baseline

docs/README-search.md

@@ -0,0 +1,25 @@
+# Search + Upload TODO (Internal)
+
+Goal: Unified search across DB data and file contents (PDF, DOCX), starting with MariaDB FULLTEXT and private file storage.
+
+## MVP Steps
+1) Create `documents` table (uuid, original_name, mime, size, storage_path, content_text, created).
+2) Add FULLTEXT index on `content_text` (and later title/metadata fields if needed).
+3) Build `FileUploadService`:
+   - Validate size/type (PDF/DOCX).
+   - Store outside `/web` (e.g. `storage/uploads/{uuid}/file.ext`).
+   - Extract text:
+     - PDF: `pdftotext` (poppler-utils).
+     - DOCX: unzip + parse `word/document.xml` (fallback: LibreOffice headless).
+   - Save `content_text` + metadata.
+4) Add upload action + form (admin area).
+5) Add search endpoint:
+   - Search DB entities.
+   - Search `documents.content_text`.
+   - Merge results into one list.
+
+## Future Enhancements
+- Async extraction (queue/worker) for large files.
+- Dedicated search engine (Meilisearch/Typesense) when scale/relevance needs grow.
+- Result ranking with field weighting (title > body).
+- Optional language detection / stemming.