@gajae-code/coding-agent 0.17.2 → 0.17.5

package/vendor/markit-ai/dist/utils/output.js

@@ -0,0 +1,31 @@
+import chalk from "chalk";
+export const success = (msg) => console.log(chalk.green("✓"), msg);
+export const info = (msg) => console.log(chalk.blue("ℹ"), msg);
+export const warn = (msg) => console.log(chalk.yellow("⚠"), msg);
+export const error = (msg) => console.error(chalk.red("✗"), msg);
+export const bold = (s) => chalk.bold(s);
+export const dim = (s) => chalk.dim(s);
+export const cmd = (s) => chalk.cyan(s);
+export const bullet = (msg) => console.log(chalk.green("●"), msg);
+export const bulletDim = (msg) => console.log(chalk.dim("●"), msg);
+export const hint = (msg) => console.log(chalk.dim(`  ${msg}`));
+export const nextStep = (command) => console.log(`  ${cmd(command)}`);
+export const header = (title) => {
+    console.log();
+    console.log(chalk.bold(title));
+    console.log();
+};
+export function jsonOutput(data) {
+    console.log(JSON.stringify(data, null, 2));
+}
+export function output(options, handlers) {
+    if (options.json && handlers.json) {
+        jsonOutput(handlers.json());
+    }
+    else if (options.quiet && handlers.quiet) {
+        handlers.quiet();
+    }
+    else {
+        handlers.human();
+    }
+}

package/vendor/markit-ai/dist/utils/turndown.d.ts

@@ -0,0 +1,8 @@
+import TurndownService from "turndown";
+export declare function createTurndown(): TurndownService;
+/**
+ * Normalize HTML tables so turndown-plugin-gfm can handle them:
+ * - Wrap first row in <thead> if missing
+ * - Strip <p> tags inside <td>/<th> cells
+ */
+export declare function normalizeTablesHtml(html: string): string;

package/vendor/markit-ai/dist/utils/turndown.js

@@ -0,0 +1,70 @@
+import TurndownService from "turndown";
+import { gfm } from "turndown-plugin-gfm";
+export function createTurndown() {
+    const turndown = new TurndownService({
+        headingStyle: "atx",
+        codeBlockStyle: "fenced",
+        bulletListMarker: "-",
+    });
+    turndown.use(gfm);
+    // Fix strikethrough: GFM spec uses ~~ (double tilde), not ~ (single)
+    turndown.addRule("strikethrough", {
+        filter: ["del", "s", "strike"],
+        replacement(content) {
+            return `~~${content}~~`;
+        },
+    });
+    // Fix heading escaping: turndown escapes "1." to "1\." to avoid ordered lists
+    turndown.addRule("heading", {
+        filter: ["h1", "h2", "h3", "h4", "h5", "h6"],
+        replacement(content, node) {
+            const level = Number(node.nodeName.charAt(1));
+            const prefix = "#".repeat(level);
+            // Unescape unnecessary backslash before periods in headings
+            const cleaned = content.replace(/\\([.])/g, "$1").trim();
+            return `\n\n${prefix} ${cleaned}\n\n`;
+        },
+    });
+    // Override listItem rule to use single space after marker (turndown hardcodes 3)
+    turndown.addRule("listItem", {
+        filter: "li",
+        replacement(content, node, options) {
+            content = content
+                .replace(/^\n+/, "")
+                .replace(/\n+$/, "\n")
+                .replace(/\n/gm, "\n  ");
+            const parent = node.parentNode;
+            let prefix = `${options.bulletListMarker} `;
+            if (parent?.nodeName === "OL") {
+                const start = parent.getAttribute("start");
+                const index = Array.prototype.indexOf.call(parent.children, node);
+                prefix = `${(start ? Number(start) : 1) + index}. `;
+            }
+            return prefix + content + (node.nextSibling ? "\n" : "");
+        },
+    });
+    return turndown;
+}
+/**
+ * Normalize HTML tables so turndown-plugin-gfm can handle them:
+ * - Wrap first row in <thead> if missing
+ * - Strip <p> tags inside <td>/<th> cells
+ */
+export function normalizeTablesHtml(html) {
+    // Strip <p> tags inside table cells, joining multiple paragraphs with <br>
+    let result = html.replace(/<(td|th)([^>]*)>([\s\S]*?)<\/(td|th)>/gi, (_match, tag, attrs, inner, closeTag) => {
+        const stripped = inner
+            .replace(/^\s*<p>/i, "")
+            .replace(/<\/p>\s*$/i, "")
+            .replace(/<\/p>\s*<p>/gi, " ");
+        return `<${tag}${attrs}>${stripped}</${closeTag}>`;
+    });
+    // Add thead to tables that lack it
+    result = result.replace(/<table([^>]*)>\s*(?:<tbody>\s*)?(<tr[\s\S]*?<\/tr>)([\s\S]*?)<\/(?:tbody>\s*<\/)?table>/gi, (_match, attrs, firstRow, rest) => {
+        const theadRow = firstRow
+            .replace(/<td/gi, "<th")
+            .replace(/<\/td>/gi, "</th>");
+        return `<table${attrs}><thead>${theadRow}</thead><tbody>${rest}</tbody></table>`;
+    });
+    return result;
+}

package/vendor/markit-ai/package.json

@@ -0,0 +1,74 @@
+{
+  "name": "markit-ai",
+  "version": "0.5.3",
+  "description": "Convert anything to markdown. PDF, DOCX, PPTX, XLSX, HTML, EPUB, Jupyter, RSS, images, audio, URLs, and more. Pluggable converters, built-in LLM providers for image description and audio transcription. Works as a CLI and as a library.",
+  "type": "module",
+  "main": "dist/index.js",
+  "types": "dist/index.d.ts",
+  "exports": {
+    ".": {
+      "types": "./dist/index.d.ts",
+      "import": "./dist/index.js",
+      "require": "./dist/index.js",
+      "default": "./dist/index.js"
+    }
+  },
+  "bin": {
+    "markit": "./dist/main.js"
+  },
+  "files": [
+    "dist"
+  ],
+  "scripts": {
+    "build": "tsc",
+    "prepublishOnly": "npm run build",
+    "dev": "bun run src/main.ts",
+    "build:bun": "bun build --compile src/main.ts --outfile mill",
+    "build:bun-linux": "bun build --compile --target=bun-linux-x64 src/main.ts --outfile mill-linux-x64",
+    "build:bun-mac-arm": "bun build --compile --target=bun-darwin-arm64 src/main.ts --outfile mill-darwin-arm64",
+    "build:bun-mac-x64": "bun build --compile --target=bun-darwin-x64 src/main.ts --outfile mill-darwin-x64",
+    "test": "bun test",
+    "format": "bunx biome format --write src/",
+    "lint": "bunx biome lint src/",
+    "check": "bunx biome check src/"
+  },
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/Michaelliv/markit.git"
+  },
+  "homepage": "https://github.com/Michaelliv/markit#readme",
+  "bugs": {
+    "url": "https://github.com/Michaelliv/markit/issues"
+  },
+  "keywords": [
+    "markdown",
+    "pdf",
+    "docx",
+    "html",
+    "converter",
+    "ai",
+    "llm",
+    "agents"
+  ],
+  "license": "MIT",
+  "devDependencies": {
+    "@biomejs/biome": "^2.3.14",
+    "@types/jszip": "^3.4.1",
+    "@types/node": "^25.5.0",
+    "@types/turndown": "^5.0.5",
+    "typescript": "^5.8.0"
+  },
+  "dependencies": {
+    "chalk": "^5.6.2",
+    "commander": "^14.0.3",
+    "exifr": "^7.1.3",
+    "fast-xml-parser": "^5.5.9",
+    "jszip": "^3.10.1",
+    "mammoth": "^1.9.0",
+    "mupdf": "^1.27.0",
+    "music-metadata": "^11.12.3",
+    "rss-parser": "^3.13.0",
+    "turndown": "^7.2.0",
+    "turndown-plugin-gfm": "^1.0.2"
+  }
+}

package/vendor/markit-ai.patch

@@ -0,0 +1,49 @@
+diff --git a/dist/converters/pdf/extract.js b/dist/converters/pdf/extract.js
+index 7e3f02ea889f166c30975cce53f7cbe4de620e54..e1dcd4517128adf1021299cb5aa937bf2f98fab7 100644
+--- a/dist/converters/pdf/extract.js
++++ b/dist/converters/pdf/extract.js
+@@ -14,6 +14,8 @@
+ const SAME_LINE_Y_TOLERANCE = 2;
+ /** Max horizontal gap (pts) to merge adjacent fragments into one text box. */
+ const MAX_MERGE_GAP = 14;
++// Reuse the asynchronously initialized module when rendering extracted images.
++let mupdf;
+ /**
+  * Merge horizontally adjacent raw text items on the same visual line into
+  * word/phrase-level text boxes.
+@@ -429,7 +431,6 @@ function extractImageRegions(stext, pageNumber, pageHeight) {
+  * Uses mupdf's DrawDevice to render just the cropped area at 2x resolution.
+  */
+ export function renderImageRegion(input, region) {
+-    const mupdf = require("mupdf");
+     const doc = mupdf.Document.openDocument(input, "application/pdf");
+     const page = doc.loadPage(region.pageNumber - 1);
+     const pad = 10;
+@@ -453,12 +454,11 @@ export function renderImageRegion(input, region) {
+  * Extract text boxes and vector segments from all pages of a PDF buffer.
+  */
+ export async function extractPages(input) {
+-    let mupdf;
+     try {
+         mupdf = await import("mupdf");
+     }
+-    catch {
+-        throw new Error("PDF support requires 'mupdf'. Install it: npm install mupdf");
++    catch (cause) {
++        throw new Error("MuPDF module initialization failed", { cause });
+     }
+     const doc = mupdf.Document.openDocument(input, "application/pdf");
+     const pages = [];
+diff --git a/dist/markit.js b/dist/markit.js
+index d9ba0a06e0d42a855f353a4df3aa7d3bfab11100..281ffc15b67b0d15a638f01ebcbcc1d842e5cc61 100644
+--- a/dist/markit.js
++++ b/dist/markit.js
+@@ -225,7 +225,7 @@ export class Markit {
+             const details = errors
+                 .map((e) => `  ${e.converter}: ${e.error.message}`)
+                 .join("\n");
+-            throw new Error(`Conversion failed:\n${details}`);
++            throw new AggregateError(errors.map((entry) => entry.error), `Conversion failed:\n${details}`, { cause: errors[0].error });
+         }
+         throw new Error(`Unsupported format: ${streamInfo.extension || streamInfo.mimetype || "unknown"}`);
+     }