cerebras-cli 1.0.5 → 1.0.138

package/src/provider/provider.ts

@@ -0,0 +1,1005 @@
+import z from "zod"
+import fuzzysort from "fuzzysort"
+import { Config } from "../config/config"
+import { mapValues, mergeDeep, sortBy } from "remeda"
+import { NoSuchModelError, type Provider as SDK } from "ai"
+import { Log } from "../util/log"
+import { BunProc } from "../bun"
+import { Plugin } from "../plugin"
+import { ModelsDev } from "./models"
+import { NamedError } from "@opencode-ai/util/error"
+import { Auth } from "../auth"
+import { Env } from "../env"
+import { Instance } from "../project/instance"
+import { Flag } from "../flag/flag"
+import { iife } from "@/util/iife"
+import { RateLimit } from "@/ratelimit"
+
+// Direct imports for bundled providers
+import { createAmazonBedrock } from "@ai-sdk/amazon-bedrock"
+import { createAnthropic } from "@ai-sdk/anthropic"
+import { createAzure } from "@ai-sdk/azure"
+import { createGoogleGenerativeAI } from "@ai-sdk/google"
+import { createVertex } from "@ai-sdk/google-vertex"
+import { createVertexAnthropic } from "@ai-sdk/google-vertex/anthropic"
+import { createOpenAI } from "@ai-sdk/openai"
+import { createOpenAICompatible } from "@ai-sdk/openai-compatible"
+import { createOpenRouter, type LanguageModelV2 } from "@openrouter/ai-sdk-provider"
+import { createOpenaiCompatible as createGitHubCopilotOpenAICompatible } from "./sdk/openai-compatible/src"
+
+export namespace Provider {
+  const log = Log.create({ service: "provider" })
+
+  const BUNDLED_PROVIDERS: Record<string, (options: any) => SDK> = {
+    "@ai-sdk/amazon-bedrock": createAmazonBedrock,
+    "@ai-sdk/anthropic": createAnthropic,
+    "@ai-sdk/azure": createAzure,
+    "@ai-sdk/google": createGoogleGenerativeAI,
+    "@ai-sdk/google-vertex": createVertex,
+    "@ai-sdk/google-vertex/anthropic": createVertexAnthropic,
+    "@ai-sdk/openai": createOpenAI,
+    "@ai-sdk/openai-compatible": createOpenAICompatible,
+    "@openrouter/ai-sdk-provider": createOpenRouter,
+    // @ts-ignore (TODO: kill this code so we dont have to maintain it)
+    "@ai-sdk/github-copilot": createGitHubCopilotOpenAICompatible,
+    // Use OpenAI-compatible SDK for Cerebras to handle response format differences
+    "@ai-sdk/cerebras": createOpenAICompatible,
+  }
+
+  type CustomModelLoader = (sdk: any, modelID: string, options?: Record<string, any>) => Promise<any>
+  type CustomLoader = (provider: Info) => Promise<{
+    autoload: boolean
+    getModel?: CustomModelLoader
+    options?: Record<string, any>
+  }>
+
+  const CUSTOM_LOADERS: Record<string, CustomLoader> = {
+    async anthropic() {
+      return {
+        autoload: false,
+        options: {
+          headers: {
+            "anthropic-beta":
+              "claude-code-20250219,interleaved-thinking-2025-05-14,fine-grained-tool-streaming-2025-05-14",
+          },
+        },
+      }
+    },
+    async opencode(input) {
+      const hasKey = await (async () => {
+        const env = Env.all()
+        if (input.env.some((item) => env[item])) return true
+        if (await Auth.get(input.id)) return true
+        return false
+      })()
+
+      if (!hasKey) {
+        for (const [key, value] of Object.entries(input.models)) {
+          if (value.cost.input === 0) continue
+          delete input.models[key]
+        }
+      }
+
+      return {
+        autoload: Object.keys(input.models).length > 0,
+        options: hasKey ? {} : { apiKey: "public" },
+      }
+    },
+    openai: async () => {
+      return {
+        autoload: false,
+        async getModel(sdk: any, modelID: string, _options?: Record<string, any>) {
+          return sdk.responses(modelID)
+        },
+        options: {},
+      }
+    },
+    cerebras: async () => {
+      return {
+        autoload: false,
+        options: {
+          baseURL: "https://api.cerebras.ai/v1",
+          headers: {
+            "X-Cerebras-3rd-Party-Integration": "cli",
+          },
+        },
+      }
+    },
+    "github-copilot": async () => {
+      return {
+        autoload: false,
+        async getModel(sdk: any, modelID: string, _options?: Record<string, any>) {
+          if (modelID.includes("codex")) {
+            return sdk.responses(modelID)
+          }
+          return sdk.chat(modelID)
+        },
+        options: {},
+      }
+    },
+    "github-copilot-enterprise": async () => {
+      return {
+        autoload: false,
+        async getModel(sdk: any, modelID: string, _options?: Record<string, any>) {
+          if (modelID.includes("codex")) {
+            return sdk.responses(modelID)
+          }
+          return sdk.chat(modelID)
+        },
+        options: {},
+      }
+    },
+    azure: async () => {
+      return {
+        autoload: false,
+        async getModel(sdk: any, modelID: string, options?: Record<string, any>) {
+          if (options?.["useCompletionUrls"]) {
+            return sdk.chat(modelID)
+          } else {
+            return sdk.responses(modelID)
+          }
+        },
+        options: {},
+      }
+    },
+    "azure-cognitive-services": async () => {
+      const resourceName = Env.get("AZURE_COGNITIVE_SERVICES_RESOURCE_NAME")
+      return {
+        autoload: false,
+        async getModel(sdk: any, modelID: string, options?: Record<string, any>) {
+          if (options?.["useCompletionUrls"]) {
+            return sdk.chat(modelID)
+          } else {
+            return sdk.responses(modelID)
+          }
+        },
+        options: {
+          baseURL: resourceName ? `https://${resourceName}.cognitiveservices.azure.com/openai` : undefined,
+        },
+      }
+    },
+    "amazon-bedrock": async () => {
+      const [awsProfile, awsAccessKeyId, awsBearerToken, awsRegion] = await Promise.all([
+        Env.get("AWS_PROFILE"),
+        Env.get("AWS_ACCESS_KEY_ID"),
+        Env.get("AWS_BEARER_TOKEN_BEDROCK"),
+        Env.get("AWS_REGION"),
+      ])
+      if (!awsProfile && !awsAccessKeyId && !awsBearerToken) return { autoload: false }
+
+      const region = awsRegion ?? "us-east-1"
+
+      const { fromNodeProviderChain } = await import(await BunProc.install("@aws-sdk/credential-providers"))
+      return {
+        autoload: true,
+        options: {
+          region,
+          credentialProvider: fromNodeProviderChain(),
+        },
+        async getModel(sdk: any, modelID: string, _options?: Record<string, any>) {
+          // Skip region prefixing if model already has global prefix
+          if (modelID.startsWith("global.")) {
+            return sdk.languageModel(modelID)
+          }
+
+          let regionPrefix = region.split("-")[0]
+
+          switch (regionPrefix) {
+            case "us": {
+              const modelRequiresPrefix = [
+                "nova-micro",
+                "nova-lite",
+                "nova-pro",
+                "nova-premier",
+                "claude",
+                "deepseek",
+              ].some((m) => modelID.includes(m))
+              const isGovCloud = region.startsWith("us-gov")
+              if (modelRequiresPrefix && !isGovCloud) {
+                modelID = `${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}.${modelID}`
+              }
+              break
+            }
+            case "ap": {
+              const isAustraliaRegion = ["ap-southeast-2", "ap-southeast-4"].includes(region)
+              if (
+                isAustraliaRegion &&
+                ["anthropic.claude-sonnet-4-5", "anthropic.claude-haiku"].some((m) => modelID.includes(m))
+              ) {
+                regionPrefix = "au"
+                modelID = `${regionPrefix}.${modelID}`
+              } else {
+                const modelRequiresPrefix = ["claude", "nova-lite", "nova-micro", "nova-pro"].some((m) =>
+                  modelID.includes(m),
+                )
+                if (modelRequiresPrefix) {
+                  regionPrefix = "apac"
+                  modelID = `${regionPrefix}.${modelID}`
+                }
+              }
+              break
+            }
+          }
+
+          return sdk.languageModel(modelID)
+        },
+      }
+    },
+    openrouter: async () => {
+      return {
+        autoload: false,
+        options: {
+          headers: {
+            "HTTP-Referer": "https://opencode.ai/",
+            "X-Title": "opencode",
+          },
+        },
+      }
+    },
+    vercel: async () => {
+      return {
+        autoload: false,
+        options: {
+          headers: {
+            "http-referer": "https://opencode.ai/",
+            "x-title": "opencode",
+          },
+        },
+      }
+    },
+    "google-vertex": async () => {
+      const project = Env.get("GOOGLE_CLOUD_PROJECT") ?? Env.get("GCP_PROJECT") ?? Env.get("GCLOUD_PROJECT")
+      const location = Env.get("GOOGLE_CLOUD_LOCATION") ?? Env.get("VERTEX_LOCATION") ?? "us-east5"
+      const autoload = Boolean(project)
+      if (!autoload) return { autoload: false }
+      return {
+        autoload: true,
+        options: {
+          project,
+          location,
+        },
+        async getModel(sdk: any, modelID: string) {
+          const id = String(modelID).trim()
+          return sdk.languageModel(id)
+        },
+      }
+    },
+    "google-vertex-anthropic": async () => {
+      const project = Env.get("GOOGLE_CLOUD_PROJECT") ?? Env.get("GCP_PROJECT") ?? Env.get("GCLOUD_PROJECT")
+      const location = Env.get("GOOGLE_CLOUD_LOCATION") ?? Env.get("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": async () => {
+      const auth = await Auth.get("sap-ai-core")
+      const envServiceKey = iife(() => {
+        const envAICoreServiceKey = Env.get("AICORE_SERVICE_KEY")
+        if (envAICoreServiceKey) return envAICoreServiceKey
+        if (auth?.type === "api") {
+          Env.set("AICORE_SERVICE_KEY", auth.key)
+          return auth.key
+        }
+        return undefined
+      })
+      const deploymentId = Env.get("AICORE_DEPLOYMENT_ID")
+      const resourceGroup = Env.get("AICORE_RESOURCE_GROUP")
+
+      return {
+        autoload: !!envServiceKey,
+        options: envServiceKey ? { deploymentId, resourceGroup } : {},
+        async getModel(sdk: any, modelID: string) {
+          return sdk(modelID)
+        },
+      }
+    },
+    zenmux: async () => {
+      return {
+        autoload: false,
+        options: {
+          headers: {
+            "HTTP-Referer": "https://opencode.ai/",
+            "X-Title": "opencode",
+          },
+        },
+      }
+    },
+  }
+
+  export const Model = z
+    .object({
+      id: z.string(),
+      providerID: z.string(),
+      api: z.object({
+        id: z.string(),
+        url: z.string(),
+        npm: z.string(),
+      }),
+      name: z.string(),
+      capabilities: z.object({
+        temperature: z.boolean(),
+        reasoning: z.boolean(),
+        attachment: z.boolean(),
+        toolcall: z.boolean(),
+        input: z.object({
+          text: z.boolean(),
+          audio: z.boolean(),
+          image: z.boolean(),
+          video: z.boolean(),
+          pdf: z.boolean(),
+        }),
+        output: z.object({
+          text: z.boolean(),
+          audio: z.boolean(),
+          image: z.boolean(),
+          video: z.boolean(),
+          pdf: z.boolean(),
+        }),
+      }),
+      cost: z.object({
+        input: z.number(),
+        output: z.number(),
+        cache: z.object({
+          read: z.number(),
+          write: z.number(),
+        }),
+        experimentalOver200K: z
+          .object({
+            input: z.number(),
+            output: z.number(),
+            cache: z.object({
+              read: z.number(),
+              write: z.number(),
+            }),
+          })
+          .optional(),
+      }),
+      limit: z.object({
+        context: z.number(),
+        output: z.number(),
+      }),
+      status: z.enum(["alpha", "beta", "deprecated", "active"]),
+      options: z.record(z.string(), z.any()),
+      headers: z.record(z.string(), z.string()),
+    })
+    .meta({
+      ref: "Model",
+    })
+  export type Model = z.infer<typeof Model>
+
+  export const Info = z
+    .object({
+      id: z.string(),
+      name: z.string(),
+      source: z.enum(["env", "config", "custom", "api"]),
+      env: z.string().array(),
+      key: z.string().optional(),
+      options: z.record(z.string(), z.any()),
+      models: z.record(z.string(), Model),
+    })
+    .meta({
+      ref: "Provider",
+    })
+  export type Info = z.infer<typeof Info>
+
+  function fromModelsDevModel(provider: ModelsDev.Provider, model: ModelsDev.Model): Model {
+    return {
+      id: model.id,
+      providerID: provider.id,
+      name: model.name,
+      api: {
+        id: model.id,
+        url: provider.api!,
+        npm: model.provider?.npm ?? provider.npm ?? provider.id,
+      },
+      status: model.status ?? "active",
+      headers: model.headers ?? {},
+      options: model.options ?? {},
+      cost: {
+        input: model.cost?.input ?? 0,
+        output: model.cost?.output ?? 0,
+        cache: {
+          read: model.cost?.cache_read ?? 0,
+          write: model.cost?.cache_write ?? 0,
+        },
+        experimentalOver200K: model.cost?.context_over_200k
+          ? {
+              cache: {
+                read: model.cost.context_over_200k.cache_read ?? 0,
+                write: model.cost.context_over_200k.cache_write ?? 0,
+              },
+              input: model.cost.context_over_200k.input,
+              output: model.cost.context_over_200k.output,
+            }
+          : undefined,
+      },
+      limit: {
+        context: model.limit.context,
+        output: model.limit.output,
+      },
+      capabilities: {
+        temperature: model.temperature,
+        reasoning: model.reasoning,
+        attachment: model.attachment,
+        toolcall: model.tool_call,
+        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,
+        },
+      },
+    }
+  }
+
+  export function fromModelsDevProvider(provider: ModelsDev.Provider): Info {
+    return {
+      id: provider.id,
+      source: "custom",
+      name: provider.name,
+      env: provider.env ?? [],
+      options: {},
+      models: mapValues(provider.models, (model) => fromModelsDevModel(provider, model)),
+    }
+  }
+
+  const state = Instance.state(async () => {
+    using _ = log.time("state")
+    const config = await Config.get()
+    const modelsDev = await ModelsDev.get()
+    const database = mapValues(modelsDev, fromModelsDevProvider)
+
+    const disabled = new Set(config.disabled_providers ?? [])
+    const enabled = config.enabled_providers ? new Set(config.enabled_providers) : null
+
+    function isProviderAllowed(providerID: string): boolean {
+      if (enabled && !enabled.has(providerID)) return false
+      if (disabled.has(providerID)) return false
+      return true
+    }
+
+    const providers: { [providerID: string]: Info } = {}
+    const languages = new Map<string, LanguageModelV2>()
+    const modelLoaders: {
+      [providerID: string]: CustomModelLoader
+    } = {}
+    const sdk = new Map<number, SDK>()
+
+    log.info("init")
+
+    const configProviders = Object.entries(config.provider ?? {})
+
+    // Add GitHub Copilot Enterprise provider that inherits from GitHub Copilot
+    if (database["github-copilot"]) {
+      const githubCopilot = database["github-copilot"]
+      database["github-copilot-enterprise"] = {
+        ...githubCopilot,
+        id: "github-copilot-enterprise",
+        name: "GitHub Copilot Enterprise",
+        models: mapValues(githubCopilot.models, (model) => ({
+          ...model,
+          providerID: "github-copilot-enterprise",
+        })),
+      }
+    }
+
+    function mergeProvider(providerID: string, 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)
+    }
+
+    // extend database from config
+    for (const [providerID, provider] of configProviders) {
+      const existing = database[providerID]
+      const parsed: Info = {
+        id: 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 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,
+          api: {
+            id: model.id ?? existingModel?.api.id ?? modelID,
+            npm:
+              model.provider?.npm ?? provider.npm ?? existingModel?.api.npm ?? modelsDev[providerID]?.npm ?? providerID,
+            url: provider?.api ?? existingModel?.api.url ?? modelsDev[providerID]?.api,
+          },
+          status: model.status ?? existingModel?.status ?? "active",
+          name,
+          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,
+            },
+          },
+          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: model.limit?.context ?? existingModel?.limit?.context ?? 0,
+            output: model.limit?.output ?? existingModel?.limit?.output ?? 0,
+          },
+          headers: mergeDeep(existingModel?.headers ?? {}, model.headers ?? {}),
+        }
+        parsed.models[modelID] = parsedModel
+      }
+      database[providerID] = parsed
+    }
+
+    // load env
+    const env = Env.all()
+    for (const [providerID, provider] of Object.entries(database)) {
+      if (disabled.has(providerID)) continue
+      const apiKey = provider.env.map((item) => env[item]).find(Boolean)
+      if (!apiKey) continue
+      mergeProvider(providerID, {
+        source: "env",
+        key: provider.env.length === 1 ? apiKey : undefined,
+      })
+    }
+
+    // load apikeys
+    for (const [providerID, provider] of Object.entries(await Auth.all())) {
+      if (disabled.has(providerID)) continue
+      if (provider.type === "api") {
+        mergeProvider(providerID, {
+          source: "api",
+          key: provider.key,
+        })
+      }
+    }
+
+    for (const plugin of await Plugin.list()) {
+      if (!plugin.auth) continue
+      const providerID = plugin.auth.provider
+      if (disabled.has(providerID)) continue
+
+      // For github-copilot plugin, check if auth exists for either github-copilot or github-copilot-enterprise
+      let hasAuth = false
+      const auth = await Auth.get(providerID)
+      if (auth) hasAuth = true
+
+      // Special handling for github-copilot: also check for enterprise auth
+      if (providerID === "github-copilot" && !hasAuth) {
+        const enterpriseAuth = await Auth.get("github-copilot-enterprise")
+        if (enterpriseAuth) hasAuth = true
+      }
+
+      if (!hasAuth) continue
+      if (!plugin.auth.loader) continue
+
+      // Load for the main provider if auth exists
+      if (auth) {
+        const options = await plugin.auth.loader(() => Auth.get(providerID) as any, database[plugin.auth.provider])
+        mergeProvider(plugin.auth.provider, {
+          source: "custom",
+          options: options,
+        })
+      }
+
+      // If this is github-copilot plugin, also register for github-copilot-enterprise if auth exists
+      if (providerID === "github-copilot") {
+        const enterpriseProviderID = "github-copilot-enterprise"
+        if (!disabled.has(enterpriseProviderID)) {
+          const enterpriseAuth = await Auth.get(enterpriseProviderID)
+          if (enterpriseAuth) {
+            const enterpriseOptions = await plugin.auth.loader(
+              () => Auth.get(enterpriseProviderID) as any,
+              database[enterpriseProviderID],
+            )
+            mergeProvider(enterpriseProviderID, {
+              source: "custom",
+              options: enterpriseOptions,
+            })
+          }
+        }
+      }
+    }
+
+    for (const [providerID, fn] of Object.entries(CUSTOM_LOADERS)) {
+      if (disabled.has(providerID)) continue
+      const result = await fn(database[providerID])
+      if (result && (result.autoload || providers[providerID])) {
+        if (result.getModel) modelLoaders[providerID] = result.getModel
+        mergeProvider(providerID, {
+          source: "custom",
+          options: result.options,
+        })
+      }
+    }
+
+    // load config
+    for (const [providerID, provider] of configProviders) {
+      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)
+    }
+
+    for (const [providerID, provider] of Object.entries(providers)) {
+      if (!isProviderAllowed(providerID)) {
+        delete providers[providerID]
+        continue
+      }
+
+      if (providerID === "github-copilot" || providerID === "github-copilot-enterprise") {
+        provider.models = mapValues(provider.models, (model) => ({
+          ...model,
+          api: {
+            ...model.api,
+            npm: "@ai-sdk/github-copilot",
+          },
+        }))
+      }
+
+      const configProvider = config.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 === "openrouter" && modelID === "openai/gpt-5-chat"))
+          delete provider.models[modelID]
+        if (model.status === "alpha" && !Flag.OPENCODE_ENABLE_EXPERIMENTAL_MODELS) delete provider.models[modelID]
+        if (
+          (configProvider?.blacklist && configProvider.blacklist.includes(modelID)) ||
+          (configProvider?.whitelist && !configProvider.whitelist.includes(modelID))
+        )
+          delete provider.models[modelID]
+      }
+
+      if (Object.keys(provider.models).length === 0) {
+        delete providers[providerID]
+        continue
+      }
+
+      log.info("found", { providerID })
+    }
+
+    return {
+      models: languages,
+      providers,
+      sdk,
+      modelLoaders,
+    }
+  })
+
+  export async function list() {
+    return state().then((state) => state.providers)
+  }
+
+  async function getSDK(model: Model) {
+    try {
+      using _ = log.time("getSDK", {
+        providerID: model.providerID,
+      })
+      const s = await state()
+      const provider = s.providers[model.providerID]
+      const options = { ...provider.options }
+
+      if (model.api.npm.includes("@ai-sdk/openai-compatible") && options["includeUsage"] !== false) {
+        options["includeUsage"] = true
+      }
+
+      if (!options["baseURL"]) options["baseURL"] = model.api.url
+      if (options["apiKey"] === undefined && provider.key) options["apiKey"] = provider.key
+      if (model.headers)
+        options["headers"] = {
+          ...options["headers"],
+          ...model.headers,
+        }
+
+      const key = Bun.hash.xxHash32(JSON.stringify({ npm: model.api.npm, options }))
+      const existing = s.sdk.get(key)
+      if (existing) return existing
+
+      const customFetch = options["fetch"]
+
+      options["fetch"] = async (input: any, init?: BunFetchRequestInit) => {
+        // Preserve custom fetch if it exists, wrap it with timeout logic
+        const fetchFn = customFetch ?? fetch
+        const opts = init ?? {}
+
+        if (options["timeout"] !== undefined && options["timeout"] !== null) {
+          const signals: AbortSignal[] = []
+          if (opts.signal) signals.push(opts.signal)
+          if (options["timeout"] !== false) signals.push(AbortSignal.timeout(options["timeout"]))
+
+          const combined = signals.length > 1 ? AbortSignal.any(signals) : signals[0]
+
+          opts.signal = combined
+        }
+
+        const response = await fetchFn(input, {
+          ...opts,
+          // @ts-ignore see here: https://github.com/oven-sh/bun/issues/16682
+          timeout: false,
+        })
+
+        // Capture rate limit headers from response
+        try {
+          const rateLimitInfo = RateLimit.parseHeaders(response.headers, model.providerID)
+          if (rateLimitInfo) {
+            RateLimit.setLatest(model.providerID, rateLimitInfo)
+          }
+        } catch {
+          // Ignore errors parsing rate limit headers
+        }
+
+        return response
+      }
+
+      // Special case: google-vertex-anthropic uses a subpath import
+      const bundledKey =
+        model.providerID === "google-vertex-anthropic" ? "@ai-sdk/google-vertex/anthropic" : model.api.npm
+      const bundledFn = BUNDLED_PROVIDERS[bundledKey]
+      if (bundledFn) {
+        log.info("using bundled provider", { providerID: model.providerID, pkg: bundledKey })
+        const loaded = bundledFn({
+          name: model.providerID,
+          ...options,
+        })
+        s.sdk.set(key, loaded)
+        return loaded as SDK
+      }
+
+      let installedPath: string
+      if (!model.api.npm.startsWith("file://")) {
+        installedPath = await BunProc.install(model.api.npm, "latest")
+      } else {
+        log.info("loading local provider", { pkg: model.api.npm })
+        installedPath = model.api.npm
+      }
+
+      const mod = await import(installedPath)
+
+      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 })
+    }
+  }
+
+  export async function getProvider(providerID: string) {
+    return state().then((s) => s.providers[providerID])
+  }
+
+  export async function getModel(providerID: string, modelID: string) {
+    const s = await state()
+    const provider = s.providers[providerID]
+    if (!provider) {
+      const availableProviders = Object.keys(s.providers)
+      const matches = fuzzysort.go(providerID, availableProviders, { limit: 3, threshold: -10000 })
+      const suggestions = matches.map((m) => m.target)
+      throw new ModelNotFoundError({ providerID, modelID, suggestions })
+    }
+
+    const info = provider.models[modelID]
+    if (!info) {
+      const availableModels = Object.keys(provider.models)
+      const matches = fuzzysort.go(modelID, availableModels, { limit: 3, threshold: -10000 })
+      const suggestions = matches.map((m) => m.target)
+      throw new ModelNotFoundError({ providerID, modelID, suggestions })
+    }
+    return info
+  }
+
+  export async function getLanguage(model: Model) {
+    const s = await state()
+    const key = `${model.providerID}/${model.id}`
+    if (s.models.has(key)) return s.models.get(key)!
+
+    const provider = s.providers[model.providerID]
+    const sdk = await getSDK(model)
+
+    try {
+      const language = s.modelLoaders[model.providerID]
+        ? await s.modelLoaders[model.providerID](sdk, model.api.id, provider.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
+    }
+  }
+
+  export async function closest(providerID: string, query: string[]) {
+    const s = await 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,
+          }
+      }
+    }
+  }
+
+  export async function getSmallModel(providerID: string) {
+    const cfg = await Config.get()
+
+    if (cfg.small_model) {
+      const parsed = parseModel(cfg.small_model)
+      return getModel(parsed.providerID, parsed.modelID)
+    }
+
+    const provider = await state().then((state) => state.providers[providerID])
+    if (provider) {
+      let priority = [
+        "claude-haiku-4-5",
+        "claude-haiku-4.5",
+        "3-5-haiku",
+        "3.5-haiku",
+        "gemini-2.5-flash",
+        "gpt-5-nano",
+      ]
+      // claude-haiku-4.5 is considered a premium model in github copilot, we shouldn't use premium requests for title gen
+      if (providerID === "github-copilot") {
+        priority = priority.filter((m) => m !== "claude-haiku-4.5")
+      }
+      if (providerID.startsWith("opencode")) {
+        priority = ["gpt-5-nano"]
+      }
+      for (const item of priority) {
+        for (const model of Object.keys(provider.models)) {
+          if (model.includes(item)) return getModel(providerID, model)
+        }
+      }
+    }
+
+    // Check if Cerebras provider is available before using it (Cerebras Code CLI default)
+    const cerebrasProvider = await state().then((state) => state.providers["cerebras"])
+    if (cerebrasProvider && cerebrasProvider.models["zai-glm-4.7"]) {
+      return getModel("cerebras", "zai-glm-4.7")
+    }
+
+    // Fallback to opencode provider
+    const opencodeProvider = await state().then((state) => state.providers["opencode"])
+    if (opencodeProvider && opencodeProvider.models["gpt-5-nano"]) {
+      return getModel("opencode", "gpt-5-nano")
+    }
+
+    return undefined
+  }
+
+  const priority = ["zai-glm-4.7", "gpt-5", "claude-sonnet-4", "big-pickle", "gemini-3-pro"]
+  export function sort(models: Model[]) {
+    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 async function defaultModel() {
+    const cfg = await Config.get()
+    if (cfg.model) return parseModel(cfg.model)
+
+    const provider = await list()
+      .then((val) => Object.values(val))
+      .then((x) => x.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,
+    }
+  }
+
+  export function parseModel(model: string) {
+    const [providerID, ...rest] = model.split("/")
+    return {
+      providerID: providerID,
+      modelID: rest.join("/"),
+    }
+  }
+
+  export const ModelNotFoundError = NamedError.create(
+    "ProviderModelNotFoundError",
+    z.object({
+      providerID: z.string(),
+      modelID: z.string(),
+      suggestions: z.array(z.string()).optional(),
+    }),
+  )
+
+  export const InitError = NamedError.create(
+    "ProviderInitError",
+    z.object({
+      providerID: z.string(),
+    }),
+  )
+}