document to markdown

Signed-off-by: Kenneth Obsequio <k80308392@gmail.com>

controllers/admin/assets.controller.js

@@ -9,6 +9,7 @@ const s3 = require("../../services/s3.service");
 const mediaToken = require("../../services/mediaToken.service");
 const uploadProgress = require("../../services/uploadProgress.service");
 const { extractVideoMeta } = require("../../services/ffprobe.service");
+const documentConversion = require("../../services/documentConversion.service");
 const R = require('../../utils/response.util');
 const { paginate } = require("../../utils/paginate.util");
 const { adminExclude, jsonbSchemas, computedAttributes } = require("../../models/assets/assets.attributes");
@@ -60,6 +61,15 @@ function resolveChecksum(buffer) {
   return crypto.createHash("sha256").update(buffer).digest("hex");
 }
 
+function streamToBuffer(stream) {
+  return new Promise((resolve, reject) => {
+    const chunks = [];
+    stream.on("data", (chunk) => chunks.push(chunk));
+    stream.on("end", () => resolve(Buffer.concat(chunks)));
+    stream.on("error", reject);
+  });
+}
+
 function resolveResolution(width, height) {
   if (!width || !height) return null;
   const h = Math.min(width, height);
@@ -524,6 +534,12 @@ exports.streamUploadProgress = (req, res) => {
   uploadProgress.subscribe(req.params.uploadId, res);
 };
 
+// Same broadcaster, reused as-is for document-conversion job progress — it's
+// just a generic string-keyed SSE channel, nothing upload-specific about it.
+exports.streamConvertProgress = (req, res) => {
+  uploadProgress.subscribe(req.params.jobId, res);
+};
+
 // ─── UPLOAD (batch) ─────────────────────────────────────────────────────────
 //
 // Accepts multiple files under the "files" field in one multipart request,
@@ -573,6 +589,84 @@ exports.uploadAssetsBatch = async (req, res) => {
   return R.success(res, `${createdCount} of ${files.length} asset(s) uploaded.`, { results }, 201);
 };
 
+// ─── CONVERT TO MARKDOWN ────────────────────────────────────────────────────
+//
+// PDF/PPTX -> Markdown, text only (see services/documentConversion.service.js
+// for the compile/validate/automate stages and why OCR/images are out of
+// scope). Nothing here is written to the database — the caller (the Lesson
+// block builder) treats the response as a draft and only persists it if/when
+// the admin explicitly inserts it into a block and saves the lesson page.
+//
+// Mirrors the upload flow's SSE progress pattern exactly: the actual work
+// runs synchronously inside this request (uploadProgress.service.js is
+// reused as-is, keyed by a client-generated jobId) while stage transitions
+// are published for a live "Compiling / Validating / Generating" UI, same as
+// AddAsset.jsx already renders for uploads.
+//
+const CONVERTIBLE_EXTENSIONS = new Set(["pdf", "pptx"]);
+const MAX_CONVERT_SIZE_BYTES = 25 * 1024 * 1024; // 25MB — keeps this comfortably synchronous
+// compile() calls the MarkItDown sidecar over HTTP now instead of running
+// officeparser in-process, so this leaves a bit more room than the original
+// 45s for network/queueing overhead — MarkItDown's own parsing is plain
+// CPU-bound work, not ML inference, so it doesn't need much more than that.
+const CONVERT_TIMEOUT_MS = 60_000;
+
+exports.convertAssetToMarkdown = async (req, res) => {
+  const { jobId } = req.body;
+
+  try {
+    const { assetId } = req.params;
+    if (!assetId || assetId === "undefined") return R.error(res, "Invalid asset ID.", 400);
+
+    const asset = await Asset.findOne({ where: { asset_id: assetId, ...notDeleted } });
+    if (!asset) return R.error(res, "Asset not found.", 404);
+
+    const extension = (asset.extension || "").toLowerCase();
+    if (!CONVERTIBLE_EXTENSIONS.has(extension)) {
+      return R.error(res, "Only PDF and PPTX documents can be converted to Markdown.", 400);
+    }
+    if (asset.storage_provider !== "s3" || !asset.storage_key) {
+      return R.error(res, "This asset has no stored file to convert.", 400);
+    }
+    if (Number(asset.file_size) > MAX_CONVERT_SIZE_BYTES) {
+      return R.error(res, "File is too large to convert (25MB max).", 400);
+    }
+
+    const publish = (phase) => { if (jobId) uploadProgress.publish(jobId, { phase }); };
+    const abortController = new AbortController();
+    const timeout = setTimeout(() => abortController.abort(), CONVERT_TIMEOUT_MS);
+
+    let markdown, warnings, stats;
+    try {
+      publish("compiling");
+      const { stream } = await s3.getObjectStream(asset.storage_key);
+      const buffer = await streamToBuffer(stream);
+      const ast = await documentConversion.compile(buffer, extension, { signal: abortController.signal });
+
+      publish("validating");
+      const validated = documentConversion.validate(ast);
+
+      publish("generating");
+      const generated = await documentConversion.automate(ast, { signal: abortController.signal });
+
+      markdown = generated.markdown;
+      warnings = [...validated.warnings, ...generated.messages];
+      stats    = validated.stats;
+    } finally {
+      clearTimeout(timeout);
+    }
+
+    if (jobId) uploadProgress.complete(jobId, { phase: "done" });
+    return R.success(res, "Document converted.", { markdown, warnings, stats });
+
+  } catch (err) {
+    if (jobId) uploadProgress.complete(jobId, { phase: "error", message: err.message });
+    console.error("[ASSET][CONVERT TO MARKDOWN]", err);
+    if (err.status) return R.error(res, err.message, err.status);
+    return R.error(res, "Internal server error.", 500);
+  }
+};
+
 // ─── UPDATE ───────────────────────────────────────────────────────────────────
 
 exports.updateAsset = async (req, res) => {