How do I get an API key?
A: Go to the SoMark API workbench to get your API key. Each user gets free parsing quota of up to500 pages per day and 2000 pages per month. After you exceed the free quota, the system automatically starts consuming paid quota.
Why do I hit QPS limits, and what should I do?
A: Each account is limited to QPS 4. Exceeding that rate within a short period triggers rate limiting. Control the request rate, submit each file only once, and query the result afterward.What file formats does SoMark support?
A: SoMark currently supports the following file formats:pdf, png, jpg, jpeg, bmp, tiff, jp2, dib, ppm, pgm, pbm, gif, heic, heif, webp, xpm, tga, dds, xbm, doc, docx, ppt, pptx, xlsx, xlsm, and xls.
What output formats are returned after parsing?
A: SoMark returns Markdown text by default. It also supports JSON structured output with detailed document elements, coordinates, and related metadata.What can you build with the SoMark plugin in Coze/Dify?
A:-
Document chat bot
With SoMark, users can upload PDFs, images, and other files directly in a Coze chat window, then ask questions such as “What are the key terms in this contract?” or “Summarize this annual report.” The bot parses the document in real time and returns an answer in the conversation. -
Information extraction workflow
Build an automated flow where users upload invoices, IDs, or contracts, then SoMark extracts key fields such as amount, document number, and expiration date, and sends the results to spreadsheets, notifications, or approval steps without manual entry. -
Multimodal document bot
Handle complex documents that contain mixed text, tables, and formulas, such as academic papers and technical manuals. The bot can understand heading hierarchy, table content, and formula meaning for professional Q&A and structured summarization. -
Knowledge base ingestion assistant
Import large volumes of unstructured documents such as Word files, PDFs, and scanned copies. After SoMark converts them into structured data, you can load them into a Coze knowledge base and quickly build an internal knowledge bot. -
Cross-platform document assistant
Publish the bot to WeChat, Feishu, and other messaging platforms so users can send documents in chat and receive summaries, translations, or key information extraction directly in their daily workflow.
How do I store images locally for Markdown that has already been parsed?
A: If the Markdown still contains remote image URLs, use the document image local storage Skill to download the images and rewrite their links. Use the Python example or full Skill script when integrating the workflow into your own program.Does the document image local storage Skill modify the source Markdown?
A: No. It writes a newmain.md and images/ directory in the output package. The source Markdown remains unchanged.
Does the document image local storage Skill create a ZIP automatically?
A: No. It creates a directory package. Compress the directory separately after verifying the output.Why are locally stored images converted to JPEG?
A: Re-encoding verifies that the downloaded content is a decodable image and provides a stable extension. Images with transparency use a white background.What happens if some images fail to download while being stored locally?
A: For one file, the Skill writes the newmain.md only after every image succeeds. In batch mode, it records the failed document, continues with other documents, and exits with a nonzero code.
Can the document image local storage Skill process multiple Markdown files in batch?
A: Yes. Pass a directory to the full Skill script. It recursively processes.md and .markdown files and creates a separate package for each document. The simplified Python example on this page processes one file only.