import { tool } from "@lmstudio/sdk";
import { z } from "zod";
import { crawlPage } from "../services/crawlPage";
import { normalizeUrl } from "../utils/url";

export const crawlSpecificPageTool = tool({
    name: "crawlSpecificPage",
    description: "Crawl a webpage. Returns optimized text content with precise contextual inline links.",
    parameters: {
        url: z.string().describe("The absolute URL of the webpage to crawl"),
        include: z.array(z.enum(["content", "rawHtml", "links", "infoboxes", "recentUpdates"])).optional().describe("Explicit data types to extract."),
    },
    implementation: async (input: {
        url: string;
        include?: string[];
    }) => {
        try {
            const rawUrl = input.url;
            if (!rawUrl) return "Error: url is required";

            const normalizedUrl = normalizeUrl(rawUrl);
            // crawlPage handles blocked domains/extensions internally via isNoiseUrl — no need to pre-check
            const result = await crawlPage(normalizedUrl, "");
            if (!result.success) return `Crawl failed: ${result.error}`;

            const include = input.include ?? ["content"];
            const output: string[] = [];

            // Single pass over structuredContent for infoboxes/recentUpdates filtering
            let metadataSection: string | null = null;
            let updateSections: string[] = [];
            if (include.includes("infoboxes") || include.includes("recentUpdates")) {
                for (const section of result.structuredContent) {
                    if (section.includes("## Page Metadata")) {
                        metadataSection = section;
                    } else if (/update|news|change|patch|version|release/i.test(section)) {
                        updateSections.push(section);
                    }
                }
            }

            // Structured content (includes infoboxes when requested)
            if ((include.includes("content") || include.includes("links")) && result.structuredContent.length > 0) {
                output.push("## Structured Content");
                for (const section of result.structuredContent) {
                    output.push(section);
                }
            }

            // Raw text content — only as fallback when no structured sections exist
            if (include.includes("content") && !output.some(s => s.startsWith("## Structured Content"))) {
                const raw = result.content.slice(0, 6000).trim();
                if (raw) {
                    output.push("## Content");
                    output.push(raw);
                }
            }

            // Links
            if (include.includes("links")) {
                const links = result.links.map(link => {
                    try {
                        const parsed = new URL(link);
                        const anchor = parsed.hash ? parsed.hash.slice(1) : null;
                        return `- [${anchor ? `${parsed.pathname}#${anchor}` : parsed.pathname}](${link})`;
                    } catch {
                        return `- ${link}`;
                    }
                });
                if (links.length > 0) {
                    output.push("## Links");
                    output.push(links.join("\n"));
                }
            }

            // Infoboxes (extracted in single pass above)
            if (include.includes("infoboxes") && metadataSection) {
                output.push(metadataSection);
            }

            // Recent updates (extracted in single pass above)
            if (include.includes("recentUpdates") && updateSections.length > 0) {
                output.push("## Recent Updates");
                output.push(updateSections.join("\n\n"));
            }

            // rawHtml: use the already-fetched HTML from crawlPage result
            if (include.includes("rawHtml") && result.rawHtml) {
                output.push("## Raw HTML");
                output.push(result.rawHtml.slice(0, 500_000));
            }

            return output.length ? output.join("\n\n") : "No content found.";
        } catch (error) {
            const message = error instanceof Error ? error.message : String(error);
            return `Error: ${message}`;
        }
    },
});

import { tool } from "@lmstudio/sdk";
import { z } from "zod";
import { crawlPage } from "../services/crawlPage";
import { normalizeUrl } from "../utils/url";

export const crawlSpecificPageTool = tool({
    name: "crawlSpecificPage",
    description: "Crawl a webpage. Returns optimized text content with precise contextual inline links.",
    parameters: {
        url: z.string().describe("The absolute URL of the webpage to crawl"),
        include: z.array(z.enum(["content", "rawHtml", "links", "infoboxes", "recentUpdates"])).optional().describe("Explicit data types to extract."),
    },
    implementation: async (input: {
        url: string;
        include?: string[];
    }) => {
        try {
            const rawUrl = input.url;
            if (!rawUrl) return "Error: url is required";

            const normalizedUrl = normalizeUrl(rawUrl);
            // crawlPage handles blocked domains/extensions internally via isNoiseUrl — no need to pre-check
            const result = await crawlPage(normalizedUrl, "");
            if (!result.success) return `Crawl failed: ${result.error}`;

            const include = input.include ?? ["content"];
            const output: string[] = [];

            // Single pass over structuredContent for infoboxes/recentUpdates filtering
            let metadataSection: string | null = null;
            let updateSections: string[] = [];
            if (include.includes("infoboxes") || include.includes("recentUpdates")) {
                for (const section of result.structuredContent) {
                    if (section.includes("## Page Metadata")) {
                        metadataSection = section;
                    } else if (/update|news|change|patch|version|release/i.test(section)) {
                        updateSections.push(section);
                    }
                }
            }

            // Structured content (includes infoboxes when requested)
            if ((include.includes("content") || include.includes("links")) && result.structuredContent.length > 0) {
                output.push("## Structured Content");
                for (const section of result.structuredContent) {
                    output.push(section);
                }
            }

            // Raw text content — only as fallback when no structured sections exist
            if (include.includes("content") && !output.some(s => s.startsWith("## Structured Content"))) {
                const raw = result.content.slice(0, 6000).trim();
                if (raw) {
                    output.push("## Content");
                    output.push(raw);
                }
            }

            // Links
            if (include.includes("links")) {
                const links = result.links.map(link => {
                    try {
                        const parsed = new URL(link);
                        const anchor = parsed.hash ? parsed.hash.slice(1) : null;
                        return `- [${anchor ? `${parsed.pathname}#${anchor}` : parsed.pathname}](${link})`;
                    } catch {
                        return `- ${link}`;
                    }
                });
                if (links.length > 0) {
                    output.push("## Links");
                    output.push(links.join("\n"));
                }
            }

            // Infoboxes (extracted in single pass above)
            if (include.includes("infoboxes") && metadataSection) {
                output.push(metadataSection);
            }

            // Recent updates (extracted in single pass above)
            if (include.includes("recentUpdates") && updateSections.length > 0) {
                output.push("## Recent Updates");
                output.push(updateSections.join("\n\n"));
            }

            // rawHtml: use the already-fetched HTML from crawlPage result
            if (include.includes("rawHtml") && result.rawHtml) {
                output.push("## Raw HTML");
                output.push(result.rawHtml.slice(0, 500_000));
            }

            return output.length ? output.join("\n\n") : "No content found.";
        } catch (error) {
            const message = error instanceof Error ? error.message : String(error);
            return `Error: ${message}`;
        }
    },
});