@happy-creative/iroder 1.0.0

package/src/provider/error.ts

@@ -0,0 +1,197 @@
+import { APICallError } from "ai"
+import { STATUS_CODES } from "http"
+import { iife } from "@/util/iife"
+import type { ProviderID } from "./schema"
+
+export namespace ProviderError {
+  // Adapted from overflow detection patterns in:
+  // https://github.com/badlogic/pi-mono/blob/main/packages/ai/src/utils/overflow.ts
+  const OVERFLOW_PATTERNS = [
+    /prompt is too long/i, // Anthropic
+    /input is too long for requested model/i, // Amazon Bedrock
+    /exceeds the context window/i, // OpenAI (Completions + Responses API message text)
+    /input token count.*exceeds the maximum/i, // Google (Gemini)
+    /maximum prompt length is \d+/i, // xAI (Grok)
+    /reduce the length of the messages/i, // Groq
+    /maximum context length is \d+ tokens/i, // OpenRouter, DeepSeek, vLLM
+    /exceeds the limit of \d+/i, // GitHub Copilot
+    /exceeds the available context size/i, // llama.cpp server
+    /greater than the context length/i, // LM Studio
+    /context window exceeds limit/i, // MiniMax
+    /exceeded model token limit/i, // Kimi For Coding, Moonshot
+    /context[_ ]length[_ ]exceeded/i, // Generic fallback
+    /request entity too large/i, // HTTP 413
+    /context length is only \d+ tokens/i, // vLLM
+    /input length.*exceeds.*context length/i, // vLLM
+    /prompt too long; exceeded (?:max )?context length/i, // Ollama explicit overflow error
+    /too large for model with \d+ maximum context length/i, // Mistral
+    /model_context_window_exceeded/i, // z.ai non-standard finish_reason surfaced as error text
+  ]
+
+  function isOpenAiErrorRetryable(e: APICallError) {
+    const status = e.statusCode
+    if (!status) return e.isRetryable
+    // openai sometimes returns 404 for models that are actually available
+    return status === 404 || e.isRetryable
+  }
+
+  // Providers not reliably handled in this function:
+  // - z.ai: can accept overflow silently (needs token-count/context-window checks)
+  function isOverflow(message: string) {
+    if (OVERFLOW_PATTERNS.some((p) => p.test(message))) return true
+
+    // Providers/status patterns handled outside of regex list:
+    // - Cerebras: often returns "400 (no body)" / "413 (no body)"
+    // - Mistral: often returns "400 (no body)" / "413 (no body)"
+    return /^4(00|13)\s*(status code)?\s*\(no body\)/i.test(message)
+  }
+
+  function message(providerID: ProviderID, e: APICallError) {
+    return iife(() => {
+      const msg = e.message
+      if (msg === "") {
+        if (e.responseBody) return e.responseBody
+        if (e.statusCode) {
+          const err = STATUS_CODES[e.statusCode]
+          if (err) return err
+        }
+        return "Unknown error"
+      }
+
+      if (!e.responseBody || (e.statusCode && msg !== STATUS_CODES[e.statusCode])) {
+        return msg
+      }
+
+      try {
+        const body = JSON.parse(e.responseBody)
+        // try to extract common error message fields
+        const errMsg = body.message || body.error || body.error?.message
+        if (errMsg && typeof errMsg === "string") {
+          return `${msg}: ${errMsg}`
+        }
+      } catch {}
+
+      // If responseBody is HTML (e.g. from a gateway or proxy error page),
+      // provide a human-readable message instead of dumping raw markup
+      if (/^\s*<!doctype|^\s*<html/i.test(e.responseBody)) {
+        if (e.statusCode === 401) {
+          return "Unauthorized: request was blocked by a gateway or proxy. Your authentication token may be missing or expired — try running `iroder auth login <your provider URL>` to re-authenticate."
+        }
+        if (e.statusCode === 403) {
+          return "Forbidden: request was blocked by a gateway or proxy. You may not have permission to access this resource — check your account and provider settings."
+        }
+        return msg
+      }
+
+      return `${msg}: ${e.responseBody}`
+    }).trim()
+  }
+
+  function json(input: unknown) {
+    if (typeof input === "string") {
+      try {
+        const result = JSON.parse(input)
+        if (result && typeof result === "object") return result
+        return undefined
+      } catch {
+        return undefined
+      }
+    }
+    if (typeof input === "object" && input !== null) {
+      return input
+    }
+    return undefined
+  }
+
+  export type ParsedStreamError =
+    | {
+        type: "context_overflow"
+        message: string
+        responseBody: string
+      }
+    | {
+        type: "api_error"
+        message: string
+        isRetryable: false
+        responseBody: string
+      }
+
+  export function parseStreamError(input: unknown): ParsedStreamError | undefined {
+    const body = json(input)
+    if (!body) return
+
+    const responseBody = JSON.stringify(body)
+    if (body.type !== "error") return
+
+    switch (body?.error?.code) {
+      case "context_length_exceeded":
+        return {
+          type: "context_overflow",
+          message: "Input exceeds context window of this model",
+          responseBody,
+        }
+      case "insufficient_quota":
+        return {
+          type: "api_error",
+          message: "Quota exceeded. Check your plan and billing details.",
+          isRetryable: false,
+          responseBody,
+        }
+      case "usage_not_included":
+        return {
+          type: "api_error",
+          message: "To use Codex with your ChatGPT plan, upgrade to Plus: https://chatgpt.com/explore/plus.",
+          isRetryable: false,
+          responseBody,
+        }
+      case "invalid_prompt":
+        return {
+          type: "api_error",
+          message: typeof body?.error?.message === "string" ? body?.error?.message : "Invalid prompt.",
+          isRetryable: false,
+          responseBody,
+        }
+    }
+  }
+
+  export type ParsedAPICallError =
+    | {
+        type: "context_overflow"
+        message: string
+        responseBody?: string
+      }
+    | {
+        type: "api_error"
+        message: string
+        statusCode?: number
+        isRetryable: boolean
+        responseHeaders?: Record<string, string>
+        responseBody?: string
+        metadata?: Record<string, string>
+      }
+
+  export function parseAPICallError(input: { providerID: ProviderID; error: APICallError }): ParsedAPICallError {
+    const m = message(input.providerID, input.error)
+    const body = json(input.error.responseBody)
+    if (isOverflow(m) || input.error.statusCode === 413 || body?.error?.code === "context_length_exceeded") {
+      return {
+        type: "context_overflow",
+        message: m,
+        responseBody: input.error.responseBody,
+      }
+    }
+
+    const metadata = input.error.url ? { url: input.error.url } : undefined
+    return {
+      type: "api_error",
+      message: m,
+      statusCode: input.error.statusCode,
+      isRetryable: input.providerID.startsWith("openai")
+        ? isOpenAiErrorRetryable(input.error)
+        : input.error.isRetryable,
+      responseHeaders: input.error.responseHeaders,
+      responseBody: input.error.responseBody,
+      metadata,
+    }
+  }
+}

package/src/provider/models.ts

@@ -0,0 +1,159 @@
+import { Global } from "../global"
+import { Log } from "../util/log"
+import path from "path"
+import z from "zod"
+import { Installation } from "../installation"
+import { Flag } from "../flag/flag"
+import { lazy } from "@/util/lazy"
+import { Filesystem } from "../util/filesystem"
+import { Flock } from "@/util/flock"
+import { Hash } from "@/util/hash"
+
+// Try to import bundled snapshot (generated at build time)
+// Falls back to undefined in dev mode when snapshot doesn't exist
+/* @ts-ignore */
+
+export namespace ModelsDev {
+  const log = Log.create({ service: "models.dev" })
+  const source = url()
+  const filepath = path.join(
+    Global.Path.cache,
+    source === "https://models.dev" ? "models.json" : `models-${Hash.fast(source)}.json`,
+  )
+  const ttl = 5 * 60 * 1000
+
+  export const Model = z.object({
+    id: z.string(),
+    name: z.string(),
+    family: z.string().optional(),
+    release_date: z.string(),
+    attachment: z.boolean(),
+    reasoning: z.boolean(),
+    temperature: z.boolean(),
+    tool_call: z.boolean(),
+    interleaved: z
+      .union([
+        z.literal(true),
+        z
+          .object({
+            field: z.enum(["reasoning_content", "reasoning_details"]),
+          })
+          .strict(),
+      ])
+      .optional(),
+    cost: z
+      .object({
+        input: z.number(),
+        output: z.number(),
+        cache_read: z.number().optional(),
+        cache_write: z.number().optional(),
+        context_over_200k: z
+          .object({
+            input: z.number(),
+            output: z.number(),
+            cache_read: z.number().optional(),
+            cache_write: z.number().optional(),
+          })
+          .optional(),
+      })
+      .optional(),
+    limit: z.object({
+      context: z.number(),
+      input: z.number().optional(),
+      output: z.number(),
+    }),
+    modalities: z
+      .object({
+        input: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
+        output: z.array(z.enum(["text", "audio", "image", "video", "pdf"])),
+      })
+      .optional(),
+    experimental: z.boolean().optional(),
+    status: z.enum(["alpha", "beta", "deprecated"]).optional(),
+    provider: z.object({ npm: z.string().optional(), api: z.string().optional() }).optional(),
+  })
+  export type Model = z.infer<typeof Model>
+
+  export const Provider = z.object({
+    api: z.string().optional(),
+    name: z.string(),
+    env: z.array(z.string()),
+    id: z.string(),
+    npm: z.string().optional(),
+    models: z.record(z.string(), Model),
+  })
+
+  export type Provider = z.infer<typeof Provider>
+
+  function url() {
+    return Flag.IRODER_MODELS_URL || "https://models.dev"
+  }
+
+  function fresh() {
+    return Date.now() - Number(Filesystem.stat(filepath)?.mtimeMs ?? 0) < ttl
+  }
+
+  function skip(force: boolean) {
+    return !force && fresh()
+  }
+
+  const fetchApi = async () => {
+    const result = await fetch(`${url()}/api.json`, {
+      headers: { "User-Agent": Installation.USER_AGENT },
+      signal: AbortSignal.timeout(10000),
+    })
+    return { ok: result.ok, text: await result.text() }
+  }
+
+  export const Data = lazy(async () => {
+    const result = await Filesystem.readJson(Flag.IRODER_MODELS_PATH ?? filepath).catch(() => {})
+    if (result) return result
+    // @ts-ignore
+    const snapshot = await import("./models-snapshot.js")
+      .then((m) => m.snapshot as Record<string, unknown>)
+      .catch(() => undefined)
+    if (snapshot) return snapshot
+    if (Flag.IRODER_DISABLE_MODELS_FETCH) return {}
+    return Flock.withLock(`models-dev:${filepath}`, async () => {
+      const result = await Filesystem.readJson(Flag.IRODER_MODELS_PATH ?? filepath).catch(() => {})
+      if (result) return result
+      const result2 = await fetchApi()
+      if (result2.ok) {
+        await Filesystem.write(filepath, result2.text).catch((e) => {
+          log.error("Failed to write models cache", { error: e })
+        })
+      }
+      return JSON.parse(result2.text)
+    })
+  })
+
+  export async function get() {
+    const result = await Data()
+    return result as Record<string, Provider>
+  }
+
+  export async function refresh(force = false) {
+    if (skip(force)) return ModelsDev.Data.reset()
+    await Flock.withLock(`models-dev:${filepath}`, async () => {
+      if (skip(force)) return ModelsDev.Data.reset()
+      const result = await fetchApi()
+      if (!result.ok) return
+      await Filesystem.write(filepath, result.text)
+      ModelsDev.Data.reset()
+    }).catch((e) => {
+      log.error("Failed to fetch models.dev", {
+        error: e,
+      })
+    })
+  }
+}
+
+if (!Flag.IRODER_DISABLE_MODELS_FETCH && !process.argv.includes("--get-yargs-completions")) {
+  ModelsDev.refresh()
+  setInterval(
+    async () => {
+      await ModelsDev.refresh()
+    },
+    60 * 1000 * 60,
+  ).unref()
+}