report-with-imageslisted

# Report with Images One task in, a finished `.docx` + `.pdf` (with images and captions) out. ## Pipeline 1. **Content** — gather facts on the topic (search + read sources). Structure as: title, intro, 3-6 sections (each a heading + 2-4 short paragraphs). 2. **Images** — search image sources, collect candidate URLs, pick the best per section. 3. **Download** — fetch images through a browser context (this also bypasses naive hotlink/referer blocks). If a URL returns 403/404, skip it and take the next candidate — do not stop. 4. **Assemble** — build the document with a real image per section and a caption; export DOCX and then PDF. ## Output contract - `<out>.docx` and `<out>.pdf` must both exist and be **> 5 KB**. - Every image has a caption and a source note. - Text reads in the user's language (BS/HR/SR/EN). ## Rules - The output path is the filename **without extension**. - An image may come from a URL or a local file path — support both. - Never publish anything anywhere; the document stays local. - Verify at the end: file exists, size > 5 KB, page count > 1. - If no image can be downloaded, deliver the document with a placeholder box and say so — never a broken image.