|
||||
import { SyncEvent } from "../sync"
|
||||
import { Database, NotFoundError, and, desc, eq, inArray, lt, or } from "@/storage"
|
||||
import { MessageTable, PartTable, SessionTable } from "./session.sql"
|
||||
import { ProviderError } from "@/provider/error"
|
||||
import { ProviderError } from "@/provider"
|
||||
import { iife } from "@/util/iife"
|
||||
import { errorMessage } from "@/util/error"
|
||||
import type { SystemError } from "bun"
|
||||
|
||||
@@ -1,6 +1,6 @@
|
||||
import type { Config } from "@/config"
|
||||
import type { Provider } from "@/provider"
|
||||
import { ProviderTransform } from "@/provider/transform"
|
||||
import { ProviderTransform } from "@/provider"
|
||||
import type { MessageV2 } from "./message-v2"
|
||||
|
||||
const COMPACTION_BUFFER = 20_000
|
||||
|
||||
@@ -12,7 +12,7 @@ import { ModelID, ProviderID } from "../provider/schema"
|
||||
import { type Tool as AITool, tool, jsonSchema, type ToolExecutionOptions, asSchema } from "ai"
|
||||
import { SessionCompaction } from "./compaction"
|
||||
import { Bus } from "../bus"
|
||||
import { ProviderTransform } from "../provider/transform"
|
||||
import { ProviderTransform } from "../provider"
|
||||
import { SystemPrompt } from "./system"
|
||||
import { Instruction } from "./instruction"
|
||||
import { Plugin } from "../plugin"
|
||||
|
||||
@@ -4,7 +4,7 @@ import fs from "fs/promises"
|
||||
import { Effect } from "effect"
|
||||
import { tmpdir } from "../fixture/fixture"
|
||||
import { Instance } from "../../src/project/instance"
|
||||
import { ProviderAuth } from "../../src/provider/auth"
|
||||
import { ProviderAuth } from "../../src/provider"
|
||||
import { ProviderID } from "../../src/provider/schema"
|
||||
|
||||
describe("plugin.auth-override", () => {
|
||||
|
||||
@@ -6,7 +6,7 @@ import { tmpdir } from "../fixture/fixture"
|
||||
import { Global } from "../../src/global"
|
||||
import { Instance } from "../../src/project/instance"
|
||||
import { Plugin } from "../../src/plugin/index"
|
||||
import { ModelsDev } from "../../src/provider/models"
|
||||
import { ModelsDev } from "../../src/provider"
|
||||
import { Provider } from "../../src/provider"
|
||||
import { ProviderID, ModelID } from "../../src/provider/schema"
|
||||
import { Filesystem } from "../../src/util"
|
||||
|
||||
@@ -1,5 +1,5 @@
|
||||
import { describe, expect, test } from "bun:test"
|
||||
import { ProviderTransform } from "../../src/provider/transform"
|
||||
import { ProviderTransform } from "../../src/provider"
|
||||
import { ModelID, ProviderID } from "../../src/provider/schema"
|
||||
|
||||
describe("ProviderTransform.options - setCacheKey", () => {
|
||||
|
||||
@@ -7,8 +7,8 @@ import { makeRuntime } from "../../src/effect/run-service"
|
||||
import { LLM } from "../../src/session/llm"
|
||||
import { Instance } from "../../src/project/instance"
|
||||
import { Provider } from "../../src/provider"
|
||||
import { ProviderTransform } from "../../src/provider/transform"
|
||||
import { ModelsDev } from "../../src/provider/models"
|
||||
import { ProviderTransform } from "../../src/provider"
|
||||
import { ModelsDev } from "../../src/provider"
|
||||
import { ProviderID, ModelID } from "../../src/provider/schema"
|
||||
import { Filesystem } from "../../src/util"
|
||||
import { tmpdir } from "../fixture/fixture"
|
||||
|
||||
Reference in New Issue
Block a user