@sleepy-ai/cli 0.1.6 → 0.1.8

package/src/provider/provider.ts

@@ -1,2098 +0,0 @@
-import z from "zod"
-import fsNode from "fs/promises"
-import os from "os"
-import fuzzysort from "fuzzysort"
-import { Config } from "../config"
-import { mapValues, mergeDeep, omit, pickBy, sortBy } from "remeda"
-import { NoSuchModelError, type Provider as SDK } from "ai"
-import { Log } from "../util"
-import { Npm } from "../npm"
-import { Hash } from "@sleepy-ai/shared/util/hash"
-import { Plugin } from "../plugin"
-import { NamedError } from "@sleepy-ai/shared/util/error"
-import { type LanguageModelV3 } from "@ai-sdk/provider"
-import * as ModelsDev from "./models"
-import { Auth } from "../auth"
-import { Env } from "../env"
-import { InstallationVersion } from "../installation/version"
-import { Flag } from "../flag/flag"
-import { zod } from "@/util/effect-zod"
-import { iife } from "@/util/iife"
-import { Global } from "../global"
-import path from "path"
-import { pathToFileURL } from "url"
-import { Effect, Layer, Context, Schema, Types } from "effect"
-import { EffectBridge } from "@/effect"
-import { InstanceState } from "@/effect"
-import { AppFileSystem } from "@sleepy-ai/shared/filesystem"
-import { isRecord } from "@/util/record"
-import { withStatics } from "@/util/schema"
-
-import * as ProviderTransform from "./transform"
-import { ModelID, ProviderID } from "./schema"
-import { createSessionChecker } from "./session-check"
-
-const log = Log.create({ service: "provider" })
-const DEFAULT_CONTEXT_WINDOW = 1_000_000
-// Reserved built-in model tiers: always resolve, falling back to the default
-// model when not configured in `model_groups` (zero-config never errors).
-const BUILTIN_TIERS = new Set(["ultra", "standard", "lite"])
-// F41: warn once per (providerID, modelID) when limit.context falls back to default
-const warnedContextDefaults = new Set<string>()
-
-export const DEFAULT_CHUNK_TIMEOUT = 480_000 // 8 minutes — bounds single-attempt SSE stall.
-// Tuned for sleepy-v2.5-pro on Sleepy Router whose cold-path TTFT after context
-// rebuild can dip to ~5 minutes silent. Reasoning models with multi-minute
-// thinking still emit partial chunks / heartbeats within this window. Override
-// per-provider via sleepycode.json's `chunkTimeout` config for tighter or looser
-// bounds.
-
-function shouldUseCopilotResponsesApi(modelID: string): boolean {
-  const match = /^gpt-(\d+)/.exec(modelID)
-  if (!match) return false
-  return Number(match[1]) >= 5 && !modelID.startsWith("gpt-5-mini")
-}
-
-function wrapSSE(res: Response, ms: number, ctl: AbortController) {
-  if (typeof ms !== "number" || ms <= 0) return res
-  if (!res.body) return res
-  if (!res.headers.get("content-type")?.includes("text/event-stream")) return res
-
-  const reader = res.body.getReader()
-  const body = new ReadableStream<Uint8Array>({
-    async pull(ctrl) {
-      const part = await new Promise<Awaited<ReturnType<typeof reader.read>>>((resolve, reject) => {
-        const id = setTimeout(() => {
-          const err = new Error("SSE read timed out")
-          ctl.abort(err)
-          void reader.cancel(err)
-          reject(err)
-        }, ms)
-
-        reader.read().then(
-          (part) => {
-            clearTimeout(id)
-            resolve(part)
-          },
-          (err) => {
-            clearTimeout(id)
-            reject(err)
-          },
-        )
-      })
-
-      if (part.done) {
-        ctrl.close()
-        return
-      }
-
-      ctrl.enqueue(part.value)
-    },
-    async cancel(reason) {
-      ctl.abort(reason)
-      await reader.cancel(reason)
-    },
-  })
-
-  return new Response(body, {
-    headers: new Headers(res.headers),
-    status: res.status,
-    statusText: res.statusText,
-  })
-}
-
-interface GatewayModel {
-  modelId: string
-  name: string
-  omniRouteModelId: string
-  contextWindow?: number
-  maxOutputLimit?: number
-  inputPrice?: number
-  outputPrice?: number
-  cacheReadPrice?: number
-  cacheWritePrice?: number
-}
-
-async function fetchModels(endpoint: string, token: string): Promise<GatewayModel[]> {
-  try {
-    const res = await fetch(`${endpoint}/api/v1/models`, {
-      headers: { Authorization: `Bearer ${token}` },
-    })
-    if (!res.ok) return []
-    const json = await res.json()
-    if (json && json.data && Array.isArray(json.data)) {
-      return json.data as GatewayModel[]
-    }
-    return []
-  } catch {
-    return []
-  }
-}
-
-type BundledSDK = {
-  languageModel(modelId: string): LanguageModelV3
-}
-
-const BUNDLED_PROVIDERS: Record<string, () => Promise<(opts: any) => BundledSDK>> = {
-  "@ai-sdk/amazon-bedrock": () => import("@ai-sdk/amazon-bedrock").then((m) => m.createAmazonBedrock),
-  "@ai-sdk/anthropic": () => import("@ai-sdk/anthropic").then((m) => m.createAnthropic),
-  "@ai-sdk/azure": () => import("@ai-sdk/azure").then((m) => m.createAzure),
-  "@ai-sdk/google": () => import("@ai-sdk/google").then((m) => m.createGoogleGenerativeAI),
-  "@ai-sdk/google-vertex": () => import("@ai-sdk/google-vertex").then((m) => m.createVertex),
-  "@ai-sdk/google-vertex/anthropic": () =>
-    import("@ai-sdk/google-vertex/anthropic").then((m) => m.createVertexAnthropic),
-  "@ai-sdk/openai": () => import("@ai-sdk/openai").then((m) => m.createOpenAI),
-  "@ai-sdk/openai-compatible": () => import("@ai-sdk/openai-compatible").then((m) => m.createOpenAICompatible),
-  "@openrouter/ai-sdk-provider": () => import("@openrouter/ai-sdk-provider").then((m) => m.createOpenRouter),
-  "@ai-sdk/xai": () => import("@ai-sdk/xai").then((m) => m.createXai),
-  "@ai-sdk/mistral": () => import("@ai-sdk/mistral").then((m) => m.createMistral),
-  "@ai-sdk/groq": () => import("@ai-sdk/groq").then((m) => m.createGroq),
-  "@ai-sdk/deepinfra": () => import("@ai-sdk/deepinfra").then((m) => m.createDeepInfra),
-  "@ai-sdk/cerebras": () => import("@ai-sdk/cerebras").then((m) => m.createCerebras),
-  "@ai-sdk/cohere": () => import("@ai-sdk/cohere").then((m) => m.createCohere),
-  "@ai-sdk/gateway": () => import("@ai-sdk/gateway").then((m) => m.createGateway),
-  "@ai-sdk/togetherai": () => import("@ai-sdk/togetherai").then((m) => m.createTogetherAI),
-  "@ai-sdk/perplexity": () => import("@ai-sdk/perplexity").then((m) => m.createPerplexity),
-  "@ai-sdk/vercel": () => import("@ai-sdk/vercel").then((m) => m.createVercel),
-  "@ai-sdk/alibaba": () => import("@ai-sdk/alibaba").then((m) => m.createAlibaba),
-  "gitlab-ai-provider": () => import("gitlab-ai-provider").then((m) => m.createGitLab),
-  "@ai-sdk/github-copilot": () => import("./sdk/copilot").then((m) => m.createOpenaiCompatible),
-  "venice-ai-sdk-provider": () => import("venice-ai-sdk-provider").then((m) => m.createVenice),
-}
-
-type CustomModelLoader = (sdk: any, modelID: string, options?: Record<string, any>) => Promise<any>
-type CustomVarsLoader = (options: Record<string, any>) => Record<string, string>
-type CustomDiscoverModels = () => Promise<Record<string, Model>>
-type CustomLoader = (provider: Info) => Effect.Effect<{
-  autoload: boolean
-  getModel?: CustomModelLoader
-  vars?: CustomVarsLoader
-  options?: Record<string, any>
-  discoverModels?: CustomDiscoverModels
-}>
-
-type CustomDep = {
-  auth: (id: string) => Effect.Effect<Auth.Info | undefined>
-  config: () => Effect.Effect<Config.Info>
-  env: () => Effect.Effect<Record<string, string | undefined>>
-  get: (key: string) => Effect.Effect<string | undefined>
-}
-
-function useLanguageModel(sdk: any) {
-  return sdk.responses === undefined && sdk.chat === undefined
-}
-
-function custom(dep: CustomDep): Record<string, CustomLoader> {
-  return {
-    anthropic: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "anthropic-beta": "interleaved-thinking-2025-05-14,fine-grained-tool-streaming-2025-05-14",
-          },
-        },
-      }),
-    opencode: Effect.fnUntraced(function* (input: Info) {
-      const env = yield* dep.env()
-      const hasKey = iife(() => {
-        if (input.env.some((item) => env[item])) return true
-        return false
-      })
-      const ok =
-        hasKey ||
-        Boolean(yield* dep.auth(input.id)) ||
-        Boolean((yield* dep.config()).provider?.["opencode"]?.options?.apiKey)
-
-      // Never surface the free/public tier (cost.input === 0). Without a
-      // subscription/key, hide the remaining (paid) models too — they can't be
-      // used unauthenticated. So: authenticated -> subscription models only,
-      // unauthenticated -> nothing.
-      for (const [key, value] of Object.entries(input.models)) {
-        if (!ok || value.cost.input === 0) delete input.models[key]
-      }
-
-      return {
-        autoload: Object.keys(input.models).length > 0,
-        options: {},
-      }
-    }),
-    openai: () =>
-      Effect.succeed({
-        autoload: false,
-        async getModel(sdk: any, modelID: string, _options?: Record<string, any>) {
-          return sdk.responses(modelID)
-        },
-        options: {},
-      }),
-    xai: () =>
-      Effect.succeed({
-        autoload: false,
-        async getModel(sdk: any, modelID: string, _options?: Record<string, any>) {
-          return sdk.responses(modelID)
-        },
-        options: {},
-      }),
-    "github-copilot": () =>
-      Effect.succeed({
-        autoload: false,
-        async getModel(sdk: any, modelID: string, _options?: Record<string, any>) {
-          if (useLanguageModel(sdk)) return sdk.languageModel(modelID)
-          return shouldUseCopilotResponsesApi(modelID) ? sdk.responses(modelID) : sdk.chat(modelID)
-        },
-        options: {},
-      }),
-    azure: Effect.fnUntraced(function* (provider: Info) {
-      const env = yield* dep.env()
-      const resource = iife(() => {
-        const name = provider.options?.resourceName
-        if (typeof name === "string" && name.trim() !== "") return name
-        return env["AZURE_RESOURCE_NAME"]
-      })
-
-      return {
-        autoload: false,
-        async getModel(sdk: any, modelID: string, options?: Record<string, any>) {
-          if (useLanguageModel(sdk)) return sdk.languageModel(modelID)
-          if (options?.["useCompletionUrls"]) {
-            return sdk.chat(modelID)
-          } else {
-            return sdk.responses(modelID)
-          }
-        },
-        options: {},
-        vars(_options) {
-          return {
-            ...(resource && { AZURE_RESOURCE_NAME: resource }),
-          }
-        },
-      }
-    }),
-    "azure-cognitive-services": Effect.fnUntraced(function* () {
-      const resourceName = yield* dep.get("AZURE_COGNITIVE_SERVICES_RESOURCE_NAME")
-      return {
-        autoload: false,
-        async getModel(sdk: any, modelID: string, options?: Record<string, any>) {
-          if (useLanguageModel(sdk)) return sdk.languageModel(modelID)
-          if (options?.["useCompletionUrls"]) {
-            return sdk.chat(modelID)
-          } else {
-            return sdk.responses(modelID)
-          }
-        },
-        options: {
-          baseURL: resourceName ? `https://${resourceName}.cognitiveservices.azure.com/openai` : undefined,
-        },
-      }
-    }),
-    "amazon-bedrock": Effect.fnUntraced(function* () {
-      const providerConfig = (yield* dep.config()).provider?.["amazon-bedrock"]
-      const auth = yield* dep.auth("amazon-bedrock")
-      const env = yield* dep.env()
-
-      // Region precedence: 1) config file, 2) env var, 3) default
-      const configRegion = providerConfig?.options?.region
-      const envRegion = env["AWS_REGION"]
-      const defaultRegion = configRegion ?? envRegion ?? "us-east-1"
-
-      // Profile: config file takes precedence over env var
-      const configProfile = providerConfig?.options?.profile
-      const envProfile = env["AWS_PROFILE"]
-      const profile = configProfile ?? envProfile
-
-      const awsAccessKeyId = env["AWS_ACCESS_KEY_ID"]
-
-      // TODO: Using process.env directly because Env.set only updates a process.env shallow copy,
-      // until the scope of the Env API is clarified (test only or runtime?)
-      const awsBearerToken = iife(() => {
-        const envToken = process.env.AWS_BEARER_TOKEN_BEDROCK
-        if (envToken) return envToken
-        if (auth?.type === "api") {
-          process.env.AWS_BEARER_TOKEN_BEDROCK = auth.key
-          return auth.key
-        }
-        return undefined
-      })
-
-      const awsWebIdentityTokenFile = env["AWS_WEB_IDENTITY_TOKEN_FILE"]
-
-      const containerCreds = Boolean(
-        process.env.AWS_CONTAINER_CREDENTIALS_RELATIVE_URI || process.env.AWS_CONTAINER_CREDENTIALS_FULL_URI,
-      )
-
-      if (!profile && !awsAccessKeyId && !awsBearerToken && !awsWebIdentityTokenFile && !containerCreds)
-        return { autoload: false }
-
-      const { fromNodeProviderChain } = yield* Effect.promise(() => import("@aws-sdk/credential-providers"))
-
-      const providerOptions: Record<string, any> = {
-        region: defaultRegion,
-      }
-
-      // Only use credential chain if no bearer token exists
-      // Bearer token takes precedence over credential chain (profiles, access keys, IAM roles, web identity tokens)
-      if (!awsBearerToken) {
-        // Build credential provider options (only pass profile if specified)
-        const credentialProviderOptions = profile ? { profile } : {}
-
-        providerOptions.credentialProvider = fromNodeProviderChain(credentialProviderOptions)
-      }
-
-      // Add custom endpoint if specified (endpoint takes precedence over baseURL)
-      const endpoint = providerConfig?.options?.endpoint ?? providerConfig?.options?.baseURL
-      if (endpoint) {
-        providerOptions.baseURL = endpoint
-      }
-
-      return {
-        autoload: true,
-        options: providerOptions,
-        async getModel(sdk: any, modelID: string, options?: Record<string, any>) {
-          // Skip region prefixing if model already has a cross-region inference profile prefix
-          // Models from models.dev may already include prefixes like us., eu., global., etc.
-          const crossRegionPrefixes = ["global.", "us.", "eu.", "jp.", "apac.", "au."]
-          if (crossRegionPrefixes.some((prefix) => modelID.startsWith(prefix))) {
-            return sdk.languageModel(modelID)
-          }
-
-          // Region resolution precedence (highest to lowest):
-          // 1. options.region from sleepycode.json provider config
-          // 2. defaultRegion from AWS_REGION environment variable
-          // 3. Default "us-east-1" (baked into defaultRegion)
-          const region = options?.region ?? defaultRegion
-
-          let regionPrefix = region.split("-")[0]
-
-          switch (regionPrefix) {
-            case "us": {
-              const modelRequiresPrefix = [
-                "nova-micro",
-                "nova-lite",
-                "nova-pro",
-                "nova-premier",
-                "nova-2",
-                "claude",
-                "deepseek",
-              ].some((m) => modelID.includes(m))
-              const isGovCloud = region.startsWith("us-gov")
-              if (modelRequiresPrefix && !isGovCloud) {
-                modelID = `${regionPrefix}.${modelID}`
-              }
-              break
-            }
-            case "eu": {
-              const regionRequiresPrefix = [
-                "eu-west-1",
-                "eu-west-2",
-                "eu-west-3",
-                "eu-north-1",
-                "eu-central-1",
-                "eu-south-1",
-                "eu-south-2",
-              ].some((r) => region.includes(r))
-              const modelRequiresPrefix = ["claude", "nova-lite", "nova-micro", "llama3", "pixtral"].some((m) =>
-                modelID.includes(m),
-              )
-              if (regionRequiresPrefix && modelRequiresPrefix) {
-                modelID = `${regionPrefix}.${modelID}`
-              }
-              break
-            }
-            case "ap": {
-              const isAustraliaRegion = ["ap-southeast-2", "ap-southeast-4"].includes(region)
-              const isTokyoRegion = region === "ap-northeast-1"
-              if (
-                isAustraliaRegion &&
-                ["anthropic.claude-sonnet-4-5", "anthropic.claude-haiku"].some((m) => modelID.includes(m))
-              ) {
-                regionPrefix = "au"
-                modelID = `${regionPrefix}.${modelID}`
-              } else if (isTokyoRegion) {
-                // Tokyo region uses jp. prefix for cross-region inference
-                const modelRequiresPrefix = ["claude", "nova-lite", "nova-micro", "nova-pro"].some((m) =>
-                  modelID.includes(m),
-                )
-                if (modelRequiresPrefix) {
-                  regionPrefix = "jp"
-                  modelID = `${regionPrefix}.${modelID}`
-                }
-              } else {
-                // Other APAC regions use apac. prefix
-                const modelRequiresPrefix = ["claude", "nova-lite", "nova-micro", "nova-pro"].some((m) =>
-                  modelID.includes(m),
-                )
-                if (modelRequiresPrefix) {
-                  regionPrefix = "apac"
-                  modelID = `${regionPrefix}.${modelID}`
-                }
-              }
-              break
-            }
-          }
-
-          return sdk.languageModel(modelID)
-        },
-      }
-    }),
-    llmgateway: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "HTTP-Referer": "https://www.sleepyai.org/",
-            "X-Title": "sleepy-code",
-            "X-Source": "sleepycode",
-          },
-        },
-      }),
-    openrouter: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "HTTP-Referer": "https://www.sleepyai.org/",
-            "X-Title": "sleepy-code",
-            "X-OpenRouter-Categories": "programming,programming-app,cli-agent",
-          },
-        },
-      }),
-    nvidia: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "HTTP-Referer": "https://www.sleepyai.org/",
-            "X-Title": "sleepy-code",
-          },
-        },
-      }),
-    vercel: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "http-referer": "https://www.sleepyai.org/",
-            "x-title": "sleepy-code",
-          },
-        },
-      }),
-    "google-vertex": Effect.fnUntraced(function* (provider: Info) {
-      const env = yield* dep.env()
-      const project =
-        provider.options?.project ?? env["GOOGLE_CLOUD_PROJECT"] ?? env["GCP_PROJECT"] ?? env["GCLOUD_PROJECT"]
-
-      const location = String(
-        provider.options?.location ??
-          env["GOOGLE_VERTEX_LOCATION"] ??
-          env["GOOGLE_CLOUD_LOCATION"] ??
-          env["VERTEX_LOCATION"] ??
-          "us-central1",
-      )
-
-      const autoload = Boolean(project)
-      if (!autoload) return { autoload: false }
-      return {
-        autoload: true,
-        vars(_options: Record<string, any>) {
-          const endpoint = location === "global" ? "aiplatform.googleapis.com" : `${location}-aiplatform.googleapis.com`
-          return {
-            ...(project && { GOOGLE_VERTEX_PROJECT: project }),
-            GOOGLE_VERTEX_LOCATION: location,
-            GOOGLE_VERTEX_ENDPOINT: endpoint,
-          }
-        },
-        options: {
-          project,
-          location,
-          fetch: async (input: RequestInfo | URL, init?: RequestInit) => {
-            const { GoogleAuth } = await import("google-auth-library")
-            const auth = new GoogleAuth()
-            const client = await auth.getApplicationDefault()
-            const token = await client.credential.getAccessToken()
-
-            const headers = new Headers(init?.headers)
-            headers.set("Authorization", `Bearer ${token.token}`)
-
-            return fetch(input, { ...init, headers })
-          },
-        },
-        async getModel(sdk: any, modelID: string) {
-          const id = String(modelID).trim()
-          return sdk.languageModel(id)
-        },
-      }
-    }),
-    "google-vertex-anthropic": Effect.fnUntraced(function* () {
-      const env = yield* dep.env()
-      const project = env["GOOGLE_CLOUD_PROJECT"] ?? env["GCP_PROJECT"] ?? env["GCLOUD_PROJECT"]
-      const location = env["GOOGLE_CLOUD_LOCATION"] ?? env["VERTEX_LOCATION"] ?? "global"
-      const autoload = Boolean(project)
-      if (!autoload) return { autoload: false }
-      return {
-        autoload: true,
-        options: {
-          project,
-          location,
-        },
-        async getModel(sdk: any, modelID) {
-          const id = String(modelID).trim()
-          return sdk.languageModel(id)
-        },
-      }
-    }),
-    "sap-ai-core": Effect.fnUntraced(function* () {
-      const auth = yield* dep.auth("sap-ai-core")
-      // TODO: Using process.env directly because Env.set only updates a shallow copy (not process.env),
-      // until the scope of the Env API is clarified (test only or runtime?)
-      const envServiceKey = iife(() => {
-        const envAICoreServiceKey = process.env.AICORE_SERVICE_KEY
-        if (envAICoreServiceKey) return envAICoreServiceKey
-        if (auth?.type === "api") {
-          process.env.AICORE_SERVICE_KEY = auth.key
-          return auth.key
-        }
-        return undefined
-      })
-      const deploymentId = process.env.AICORE_DEPLOYMENT_ID
-      const resourceGroup = process.env.AICORE_RESOURCE_GROUP
-
-      return {
-        autoload: !!envServiceKey,
-        options: envServiceKey ? { deploymentId, resourceGroup } : {},
-        async getModel(sdk: any, modelID: string) {
-          return sdk(modelID)
-        },
-      }
-    }),
-    zenmux: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "HTTP-Referer": "https://www.sleepyai.org/",
-            "X-Title": "sleepy-code",
-          },
-        },
-      }),
-    gitlab: Effect.fnUntraced(function* (input: Info) {
-      const {
-        VERSION: GITLAB_PROVIDER_VERSION,
-        isWorkflowModel,
-        discoverWorkflowModels,
-      } = yield* Effect.promise(() => import("gitlab-ai-provider"))
-
-      const instanceUrl = (yield* dep.get("GITLAB_INSTANCE_URL")) || "https://gitlab.com"
-
-      const auth = yield* dep.auth(input.id)
-      const apiKey = yield* Effect.sync(() => {
-        if (auth?.type === "oauth") return auth.access
-        if (auth?.type === "api") return auth.key
-        return undefined
-      })
-      const token = apiKey ?? (yield* dep.get("GITLAB_TOKEN"))
-
-      const providerConfig = (yield* dep.config()).provider?.["gitlab"]
-      const directory = yield* InstanceState.directory
-
-      const aiGatewayHeaders = {
-        "User-Agent": `sleepycode/${InstallationVersion} gitlab-ai-provider/${GITLAB_PROVIDER_VERSION} (${os.platform()} ${os.release()}; ${os.arch()})`,
-        "anthropic-beta": "context-1m-2025-08-07",
-        ...providerConfig?.options?.aiGatewayHeaders,
-      }
-
-      const featureFlags = {
-        duo_agent_platform_agentic_chat: true,
-        duo_agent_platform: true,
-        ...providerConfig?.options?.featureFlags,
-      }
-
-      return {
-        autoload: !!token,
-        options: {
-          instanceUrl,
-          apiKey: token,
-          aiGatewayHeaders,
-          featureFlags,
-        },
-        async getModel(sdk: any, modelID: string, options?: Record<string, any>) {
-          if (modelID.startsWith("duo-workflow-")) {
-            const workflowRef = typeof options?.workflowRef === "string" ? options.workflowRef : undefined
-            // Use the static mapping if it exists, otherwise use duo-workflow with selectedModelRef
-            const sdkModelID = isWorkflowModel(modelID) ? modelID : "duo-workflow"
-            const workflowDefinition =
-              typeof options?.workflowDefinition === "string" ? options.workflowDefinition : undefined
-            const model = sdk.workflowChat(sdkModelID, {
-              featureFlags,
-              workflowDefinition,
-            })
-            if (workflowRef) {
-              model.selectedModelRef = workflowRef
-            }
-            return model
-          }
-          return sdk.agenticChat(modelID, {
-            aiGatewayHeaders,
-            featureFlags,
-          })
-        },
-        async discoverModels(): Promise<Record<string, Model>> {
-          if (!apiKey) {
-            log.info("gitlab model discovery skipped: no apiKey")
-            return {}
-          }
-
-          try {
-            const token = apiKey
-            const getHeaders = (): Record<string, string> =>
-              auth?.type === "api" ? { "PRIVATE-TOKEN": token } : { Authorization: `Bearer ${token}` }
-
-            log.info("gitlab model discovery starting", { instanceUrl })
-            const result = await discoverWorkflowModels({ instanceUrl, getHeaders }, { workingDirectory: directory })
-
-            if (!result.models.length) {
-              log.info("gitlab model discovery skipped: no models found", {
-                project: result.project
-                  ? {
-                      id: result.project.id,
-                      path: result.project.pathWithNamespace,
-                    }
-                  : null,
-              })
-              return {}
-            }
-
-            const models: Record<string, Model> = {}
-            for (const m of result.models) {
-              if (!input.models[m.id]) {
-                models[m.id] = {
-                  id: ModelID.make(m.id),
-                  providerID: ProviderID.make("gitlab"),
-                  name: `Agent Platform (${m.name})`,
-                  family: "",
-                  api: {
-                    id: m.id,
-                    url: instanceUrl,
-                    npm: "gitlab-ai-provider",
-                  },
-                  status: "active",
-                  headers: {},
-                  options: { workflowRef: m.ref },
-                  cost: { input: 0, output: 0, cache: { read: 0, write: 0 } },
-                  limit: { context: m.context, output: m.output },
-                  capabilities: {
-                    temperature: false,
-                    reasoning: true,
-                    attachment: true,
-                    toolcall: true,
-                    input: {
-                      text: true,
-                      audio: false,
-                      image: true,
-                      video: false,
-                      pdf: true,
-                    },
-                    output: {
-                      text: true,
-                      audio: false,
-                      image: false,
-                      video: false,
-                      pdf: false,
-                    },
-                    interleaved: false,
-                  },
-                  release_date: "",
-                  variants: {},
-                }
-              }
-            }
-
-            log.info("gitlab model discovery complete", {
-              count: Object.keys(models).length,
-              models: Object.keys(models),
-            })
-            return models
-          } catch (e) {
-            log.warn("gitlab model discovery failed", { error: e })
-            return {}
-          }
-        },
-      }
-    }),
-    "cloudflare-workers-ai": Effect.fnUntraced(function* (input: Info) {
-      // When baseURL is already configured (e.g. corporate config routing through a proxy/gateway),
-      // skip the account ID check because the URL is already fully specified.
-      if (input.options?.baseURL) return { autoload: false }
-
-      const auth = yield* dep.auth(input.id)
-      const env = yield* dep.env()
-      const accountId = env["CLOUDFLARE_ACCOUNT_ID"] || (auth?.type === "api" ? auth.metadata?.accountId : undefined)
-      if (!accountId)
-        return {
-          autoload: false,
-          async getModel() {
-            throw new Error(
-              "CLOUDFLARE_ACCOUNT_ID is missing. Set it with: export CLOUDFLARE_ACCOUNT_ID=<your-account-id>",
-            )
-          },
-        }
-
-      const apiKey = yield* Effect.gen(function* () {
-        const envToken = env["CLOUDFLARE_API_KEY"]
-        if (envToken) return envToken
-        if (auth?.type === "api") return auth.key
-        return undefined
-      })
-
-      return {
-        autoload: !!apiKey,
-        options: {
-          apiKey,
-          headers: {
-            "User-Agent": `sleepycode/${InstallationVersion} cloudflare-workers-ai (${os.platform()} ${os.release()}; ${os.arch()})`,
-          },
-        },
-        async getModel(sdk: any, modelID: string) {
-          return sdk.languageModel(modelID)
-        },
-        vars(_options) {
-          return {
-            CLOUDFLARE_ACCOUNT_ID: accountId,
-          }
-        },
-      }
-    }),
-    "cloudflare-ai-gateway": Effect.fnUntraced(function* (input: Info) {
-      // When baseURL is already configured (e.g. corporate config), skip the ID checks.
-      if (input.options?.baseURL) return { autoload: false }
-
-      const auth = yield* dep.auth(input.id)
-      const env = yield* dep.env()
-      const accountId = env["CLOUDFLARE_ACCOUNT_ID"] || (auth?.type === "api" ? auth.metadata?.accountId : undefined)
-      const gateway = env["CLOUDFLARE_GATEWAY_ID"] || (auth?.type === "api" ? auth.metadata?.gatewayId : undefined)
-
-      if (!accountId || !gateway) {
-        const missing = [
-          !accountId ? "CLOUDFLARE_ACCOUNT_ID" : undefined,
-          !gateway ? "CLOUDFLARE_GATEWAY_ID" : undefined,
-        ].filter((x): x is string => Boolean(x))
-        return {
-          autoload: false,
-          async getModel() {
-            throw new Error(
-              `${missing.join(" and ")} missing. Set with: ${missing.map((x) => `export ${x}=<value>`).join(" && ")}`,
-            )
-          },
-        }
-      }
-
-      // Get API token from env or auth - required for authenticated gateways
-      const apiToken = yield* Effect.gen(function* () {
-        const envToken = env["CLOUDFLARE_API_TOKEN"] || env["CF_AIG_TOKEN"]
-        if (envToken) return envToken
-        if (auth?.type === "api") return auth.key
-        return undefined
-      })
-
-      if (!apiToken) {
-        throw new Error(
-          "CLOUDFLARE_API_TOKEN (or CF_AIG_TOKEN) is required for Cloudflare AI Gateway. " +
-            "Set it via environment variable or run `sleepycode auth cloudflare-ai-gateway`.",
-        )
-      }
-
-      // Use official ai-gateway-provider package (v2.x for AI SDK v5 compatibility)
-      const { createAiGateway } = yield* Effect.promise(() => import("ai-gateway-provider"))
-      const { createUnified } = yield* Effect.promise(() => import("ai-gateway-provider/providers/unified"))
-
-      const metadata = iife(() => {
-        if (input.options?.metadata) return input.options.metadata
-        try {
-          return JSON.parse(input.options?.headers?.["cf-aig-metadata"])
-        } catch {
-          return undefined
-        }
-      })
-      const opts = {
-        metadata,
-        cacheTtl: input.options?.cacheTtl,
-        cacheKey: input.options?.cacheKey,
-        skipCache: input.options?.skipCache,
-        collectLog: input.options?.collectLog,
-        headers: {
-          "User-Agent": `sleepycode/${InstallationVersion} cloudflare-ai-gateway (${os.platform()} ${os.release()}; ${os.arch()})`,
-        },
-      }
-
-      const aigateway = createAiGateway({
-        accountId,
-        gateway,
-        apiKey: apiToken,
-        ...(Object.values(opts).some((v) => v !== undefined) ? { options: opts } : {}),
-      })
-      const unified = createUnified()
-
-      return {
-        autoload: true,
-        async getModel(_sdk: any, modelID: string, _options?: Record<string, any>) {
-          // Model IDs use Unified API format: provider/model (e.g., "anthropic/claude-sonnet-4-5")
-          return aigateway(unified(modelID))
-        },
-        options: {},
-      }
-    }),
-    cerebras: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "X-Cerebras-3rd-Party-Integration": "sleepycode",
-          },
-        },
-      }),
-    kilo: () =>
-      Effect.succeed({
-        autoload: false,
-        options: {
-          headers: {
-            "HTTP-Referer": "https://www.sleepyai.org/",
-            "X-Title": "sleepy-code",
-          },
-        },
-      }),
-  }
-}
-
-const ProviderApiInfo = Schema.Struct({
-  id: Schema.String,
-  url: Schema.String,
-  npm: Schema.String,
-})
-
-const ProviderModalities = Schema.Struct({
-  text: Schema.Boolean,
-  audio: Schema.Boolean,
-  image: Schema.Boolean,
-  video: Schema.Boolean,
-  pdf: Schema.Boolean,
-})
-
-const ProviderInterleaved = Schema.Union([
-  Schema.Boolean,
-  Schema.Struct({
-    field: Schema.Literals(["reasoning_content", "reasoning_details"]),
-  }),
-])
-
-const ProviderCapabilities = Schema.Struct({
-  temperature: Schema.Boolean,
-  reasoning: Schema.Boolean,
-  attachment: Schema.Boolean,
-  toolcall: Schema.Boolean,
-  input: ProviderModalities,
-  output: ProviderModalities,
-  interleaved: ProviderInterleaved,
-})
-
-const ProviderCacheCost = Schema.Struct({
-  read: Schema.Number,
-  write: Schema.Number,
-})
-
-const ProviderCost = Schema.Struct({
-  input: Schema.Number,
-  output: Schema.Number,
-  cache: ProviderCacheCost,
-  experimentalOver200K: Schema.optional(
-    Schema.Struct({
-      input: Schema.Number,
-      output: Schema.Number,
-      cache: ProviderCacheCost,
-    }),
-  ),
-})
-
-const ProviderLimit = Schema.Struct({
-  context: Schema.Number,
-  input: Schema.optional(Schema.Number),
-  output: Schema.Number,
-})
-
-export const Model = Schema.Struct({
-  id: ModelID,
-  providerID: ProviderID,
-  api: ProviderApiInfo,
-  name: Schema.String,
-  family: Schema.optional(Schema.String),
-  capabilities: ProviderCapabilities,
-  cost: ProviderCost,
-  limit: ProviderLimit,
-  status: Schema.Literals(["alpha", "beta", "deprecated", "active"]),
-  options: Schema.Record(Schema.String, Schema.Any),
-  headers: Schema.Record(Schema.String, Schema.String),
-  release_date: Schema.String,
-  variants: Schema.optional(Schema.Record(Schema.String, Schema.Record(Schema.String, Schema.Any))),
-  cacheTTL: Schema.optional(Schema.Number),
-  cachePromptTTL: Schema.optional(Schema.Literals(["5m", "1h"])),
-})
-  .annotate({ identifier: "Model" })
-  .pipe(withStatics((s) => ({ zod: zod(s) })))
-export type Model = Types.DeepMutable<Schema.Schema.Type<typeof Model>>
-
-export const Info = Schema.Struct({
-  id: ProviderID,
-  name: Schema.String,
-  source: Schema.Literals(["env", "config", "custom", "api"]),
-  env: Schema.Array(Schema.String),
-  key: Schema.optional(Schema.String),
-  options: Schema.Record(Schema.String, Schema.Any),
-  models: Schema.Record(Schema.String, Model),
-})
-  .annotate({ identifier: "Provider" })
-  .pipe(withStatics((s) => ({ zod: zod(s) })))
-export type Info = Types.DeepMutable<Schema.Schema.Type<typeof Info>>
-
-const DefaultModelIDs = Schema.Record(Schema.String, Schema.String)
-
-export const ListResult = Schema.Struct({
-  all: Schema.Array(Info),
-  default: DefaultModelIDs,
-  connected: Schema.Array(Schema.String),
-}).pipe(withStatics((s) => ({ zod: zod(s) })))
-export type ListResult = Types.DeepMutable<Schema.Schema.Type<typeof ListResult>>
-
-export const ConfigProvidersResult = Schema.Struct({
-  providers: Schema.Array(Info),
-  default: DefaultModelIDs,
-}).pipe(withStatics((s) => ({ zod: zod(s) })))
-export type ConfigProvidersResult = Types.DeepMutable<Schema.Schema.Type<typeof ConfigProvidersResult>>
-
-export function defaultModelIDs<T extends { models: Record<string, { id: string }> }>(providers: Record<string, T>) {
-  return mapValues(providers, (item) => sort(Object.values(item.models))[0].id)
-}
-
-export interface Interface {
-  readonly list: () => Effect.Effect<Record<ProviderID, Info>>
-  readonly getProvider: (providerID: ProviderID) => Effect.Effect<Info>
-  readonly getModel: (providerID: ProviderID, modelID: ModelID) => Effect.Effect<Model>
-  readonly getLanguage: (model: Model) => Effect.Effect<LanguageModelV3>
-  readonly closest: (
-    providerID: ProviderID,
-    query: string[],
-  ) => Effect.Effect<{ providerID: ProviderID; modelID: string } | undefined>
-  readonly getSmallModel: (providerID: ProviderID) => Effect.Effect<Model | undefined>
-  readonly resolveModelRef: (ref: string, contextProviderID?: ProviderID) => Effect.Effect<Model>
-  readonly defaultModel: () => Effect.Effect<{ providerID: ProviderID; modelID: ModelID }>
-}
-
-interface State {
-  models: Map<string, LanguageModelV3>
-  providers: Record<ProviderID, Info>
-  sdk: Map<string, BundledSDK>
-  modelLoaders: Record<string, CustomModelLoader>
-  varsLoaders: Record<string, CustomVarsLoader>
-}
-
-export class Service extends Context.Service<Service, Interface>()("@opencode/Provider") {}
-
-function cost(c: ModelsDev.Model["cost"]): Model["cost"] {
-  const result: Model["cost"] = {
-    input: c?.input ?? 0,
-    output: c?.output ?? 0,
-    cache: {
-      read: c?.cache_read ?? 0,
-      write: c?.cache_write ?? 0,
-    },
-  }
-  if (c?.context_over_200k) {
-    result.experimentalOver200K = {
-      cache: {
-        read: c.context_over_200k.cache_read ?? 0,
-        write: c.context_over_200k.cache_write ?? 0,
-      },
-      input: c.context_over_200k.input,
-      output: c.context_over_200k.output,
-    }
-  }
-  return result
-}
-
-function fromModelsDevModel(provider: ModelsDev.Provider, model: ModelsDev.Model): Model {
-  const base: Model = {
-    id: ModelID.make(model.id),
-    providerID: ProviderID.make(provider.id),
-    name: model.name,
-    family: model.family,
-    api: {
-      id: model.id,
-      url: model.provider?.api ?? provider.api ?? "",
-      npm: model.provider?.npm ?? provider.npm ?? "@ai-sdk/openai-compatible",
-    },
-    status: model.status ?? "active",
-    headers: {},
-    options: {},
-    cost: cost(model.cost),
-    limit: {
-      context: model.limit.context,
-      input: model.limit.input,
-      output: model.limit.output,
-    },
-    capabilities: {
-      temperature: model.temperature ?? false,
-      reasoning: model.reasoning ?? false,
-      attachment: model.attachment ?? false,
-      toolcall: model.tool_call ?? true,
-      input: {
-        text: model.modalities?.input?.includes("text") ?? false,
-        audio: model.modalities?.input?.includes("audio") ?? false,
-        image: model.modalities?.input?.includes("image") ?? false,
-        video: model.modalities?.input?.includes("video") ?? false,
-        pdf: model.modalities?.input?.includes("pdf") ?? false,
-      },
-      output: {
-        text: model.modalities?.output?.includes("text") ?? false,
-        audio: model.modalities?.output?.includes("audio") ?? false,
-        image: model.modalities?.output?.includes("image") ?? false,
-        video: model.modalities?.output?.includes("video") ?? false,
-        pdf: model.modalities?.output?.includes("pdf") ?? false,
-      },
-      interleaved: model.interleaved ?? false,
-    },
-    release_date: model.release_date ?? "",
-    variants: {},
-  }
-
-  return {
-    ...base,
-    variants: mapValues(ProviderTransform.variants(base), (v) => v),
-  }
-}
-
-export function fromModelsDevProvider(provider: ModelsDev.Provider): Info {
-  const models: Record<string, Model> = {}
-  for (const [key, model] of Object.entries(provider.models)) {
-    models[key] = fromModelsDevModel(provider, model)
-    for (const [mode, opts] of Object.entries(model.experimental?.modes ?? {})) {
-      const id = `${model.id}-${mode}`
-      const base = fromModelsDevModel(provider, model)
-      models[id] = {
-        ...base,
-        id: ModelID.make(id),
-        name: `${model.name} ${mode[0].toUpperCase()}${mode.slice(1)}`,
-        cost: opts.cost ? mergeDeep(base.cost, cost(opts.cost)) : base.cost,
-        options: opts.provider?.body
-          ? Object.fromEntries(
-              Object.entries(opts.provider.body).map(([k, v]) => [
-                k.replace(/_([a-z])/g, (_, c) => c.toUpperCase()),
-                v,
-              ]),
-            )
-          : base.options,
-        headers: opts.provider?.headers ?? base.headers,
-      }
-    }
-  }
-  return {
-    id: ProviderID.make(provider.id),
-    source: "custom",
-    name: provider.name,
-    env: provider.env ?? [],
-    options: {},
-    models,
-  }
-}
-
-const layer: Layer.Layer<
-  Service,
-  never,
-  Config.Service | Auth.Service | Plugin.Service | AppFileSystem.Service | Env.Service
-> = Layer.effect(
-  Service,
-  Effect.gen(function* () {
-    const fs = yield* AppFileSystem.Service
-    const config = yield* Config.Service
-    const auth = yield* Auth.Service
-    const env = yield* Env.Service
-    const plugin = yield* Plugin.Service
-
-    const state = yield* InstanceState.make<State>(() =>
-      Effect.gen(function* () {
-        using _ = log.time("state")
-        const bridge = yield* EffectBridge.make()
-        const cfg = yield* config.get()
-        const modelsDev = yield* Effect.promise(() => ModelsDev.get())
-        const database = mapValues(modelsDev, fromModelsDevProvider)
-
-        let sleepyCredentials: { token: string; dashboardUrl: string } | null = null
-
-        // Inject Sleepy Gateway provider from ~/.sleepy/gateway.json if credentials exist
-        try {
-          const sleepyConfigPath = path.join(Global.Path.config, "gateway.json")
-          const sleepyConfigRaw = yield* Effect.promise(() =>
-            fsNode.readFile(sleepyConfigPath, "utf-8").catch(() => null as string | null),
-          )
-          if (sleepyConfigRaw) {
-            const sleepyConfig = JSON.parse(sleepyConfigRaw)
-            const endpoint = sleepyConfig.endpoint
-            let token = sleepyConfig.access_token || sleepyConfig.token
-            const dashboardUrl = sleepyConfig.dashboard_url || "https://www.sleepyai.org"
-
-            if (token && sleepyConfig.refresh_token && sleepyConfig.expires_at && Date.now() > sleepyConfig.expires_at) {
-              try {
-                const refreshRes = yield* Effect.promise(() =>
-                  fetch(`${dashboardUrl}/api/auth/token/refresh`, {
-                    method: "POST",
-                    headers: { "Content-Type": "application/json" },
-                    body: JSON.stringify({ refresh_token: sleepyConfig.refresh_token }),
-                  }).then(async (r) => (r.ok ? r.json() : null)),
-                )
-                if (refreshRes) {
-                  token = refreshRes.access_token
-                  const updated = {
-                    ...sleepyConfig,
-                    access_token: refreshRes.access_token,
-                    refresh_token: refreshRes.refresh_token,
-                    expires_at: Date.now() + (refreshRes.expires_in ?? 3600) * 1000,
-                  }
-                  const tmpPath = sleepyConfigPath + ".tmp"
-                  yield* Effect.promise(() =>
-                    fsNode.writeFile(tmpPath, JSON.stringify(updated, null, 2)).then(() => fsNode.rename(tmpPath, sleepyConfigPath)).catch((err) => {
-                      log.error("failed to persist refreshed token — next startup will re-refresh", { error: err.message })
-                    }),
-                  )
-                }
-              } catch {}
-            }
-
-            if (endpoint && token) {
-              sleepyCredentials = { token, dashboardUrl }
-              const sleepyProviderID = ProviderID.make("sleepy")
-              const baseUrl = `${dashboardUrl}/api/v1`
-              const authHeader = `Bearer ${token}`
-
-              const fetchedModels = yield* Effect.promise(() => fetchModels(dashboardUrl, token))
-
-              let sleepyModels: Record<string, Model> = {}
-
-              if (fetchedModels.length > 0) {
-                for (const m of fetchedModels) {
-                  sleepyModels[m.modelId] = {
-                    id: ModelID.make(m.modelId),
-                    providerID: sleepyProviderID,
-                    name: m.name,
-                    family: "sleepy",
-                    api: {
-                      id: m.omniRouteModelId,
-                      url: baseUrl,
-                      npm: "@ai-sdk/openai-compatible",
-                    },
-                    status: "active",
-                    headers: { Authorization: authHeader },
-                    options: {},
-                    cost: {
-                      input: m.inputPrice ?? 0.0015,
-                      output: m.outputPrice ?? 0.005,
-                      cache: {
-                        read: m.cacheReadPrice ?? 0.00075,
-                        write: m.cacheWritePrice ?? 0.0015,
-                      },
-                    },
-                    limit: {
-                      context: m.contextWindow ?? 128000,
-                      output: m.maxOutputLimit ?? 4096,
-                    },
-                    capabilities: {
-                      temperature: true,
-                      reasoning: true,
-                      attachment: true,
-                      toolcall: true,
-                      input: { text: true, audio: false, image: true, video: false, pdf: true },
-                      output: { text: true, audio: false, image: false, video: false, pdf: false },
-                      interleaved: false,
-                    },
-                    release_date: "",
-                    variants: {},
-                  }
-                }
-              } else {
-                sleepyModels = {
-                  smart: {
-                    id: ModelID.make("smart"),
-                    providerID: sleepyProviderID,
-                    name: "Sleepy Smart Routing (Auto)",
-                    family: "sleepy",
-                    api: { id: "smart", url: baseUrl, npm: "@ai-sdk/openai-compatible" },
-                    status: "active",
-                    headers: { Authorization: authHeader },
-                    options: {},
-                    cost: { input: 0.0015, output: 0.005, cache: { read: 0, write: 0 } },
-                    limit: { context: 128000, output: 4096 },
-                    capabilities: {
-                      temperature: true, reasoning: true, attachment: true, toolcall: true,
-                      input: { text: true, audio: false, image: true, video: false, pdf: true },
-                      output: { text: true, audio: false, image: false, video: false, pdf: false },
-                      interleaved: false,
-                    },
-                    release_date: "",
-                    variants: {},
-                  },
-                  cheap: {
-                    id: ModelID.make("cheap"),
-                    providerID: sleepyProviderID,
-                    name: "Sleepy Eco Route (Low Cost)",
-                    family: "sleepy",
-                    api: { id: "cheap", url: baseUrl, npm: "@ai-sdk/openai-compatible" },
-                    status: "active",
-                    headers: { Authorization: authHeader },
-                    options: {},
-                    cost: { input: 0.00015, output: 0.0006, cache: { read: 0, write: 0 } },
-                    limit: { context: 128000, output: 4096 },
-                    capabilities: {
-                      temperature: true, reasoning: false, attachment: true, toolcall: true,
-                      input: { text: true, audio: false, image: true, video: false, pdf: true },
-                      output: { text: true, audio: false, image: false, video: false, pdf: false },
-                      interleaved: false,
-                    },
-                    release_date: "",
-                    variants: {},
-                  },
-                  fast: {
-                    id: ModelID.make("fast"),
-                    providerID: sleepyProviderID,
-                    name: "Sleepy Sprint Route (Low Latency)",
-                    family: "sleepy",
-                    api: { id: "fast", url: baseUrl, npm: "@ai-sdk/openai-compatible" },
-                    status: "active",
-                    headers: { Authorization: authHeader },
-                    options: {},
-                    cost: { input: 0.00015, output: 0.0006, cache: { read: 0, write: 0 } },
-                    limit: { context: 128000, output: 4096 },
-                    capabilities: {
-                      temperature: true, reasoning: false, attachment: true, toolcall: true,
-                      input: { text: true, audio: false, image: true, video: false, pdf: true },
-                      output: { text: true, audio: false, image: false, video: false, pdf: false },
-                      interleaved: false,
-                    },
-                    release_date: "",
-                    variants: {},
-                  },
-                  high: {
-                    id: ModelID.make("high"),
-                    providerID: sleepyProviderID,
-                    name: "Sleepy Reasoning Route (High Quality)",
-                    family: "sleepy",
-                    api: { id: "high", url: baseUrl, npm: "@ai-sdk/openai-compatible" },
-                    status: "active",
-                    headers: { Authorization: authHeader },
-                    options: {},
-                    cost: { input: 0.003, output: 0.015, cache: { read: 0, write: 0 } },
-                    limit: { context: 128000, output: 4096 },
-                    capabilities: {
-                      temperature: true, reasoning: true, attachment: true, toolcall: true,
-                      input: { text: true, audio: false, image: true, video: false, pdf: true },
-                      output: { text: true, audio: false, image: false, video: false, pdf: false },
-                      interleaved: false,
-                    },
-                    release_date: "",
-                    variants: {},
-                  },
-                }
-              }
-
-              database[sleepyProviderID] = {
-                id: sleepyProviderID,
-                source: "config",
-                name: "Sleepy Gateway",
-                env: [],
-                options: {
-                  baseURL: baseUrl,
-                  headers: {
-                    Authorization: authHeader,
-                  },
-                },
-                models: sleepyModels,
-              }
-              log.info("sleepy provider injected from config", { endpoint, dashboardUrl })
-
-              // Start background session polling — checks every 5 minutes if the session is still valid.
-              // Network errors are ignored (offline = still valid). 401/403 = session revoked/expired.
-              // Proactively refreshes tokens before expiry to prevent mid-session logouts.
-              const checker = createSessionChecker({ dashboardUrl, token, configPath: sleepyConfigPath })
-              checker.start()
-            }
-          }
-        } catch {
-          // Config file missing or malformed — silently skip sleepy provider injection
-        }
-
-        const providers: Record<ProviderID, Info> = {} as Record<ProviderID, Info>
-        const languages = new Map<string, LanguageModelV3>()
-        const modelLoaders: {
-          [providerID: string]: CustomModelLoader
-        } = {}
-        const varsLoaders: {
-          [providerID: string]: CustomVarsLoader
-        } = {}
-        const sdk = new Map<string, BundledSDK>()
-        const discoveryLoaders: {
-          [providerID: string]: CustomDiscoverModels
-        } = {}
-        const dep = {
-          auth: (id: string) => auth.get(id).pipe(Effect.orDie),
-          config: () => config.get(),
-          env: () => env.all(),
-          get: (key: string) => env.get(key),
-        }
-
-        log.info("init")
-
-        function mergeProvider(providerID: ProviderID, provider: Partial<Info>) {
-          const existing = providers[providerID]
-          if (existing) {
-            // @ts-expect-error
-            providers[providerID] = mergeDeep(existing, provider)
-            return
-          }
-          const match = database[providerID]
-          if (!match) return
-          // @ts-expect-error
-          providers[providerID] = mergeDeep(match, provider)
-        }
-
-        // load plugins first so config() hook runs before reading cfg.provider
-        const plugins = yield* plugin.list()
-
-        // promote sleepy provider from database to providers if injected
-        if (database[ProviderID.make("sleepy")]) {
-          mergeProvider(ProviderID.make("sleepy"), {})
-        }
-
-        // now read config providers - includes any modifications from plugin config() hook
-        const configProviders = Object.entries(cfg.provider ?? {})
-        const disabled = new Set(cfg.disabled_providers ?? [])
-        const enabled = cfg.enabled_providers ? new Set(cfg.enabled_providers) : null
-
-        function isProviderAllowed(providerID: ProviderID): boolean {
-          if (enabled && !enabled.has(providerID)) return false
-          if (disabled.has(providerID)) return false
-          return true
-        }
-
-        // extend database from config
-        for (const [providerID, provider] of configProviders) {
-          const existing = database[providerID]
-          const parsed: Info = {
-            id: ProviderID.make(providerID),
-            name: provider.name ?? existing?.name ?? providerID,
-            env: provider.env ?? existing?.env ?? [],
-            options: mergeDeep(existing?.options ?? {}, provider.options ?? {}),
-            source: "config",
-            models: existing?.models ?? {},
-          }
-
-          for (const [modelID, model] of Object.entries(provider.models ?? {})) {
-            const existingModel = parsed.models[model.id ?? modelID]
-            const apiID = model.id ?? existingModel?.api.id ?? modelID
-            const apiNpm =
-              model.provider?.npm ??
-              provider.npm ??
-              existingModel?.api.npm ??
-              modelsDev[providerID]?.npm ??
-              "@ai-sdk/openai-compatible"
-            const name = iife(() => {
-              if (model.name) return model.name
-              if (model.id && model.id !== modelID) return modelID
-              return existingModel?.name ?? modelID
-            })
-            const parsedModel: Model = {
-              id: ModelID.make(modelID),
-              api: {
-                id: apiID,
-                npm: apiNpm,
-                url: model.provider?.api ?? provider?.api ?? existingModel?.api.url ?? modelsDev[providerID]?.api ?? "",
-              },
-              status: model.status ?? existingModel?.status ?? "active",
-              name,
-              providerID: ProviderID.make(providerID),
-              capabilities: {
-                temperature: model.temperature ?? existingModel?.capabilities.temperature ?? false,
-                reasoning: model.reasoning ?? existingModel?.capabilities.reasoning ?? false,
-                attachment: model.attachment ?? existingModel?.capabilities.attachment ?? false,
-                toolcall: model.tool_call ?? existingModel?.capabilities.toolcall ?? true,
-                input: {
-                  text: model.modalities?.input?.includes("text") ?? existingModel?.capabilities.input.text ?? true,
-                  audio: model.modalities?.input?.includes("audio") ?? existingModel?.capabilities.input.audio ?? false,
-                  image: model.modalities?.input?.includes("image") ?? existingModel?.capabilities.input.image ?? false,
-                  video: model.modalities?.input?.includes("video") ?? existingModel?.capabilities.input.video ?? false,
-                  pdf: model.modalities?.input?.includes("pdf") ?? existingModel?.capabilities.input.pdf ?? false,
-                },
-                output: {
-                  text: model.modalities?.output?.includes("text") ?? existingModel?.capabilities.output.text ?? true,
-                  audio:
-                    model.modalities?.output?.includes("audio") ?? existingModel?.capabilities.output.audio ?? false,
-                  image:
-                    model.modalities?.output?.includes("image") ?? existingModel?.capabilities.output.image ?? false,
-                  video:
-                    model.modalities?.output?.includes("video") ?? existingModel?.capabilities.output.video ?? false,
-                  pdf: model.modalities?.output?.includes("pdf") ?? existingModel?.capabilities.output.pdf ?? false,
-                },
-                interleaved:
-                  model.interleaved ??
-                  existingModel?.capabilities.interleaved ??
-                  (!existingModel && apiNpm === "@ai-sdk/openai-compatible" && apiID.includes("deepseek")
-                    ? { field: "reasoning_content" }
-                    : false),
-              },
-              cost: {
-                input: model?.cost?.input ?? existingModel?.cost?.input ?? 0,
-                output: model?.cost?.output ?? existingModel?.cost?.output ?? 0,
-                cache: {
-                  read: model?.cost?.cache_read ?? existingModel?.cost?.cache.read ?? 0,
-                  write: model?.cost?.cache_write ?? existingModel?.cost?.cache.write ?? 0,
-                },
-              },
-              options: mergeDeep(existingModel?.options ?? {}, model.options ?? {}),
-              limit: {
-                context: (() => {
-                  const explicit = model.limit?.context ?? existingModel?.limit?.context
-                  if (explicit !== undefined) return explicit
-                  const key = `${providerID}/${modelID}`
-                  if (!warnedContextDefaults.has(key)) {
-                    warnedContextDefaults.add(key)
-                    log.warn("limit.context not configured and not found in models.dev", {
-                      providerID,
-                      modelID,
-                      defaulting_to: DEFAULT_CONTEXT_WINDOW,
-                      fix: `Set limit.context explicitly in sleepycode.json under provider.${providerID}.models.${modelID}`,
-                    })
-                  }
-                  return DEFAULT_CONTEXT_WINDOW
-                })(),
-                input: model.limit?.input ?? existingModel?.limit?.input,
-                output: model.limit?.output ?? existingModel?.limit?.output ?? 0,
-              },
-              headers: mergeDeep(existingModel?.headers ?? {}, model.headers ?? {}),
-              family: model.family ?? existingModel?.family ?? "",
-              release_date: model.release_date ?? existingModel?.release_date ?? "",
-              cachePromptTTL: model.cachePromptTTL ?? existingModel?.cachePromptTTL,
-              variants: {},
-            }
-            const merged = mergeDeep(ProviderTransform.variants(parsedModel), model.variants ?? {})
-            parsedModel.variants = mapValues(
-              pickBy(merged, (v) => !v.disabled),
-              (v) => omit(v, ["disabled"]),
-            )
-            parsed.models[modelID] = parsedModel
-          }
-          database[providerID] = parsed
-        }
-
-        // load env (skipped in sleepy-only mode so ANTHROPIC_API_KEY etc. don't auto-light other providers)
-        if (!Flag.SLEEPYCODE_DISABLE_PROVIDER_ENV) {
-          const envs = yield* env.all()
-          for (const [id, provider] of Object.entries(database)) {
-            const providerID = ProviderID.make(id)
-            if (disabled.has(providerID)) continue
-            const apiKey = provider.env.map((item) => envs[item]).find(Boolean)
-            if (!apiKey) continue
-            mergeProvider(providerID, {
-              source: "env",
-              key: provider.env.length === 1 ? apiKey : undefined,
-            })
-          }
-        }
-
-        // load apikeys
-        const auths = yield* auth.all().pipe(Effect.orDie)
-        for (const [id, provider] of Object.entries(auths)) {
-          const providerID = ProviderID.make(id)
-          if (disabled.has(providerID)) continue
-          if (provider.type === "api") {
-            mergeProvider(providerID, {
-              source: "api",
-              key: provider.key,
-            })
-          }
-        }
-
-        // plugin auth loader - database now has entries for config providers
-        for (const plugin of plugins) {
-          if (!plugin.auth) continue
-          const providerID = ProviderID.make(plugin.auth.provider)
-          if (disabled.has(providerID)) continue
-
-          const stored = yield* auth.get(providerID).pipe(Effect.orDie)
-          if (!stored) continue
-          if (!plugin.auth.loader) continue
-
-          const options = yield* Effect.promise(() =>
-            plugin.auth!.loader!(
-              () => bridge.promise(auth.get(providerID).pipe(Effect.orDie)) as any,
-              database[plugin.auth!.provider],
-            ),
-          )
-          const opts = options ?? {}
-          const patch: Partial<Info> = providers[providerID] ? { options: opts } : { source: "custom", options: opts }
-          mergeProvider(providerID, patch)
-        }
-
-        for (const [id, fn] of Object.entries(custom(dep))) {
-          const providerID = ProviderID.make(id)
-          if (disabled.has(providerID)) continue
-          const data = database[providerID]
-          if (!data) {
-            log.error("Provider does not exist in model list " + providerID)
-            continue
-          }
-          const result = yield* fn(data)
-          if (result && (result.autoload || providers[providerID])) {
-            if (result.getModel) modelLoaders[providerID] = result.getModel
-            if (result.vars) varsLoaders[providerID] = result.vars
-            if (result.discoverModels) discoveryLoaders[providerID] = result.discoverModels
-            const opts = result.options ?? {}
-            const patch: Partial<Info> = providers[providerID] ? { options: opts } : { source: "custom", options: opts }
-            mergeProvider(providerID, patch)
-          }
-        }
-
-        // load config - re-apply with updated data
-        for (const [id, provider] of configProviders) {
-          const providerID = ProviderID.make(id)
-          const partial: Partial<Info> = { source: "config" }
-          if (provider.env) partial.env = provider.env
-          if (provider.name) partial.name = provider.name
-          if (provider.options) partial.options = provider.options
-          mergeProvider(providerID, partial)
-        }
-
-        const gitlab = ProviderID.make("gitlab")
-        if (discoveryLoaders[gitlab] && providers[gitlab] && isProviderAllowed(gitlab)) {
-          yield* Effect.promise(async () => {
-            try {
-              const discovered = await discoveryLoaders[gitlab]()
-              for (const [modelID, model] of Object.entries(discovered)) {
-                if (!providers[gitlab].models[modelID]) {
-                  providers[gitlab].models[modelID] = model
-                }
-              }
-            } catch (e) {
-              log.warn("state discovery error", { id: "gitlab", error: e })
-            }
-          })
-        }
-
-        for (const hook of plugins) {
-          const p = hook.provider
-          const models = p?.models
-          if (!p || !models) continue
-
-          const providerID = ProviderID.make(p.id)
-          if (disabled.has(providerID)) continue
-
-          const provider = providers[providerID]
-          if (!provider) continue
-          const pluginAuth = yield* auth.get(providerID).pipe(Effect.orDie)
-
-          provider.models = yield* Effect.promise(async () => {
-            const next = await models(provider, { auth: pluginAuth })
-            return Object.fromEntries(
-              Object.entries(next).map(([id, model]) => [
-                id,
-                {
-                  ...model,
-                  id: ModelID.make(id),
-                  providerID,
-                },
-              ]),
-            )
-          })
-        }
-
-        for (const [id, provider] of Object.entries(providers)) {
-          const providerID = ProviderID.make(id)
-          if (!isProviderAllowed(providerID)) {
-            delete providers[providerID]
-            continue
-          }
-
-          const configProvider = cfg.provider?.[providerID]
-
-          for (const [modelID, model] of Object.entries(provider.models)) {
-            model.api.id = model.api.id ?? model.id ?? modelID
-            if (
-              modelID === "gpt-5-chat-latest" ||
-              (providerID === ProviderID.openrouter && modelID === "openai/gpt-5-chat")
-            )
-              delete provider.models[modelID]
-            if (model.status === "alpha" && !Flag.SLEEPYCODE_ENABLE_EXPERIMENTAL_MODELS) delete provider.models[modelID]
-            if (model.status === "deprecated") delete provider.models[modelID]
-            if (
-              (configProvider?.blacklist && configProvider.blacklist.includes(modelID)) ||
-              (configProvider?.whitelist && !configProvider.whitelist.includes(modelID))
-            )
-              delete provider.models[modelID]
-
-            model.variants = mapValues(ProviderTransform.variants(model), (v) => v)
-
-            const configVariants = configProvider?.models?.[modelID]?.variants
-            if (configVariants && model.variants) {
-              const merged = mergeDeep(model.variants, configVariants)
-              model.variants = mapValues(
-                pickBy(merged, (v) => !v.disabled),
-                (v) => omit(v, ["disabled"]),
-              )
-            }
-          }
-
-          if (Object.keys(provider.models).length === 0) {
-            delete providers[providerID]
-            continue
-          }
-
-          log.info("found", { providerID })
-        }
-
-        if (sleepyCredentials) {
-          const { token, dashboardUrl } = sleepyCredentials
-          setInterval(async () => {
-            const freshModels = await fetchModels(dashboardUrl, token)
-            if (freshModels.length === 0) return
-
-            const sleepy = providers[ProviderID.make("sleepy")]
-            if (!sleepy) return
-
-            const sleepyProviderID = ProviderID.make("sleepy")
-            const baseUrl = `${dashboardUrl}/api/v1`
-            const authHeader = `Bearer ${token}`
-
-            for (const m of freshModels) {
-              const existing = sleepy.models[m.modelId]
-              if (existing) {
-                existing.name = m.name
-                existing.api.id = m.omniRouteModelId
-                existing.cost = {
-                  input: m.inputPrice ?? 0.0015,
-                  output: m.outputPrice ?? 0.005,
-                  cache: {
-                    read: m.cacheReadPrice ?? 0.00075,
-                    write: m.cacheWritePrice ?? 0.0015,
-                  },
-                }
-                existing.limit = {
-                  context: m.contextWindow ?? 128000,
-                  output: m.maxOutputLimit ?? 4096,
-                }
-              } else {
-                sleepy.models[m.modelId] = {
-                  id: ModelID.make(m.modelId),
-                  providerID: sleepyProviderID,
-                  name: m.name,
-                  family: "sleepy",
-                  api: {
-                    id: m.omniRouteModelId,
-                    url: baseUrl,
-                    npm: "@ai-sdk/openai-compatible",
-                  },
-                  status: "active",
-                  headers: { Authorization: authHeader },
-                  options: {},
-                  cost: {
-                    input: m.inputPrice ?? 0.0015,
-                    output: m.outputPrice ?? 0.005,
-                    cache: {
-                      read: m.cacheReadPrice ?? 0.00075,
-                      write: m.cacheWritePrice ?? 0.0015,
-                    },
-                  },
-                  limit: {
-                    context: m.contextWindow ?? 128000,
-                    output: m.maxOutputLimit ?? 4096,
-                  },
-                  capabilities: {
-                    temperature: true,
-                    reasoning: true,
-                    attachment: true,
-                    toolcall: true,
-                    input: { text: true, audio: false, image: true, video: false, pdf: true },
-                    output: { text: true, audio: false, image: false, video: false, pdf: false },
-                    interleaved: false,
-                  },
-                  release_date: "",
-                  variants: {},
-                }
-              }
-            }
-          }, 30 * 1000).unref()
-        }
-
-        return {
-          models: languages,
-          providers,
-          sdk,
-          modelLoaders,
-          varsLoaders,
-        }
-      }),
-    )
-
-    const list = Effect.fn("Provider.list")(() => InstanceState.use(state, (s) => s.providers))
-
-    async function resolveSDK(model: Model, s: State, envs: Record<string, string | undefined>) {
-      try {
-        using _ = log.time("getSDK", {
-          providerID: model.providerID,
-        })
-        const provider = s.providers[model.providerID]
-        const options = { ...provider.options }
-
-        if (model.providerID === "google-vertex" && !model.api.npm.includes("@ai-sdk/openai-compatible")) {
-          delete options.fetch
-        }
-
-        if (model.api.npm.includes("@ai-sdk/openai-compatible") && options["includeUsage"] !== false) {
-          options["includeUsage"] = true
-        }
-
-        const baseURL = iife(() => {
-          let url =
-            typeof options["baseURL"] === "string" && options["baseURL"] !== "" ? options["baseURL"] : model.api.url
-          if (!url) return
-
-          const loader = s.varsLoaders[model.providerID]
-          if (loader) {
-            const vars = loader(options)
-            for (const [key, value] of Object.entries(vars)) {
-              const field = "${" + key + "}"
-              url = url.replaceAll(field, value)
-            }
-          }
-
-          url = url.replace(/\$\{([^}]+)\}/g, (item, key) => {
-            const val = envs[String(key)]
-            return val ?? item
-          })
-          return url
-        })
-
-        if (baseURL !== undefined) options["baseURL"] = baseURL
-        if (options["apiKey"] === undefined && provider.key) options["apiKey"] = provider.key
-        if (model.headers)
-          options["headers"] = {
-            ...options["headers"],
-            ...model.headers,
-          }
-
-        const key = Hash.fast(
-          JSON.stringify({
-            providerID: model.providerID,
-            npm: model.api.npm,
-            options,
-          }),
-        )
-        const existing = s.sdk.get(key)
-        if (existing) return existing
-
-        const customFetch = options["fetch"]
-        const userChunkTimeout = options["chunkTimeout"]
-        const chunkTimeout =
-          typeof userChunkTimeout === "number"
-            ? userChunkTimeout  // user-set value (incl. 0 / negative to disable)
-            : DEFAULT_CHUNK_TIMEOUT
-        delete options["chunkTimeout"]
-
-        options["fetch"] = async (input: any, init?: BunFetchRequestInit) => {
-          const fetchFn = customFetch ?? fetch
-          const opts = init ?? {}
-          const chunkAbortCtl = typeof chunkTimeout === "number" && chunkTimeout > 0 ? new AbortController() : undefined
-          const signals: AbortSignal[] = []
-
-          if (opts.signal) signals.push(opts.signal)
-          if (chunkAbortCtl) signals.push(chunkAbortCtl.signal)
-          if (options["timeout"] !== undefined && options["timeout"] !== null && options["timeout"] !== false)
-            signals.push(AbortSignal.timeout(options["timeout"]))
-
-          const combined = signals.length === 0 ? null : signals.length === 1 ? signals[0] : AbortSignal.any(signals)
-          if (combined) opts.signal = combined
-
-          // Strip openai itemId metadata following what codex does
-          if (model.api.npm === "@ai-sdk/openai" && opts.body && opts.method === "POST") {
-            const body = JSON.parse(opts.body as string)
-            const isAzure = model.providerID.includes("azure")
-            const keepIds = isAzure && body.store === true
-            if (!keepIds && Array.isArray(body.input)) {
-              for (const item of body.input) {
-                if ("id" in item) {
-                  delete item.id
-                }
-              }
-              opts.body = JSON.stringify(body)
-            }
-          }
-
-          const res = await fetchFn(input, {
-            ...opts,
-            // @ts-ignore see here: https://github.com/oven-sh/bun/issues/16682
-            timeout: false,
-          })
-
-          if (!chunkAbortCtl) return res
-          return wrapSSE(res, chunkTimeout, chunkAbortCtl)
-        }
-
-        const bundledLoader = BUNDLED_PROVIDERS[model.api.npm]
-        if (bundledLoader) {
-          log.info("using bundled provider", {
-            providerID: model.providerID,
-            pkg: model.api.npm,
-          })
-          const factory = await bundledLoader()
-          const loaded = factory({
-            name: model.providerID,
-            ...options,
-          })
-          s.sdk.set(key, loaded)
-          return loaded as SDK
-        }
-
-        let installedPath: string
-        if (!model.api.npm.startsWith("file://")) {
-          const item = await Npm.add(model.api.npm)
-          if (!item.entrypoint) throw new Error(`Package ${model.api.npm} has no import entrypoint`)
-          installedPath = item.entrypoint
-        } else {
-          log.info("loading local provider", { pkg: model.api.npm })
-          installedPath = model.api.npm
-        }
-
-        // `installedPath` is a local entry path or an existing `file://` URL. Normalize
-        // only path inputs so Node on Windows accepts the dynamic import.
-        const importSpec = installedPath.startsWith("file://") ? installedPath : pathToFileURL(installedPath).href
-        const mod = await import(importSpec)
-
-        const fn = mod[Object.keys(mod).find((key) => key.startsWith("create"))!]
-        const loaded = fn({
-          name: model.providerID,
-          ...options,
-        })
-        s.sdk.set(key, loaded)
-        return loaded as SDK
-      } catch (e) {
-        throw new InitError({ providerID: model.providerID }, { cause: e })
-      }
-    }
-
-    const getProvider = Effect.fn("Provider.getProvider")((providerID: ProviderID) =>
-      InstanceState.use(state, (s) => s.providers[providerID]),
-    )
-
-    const getModel = Effect.fn("Provider.getModel")(function* (providerID: ProviderID, modelID: ModelID) {
-      const s = yield* InstanceState.get(state)
-      const provider = s.providers[providerID]
-      if (!provider) {
-        const available = Object.keys(s.providers)
-        const matches = fuzzysort.go(providerID, available, { limit: 3, threshold: -10000 })
-        throw new ModelNotFoundError({ providerID, modelID, suggestions: matches.map((m) => m.target) })
-      }
-
-      const info = provider.models[modelID]
-      if (!info) {
-        const available = Object.keys(provider.models)
-        const matches = fuzzysort.go(modelID, available, { limit: 3, threshold: -10000 })
-        throw new ModelNotFoundError({ providerID, modelID, suggestions: matches.map((m) => m.target) })
-      }
-      return info
-    })
-
-    const getLanguage = Effect.fn("Provider.getLanguage")(function* (model: Model) {
-      const s = yield* InstanceState.get(state)
-      const envs = yield* env.all()
-      const key = `${model.providerID}/${model.id}`
-      if (s.models.has(key)) return s.models.get(key)!
-
-      return yield* Effect.promise(async () => {
-        const provider = s.providers[model.providerID]
-        const sdk = await resolveSDK(model, s, envs)
-
-        try {
-          const language = s.modelLoaders[model.providerID]
-            ? await s.modelLoaders[model.providerID](sdk, model.api.id, {
-                ...provider.options,
-                ...model.options,
-              })
-            : sdk.languageModel(model.api.id)
-          s.models.set(key, language)
-          return language
-        } catch (e) {
-          if (e instanceof NoSuchModelError)
-            throw new ModelNotFoundError(
-              {
-                modelID: model.id,
-                providerID: model.providerID,
-              },
-              { cause: e },
-            )
-          throw e
-        }
-      })
-    })
-
-    const closest = Effect.fn("Provider.closest")(function* (providerID: ProviderID, query: string[]) {
-      const s = yield* InstanceState.get(state)
-      const provider = s.providers[providerID]
-      if (!provider) return undefined
-      for (const item of query) {
-        for (const modelID of Object.keys(provider.models)) {
-          if (modelID.includes(item)) return { providerID, modelID }
-        }
-      }
-      return undefined
-    })
-
-    const resolveModelRef = Effect.fn("Provider.resolveModelRef")(function* (
-      ref: string,
-      contextProviderID?: ProviderID,
-    ) {
-      // Literal "provider/model" — unchanged path.
-      if (ref.includes("/")) {
-        const parsed = parseModel(ref)
-        return yield* getModel(parsed.providerID, parsed.modelID)
-      }
-
-      // Otherwise it's a group name.
-      const cfg = yield* config.get()
-      const group = cfg.model_groups?.[ref]
-      if (!group) {
-        if (BUILTIN_TIERS.has(ref)) {
-          const fallback = yield* defaultModel()
-          return yield* getModel(fallback.providerID, fallback.modelID)
-        }
-        const names = Object.keys(cfg.model_groups ?? {})
-        const matches = fuzzysort.go(ref, names, { limit: 3, threshold: -10000 })
-        throw new ModelGroupNotFoundError({ group: ref, suggestions: matches.map((m) => m.target) })
-      }
-
-      const def = typeof group === "string" ? group : group.default
-      const members = typeof group === "string" ? [group] : (group.models ?? [group.default])
-
-      // Provider-aware: prefer a member on the caller's provider.
-      if (contextProviderID) {
-        for (const member of members) {
-          const parsed = parseModel(member)
-          if (parsed.providerID !== contextProviderID) continue
-          const resolved = yield* getModel(parsed.providerID, parsed.modelID).pipe(Effect.exit)
-          if (resolved._tag === "Success") return resolved.value
-        }
-      }
-
-      // Fall back to the group default.
-      const parsedDefault = parseModel(def)
-      return yield* getModel(parsedDefault.providerID, parsedDefault.modelID)
-    })
-
-    const getSmallModel = Effect.fn("Provider.getSmallModel")(function* (providerID: ProviderID) {
-      const cfg = yield* config.get()
-
-      // Explicit small_model literal wins (back-compat).
-      if (cfg.small_model) {
-        const parsed = parseModel(cfg.small_model)
-        return yield* getModel(parsed.providerID, parsed.modelID)
-      }
-
-      // Otherwise route through the `lite` tier: configured lite group
-      // (provider-aware) → else the built-in tier fallback to defaultModel().
-      return yield* resolveModelRef("lite", providerID)
-    })
-
-    const defaultModel = Effect.fn("Provider.defaultModel")(function* () {
-      const cfg = yield* config.get()
-      if (cfg.model) return parseModel(cfg.model)
-
-      const s = yield* InstanceState.get(state)
-      const recent = yield* fs.readJson(path.join(Global.Path.state, "model.json")).pipe(
-        Effect.map((x): { providerID: ProviderID; modelID: ModelID }[] => {
-          if (!isRecord(x) || !Array.isArray(x.recent)) return []
-          return x.recent.flatMap((item) => {
-            if (!isRecord(item)) return []
-            if (typeof item.providerID !== "string") return []
-            if (typeof item.modelID !== "string") return []
-            return [{ providerID: ProviderID.make(item.providerID), modelID: ModelID.make(item.modelID) }]
-          })
-        }),
-        Effect.catch(() => Effect.succeed([] as { providerID: ProviderID; modelID: ModelID }[])),
-      )
-      for (const entry of recent) {
-        const provider = s.providers[entry.providerID]
-        if (!provider) continue
-        if (!provider.models[entry.modelID]) continue
-        return { providerID: entry.providerID, modelID: entry.modelID }
-      }
-
-      const sleepy = s.providers[ProviderID.make("sleepy")]
-      if (sleepy?.models[ModelID.make("smart")]) {
-        return { providerID: sleepy.id, modelID: ModelID.make("smart") }
-      }
-
-      const provider = Object.values(s.providers).find((p) => !cfg.provider || Object.keys(cfg.provider).includes(p.id))
-      if (!provider) throw new Error("no providers found")
-      const [model] = sort(Object.values(provider.models))
-      if (!model) throw new Error("no models found")
-      return {
-        providerID: provider.id,
-        modelID: model.id,
-      }
-    })
-
-    return Service.of({ list, getProvider, getModel, getLanguage, closest, getSmallModel, defaultModel, resolveModelRef })
-  }),
-)
-
-export const defaultLayer = Layer.suspend(() =>
-  layer.pipe(
-    Layer.provide(AppFileSystem.defaultLayer),
-    Layer.provide(Env.defaultLayer),
-    Layer.provide(Config.defaultLayer),
-    Layer.provide(Auth.defaultLayer),
-    Layer.provide(Plugin.defaultLayer),
-  ),
-)
-
-const priority = ["gpt-5", "claude-sonnet-4", "big-pickle", "gemini-3-pro"]
-export function sort<T extends { id: string }>(models: T[]) {
-  return sortBy(
-    models,
-    [(model) => priority.findIndex((filter) => model.id.includes(filter)), "desc"],
-    [(model) => (model.id.includes("latest") ? 0 : 1), "asc"],
-    [(model) => model.id, "desc"],
-  )
-}
-
-export function parseModel(model: string) {
-  const [providerID, ...rest] = model.split("/")
-  return {
-    providerID: ProviderID.make(providerID),
-    modelID: ModelID.make(rest.join("/")),
-  }
-}
-
-export const ModelNotFoundError = NamedError.create(
-  "ProviderModelNotFoundError",
-  z.object({
-    providerID: ProviderID.zod,
-    modelID: ModelID.zod,
-    suggestions: z.array(z.string()).optional(),
-  }),
-)
-
-export const ModelGroupNotFoundError = NamedError.create(
-  "ProviderModelGroupNotFoundError",
-  z.object({
-    group: z.string(),
-    suggestions: z.array(z.string()).optional(),
-  }),
-)
-
-export const InitError = NamedError.create(
-  "ProviderInitError",
-  z.object({
-    providerID: ProviderID.zod,
-  }),
-)