"use strict";

import type { ExtensionAPI } from "@earendil-works/pi-coding-agent";
import { Type } from "typebox";
import { getConfig } from "../config";
import { processVideo } from "../media";
import { visionCompletion } from "../client";
import { VIDEO_ANALYSIS } from "../prompts";
import {
  DEFAULT_MAX_BYTES,
  DEFAULT_MAX_LINES,
  truncateHead,
} from "@earendil-works/pi-coding-agent";

export function registerVideoAnalysisTool(pi: ExtensionAPI) {
  pi.registerTool({
    name: "video_analysis",
    label: "Video Analysis",
    description:
      "Analyze video content (MP4, MOV, M4V, AVI, WebM, WMV up to 8MB). Identifies scenes, actions, events, and key information by sending video frames to GLM-4.6V for visual understanding. Supports local files and HTTP(S) URLs.",
    promptSnippet: "Analyze video content — identify scenes, actions, events, and key moments",
    promptGuidelines: [
      "Use video_analysis when the user provides a video file and wants to understand its content. Supported formats: mp4, mov, m4v, avi, webm, wmv. Max local file size: 8MB. Use HTTP(S) URLs for larger videos.",
    ],
    parameters: Type.Object({
      videoPath: Type.String({
        description:
          "Path to the video file (relative to cwd, supports @ prefix) or HTTP(S) URL. Supported: mp4, mov, m4v, avi, webm, wmv (max 8MB local).",
      }),
      prompt: Type.String({
        description:
          "What should the AI analyze about this video? (e.g., 'Describe what happens', 'Identify all UI interactions shown', 'What is the error in this screen recording?').",
      }),
    }),
    async execute(_toolCallId, params, signal, onUpdate, ctx) {
      const config = getConfig();

      onUpdate?.({
        content: [{ type: "text", text: "正在调用 GLM-4.6V 分析视频..." }],
      });

      const video = processVideo(params.videoPath, ctx.cwd, config.maxVideoSizeMB);

      const result = await visionCompletion(
        config,
        VIDEO_ANALYSIS,
        [video],
        params.prompt,
        signal
      );

      const truncated = truncateHead(result, {
        maxLines: DEFAULT_MAX_LINES,
        maxBytes: DEFAULT_MAX_BYTES,
      });

      let text = truncated.content;
      if (truncated.truncated) {
        text += `\n\n[输出已截断: ${truncated.outputLines}/${truncated.totalLines} 行]`;
      }

      return {
        content: [{ type: "text", text }],
        details: { raw: result },
      };
    },
  });
}