/**
 * QwenCloud Provider for pi
 *
 * Adds QwenCloud's Token Plan subscription as a pi provider, giving access to
 * Qwen3.8, DeepSeek V4, GLM-5.2, Wan image generation, and HappyHorse video
 * generation through QwenCloud's OpenAI-compatible API.
 *
 * Setup:
 *   1. Sign up at home.qwencloud.com and get a Token Plan
 *   2. Create an API key (API Keys section)
 *   3. Set QWENCLOUD_API_KEY env var, or run `pi /login` and select QwenCloud
 *   4. Install: pi install git:github.com/jellydn/pi-qwencloud-provider
 *   5. Use /model to select a QwenCloud model
 *
 * @module pi-qwencloud-provider
 */

import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
import { resolveApiBase, ENV_API_KEY, PROVIDER_NAME } from "./env.js";
import { resolveApiKey } from "./auth.js";
import { resolveModels } from "./models.js";
import { handleQwenCloudError } from "./error-handler.js";
import { getApiKey as oauthGetApiKey, login, refreshToken } from "./oauth.js";
import { generateAndDownloadWanImage } from "./wan.js";
import { generateAndDownloadHappyHorseVideo } from "./happyhorse.js";

// QwenCloud exposes a standard OpenAI-compatible chat completions API.
// It supports both `system` and `developer` roles, so `supportsDeveloperRole`
// is true by default for all models.

// ─── Extension Entry Point ─────────────────────────────────────────────────

export default async function (pi: ExtensionAPI) {
  const apiBase = resolveApiBase();

  // Attempt dynamic model discovery from the QwenCloud API. Falls back to the
  // static MODELS array on any error (network failure, 404, parse error).
  // The fetch has a 5-second timeout so startup is never blocked for long.
  const apiKey = resolveApiKey();
  const models = await resolveModels(apiKey, { apiBase });

  // Only register the $QWENCLOUD_API_KEY sigil when the env var is set at
  // extension load time. OAuth-only installs should not advertise an
  // unconfigured env-key fallback.
  const envApiKey = process.env[ENV_API_KEY]?.trim();

  pi.registerProvider(PROVIDER_NAME, {
    name: "QwenCloud",
    baseUrl: apiBase,
    // Match the clinepass pattern: only pass apiKey when the env var
    // is set. When absent, pi falls back to the OAuth flow which reads
    // from ~/.pi/agent/auth.json natively.
    ...(envApiKey ? { apiKey: envApiKey } : {}),
    authHeader: true,
    // QwenCloud uses the standard OpenAI Chat Completions format, so pi's
    // built-in OpenAI streaming handles SSE + tool calls + usage.
    api: "openai-completions",
    oauth: {
      name: "QwenCloud",
      login,
      refreshToken,
      getApiKey: oauthGetApiKey,
    },
    // Spread the model object so all fields propagate to pi automatically.
    // Only `input` needs transformation: readonly tuple → mutable array.
    models: models.map((model) => ({
      ...model,
      input: [...model.input],
    })),
  });

  // ─── Error Surface ─────────────────────────────────────────────────────
  //
  // QwenCloud returns standard HTTP error codes. Without this handler, the
  // user sees a generic "Provider returned an error stop reason" message.
  // The handler in error-handler.ts owns the full pipeline:
  // filter → classify → deliver.
  pi.on("message_end", handleQwenCloudError);

  // ─── Wan Image Generation Slash Command ───────────────────────────────
  //
  // Wan uses a separate synchronous API endpoint (not chat/completions).
  // The /wan slash command generates images from text prompts and saves
  // them to the current working directory.
  //
  // Usage: /wan a cyberpunk cat on a rainy street
  // Options: /wan --model wan2.7-image-pro --size 2K a photorealistic dog
  pi.registerCommand("wan", {
    description: "Generate an image with QwenCloud Wan (e.g. /wan a cyberpunk cat)",
    async handler(args, ctx) {
      // Parse optional flags
      let model = "wan2.7-image";
      let size = "1K";
      let prompt = args.trim();

      const modelMatch = prompt.match(/^--model\s+(\S+)\s*/);
      if (modelMatch) {
        model = modelMatch[1];
        prompt = prompt.slice(modelMatch[0].length).trim();
      }

      const sizeMatch = prompt.match(/^--size\s+(\S+)\s*/);
      if (sizeMatch) {
        size = sizeMatch[1];
        prompt = prompt.slice(sizeMatch[0].length).trim();
      }

      if (!prompt) {
        ctx.ui.notify("Usage: /wan [--model <model>] [--size <size>] <prompt>", "warning");
        ctx.ui.notify("Models: wan2.7-image, wan2.7-image-pro  |  Sizes: 1K, 2K, 4K", "info");
        return;
      }

      try {
        ctx.ui.setWorkingMessage(`Generating image with ${model}...`);
        const result = await generateAndDownloadWanImage(prompt, ctx.cwd, {
          model,
          size,
        });
        ctx.ui.setWorkingMessage(undefined);
        ctx.ui.notify(`Wan image saved: ${result.localPath}`, "info");
      } catch (err: unknown) {
        ctx.ui.setWorkingMessage(undefined);
        const message = err instanceof Error ? err.message : "Unknown error";
        ctx.ui.notify(`Wan generation failed: ${message}`, "error");
      }
    },
  });
}