What might be the reason if DeepResearch doesn't seem to analyze an uploaded PDF or image that you provided?

If DeepResearch isn’t analyzing an uploaded PDF or image, the most likely causes fall into three categories: file format compatibility, resource limitations, or content extraction challenges. Let’s break these down for clarity.

First, file format or structure issues could prevent analysis. For example, PDFs might be scanned images (not searchable text), encrypted, or use unsupported compression methods. DeepResearch may rely on optical character recognition (OCR) to extract text from images, but if the PDF is a low-resolution scan, OCR could fail. Similarly, image formats like TIFF or HEIC might not be supported, even if common formats like PNG or JPEG are. Developers should verify if the file meets the tool’s documented requirements. For instance, a PDF with embedded fonts or complex vector graphics might parse correctly in a viewer like Adobe Acrobat but crash a lightweight parser used by DeepResearch. Checking the tool’s logs for errors like “unsupported format” or “corrupted file” can confirm this.

Second, resource constraints might block processing. Large files (e.g., a 500-page PDF or a 50MB image) could exceed memory limits or timeouts set in DeepResearch’s configuration. For instance, a serverless function processing uploads might terminate after 30 seconds, leaving large files unprocessed. Similarly, concurrent uploads could overload system resources, causing silent failures. Developers can test this by uploading smaller files or checking monitoring tools for CPU/memory spikes. If the issue persists, scaling up compute resources or optimizing the pipeline (e.g., splitting PDFs into smaller chunks) might resolve it.

Third, content extraction logic might fail silently. For example, DeepResearch’s text extraction library (like PyPDF2 or Tesseract) could misinterpret certain layouts, such as multi-column text, tables, or handwritten content. A PDF with text rendered as vector paths instead of standard fonts might return empty results. Similarly, images with overlapping objects or poor contrast might not trigger analysis. Developers can debug this by testing the same file with underlying tools directly—like running Tesseract OCR on an image—to isolate the issue. Updating dependencies or adding preprocessing steps (e.g., image binarization for better OCR accuracy) could address gaps in the extraction pipeline.