mirror of
https://github.com/open-webui/open-webui.git
synced 2026-03-22 22:21:27 -05:00
Improving Multilingual Support and Image Processing with Tika Engine in Open WebUI: Challenges and Recommendations #2798
No Assignees
Notifications
Due Date
No due date set.
Dependencies
No dependencies set.
Reference: github-starred/open-webui#2798
Reference in New Issue
Block a user
Delete Branch "%!s()"
Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?
Originally created by @ahm631 on GitHub (Nov 23, 2024).
I would like to express my sincere appreciation for your outstanding efforts in developing Open WebUI. I have been utilizing Tika as a file-reading engine, and it performs excellently with English text. However, I have encountered some challenges when working with other languages, such as Arabic.
After installing the tesseract-ocr-ara package on Tika, I did not observe any effect within the Open WebUI interface. Through testing with Postman, I discovered that it is necessary to specify the OCR language of the image by passing it via the X-Tika-OCRLanguage parameter. Currently, Open WebUI does not provide an option to set this language, and it seems that the system defaults to sending the primary language as English (eng).
I attempted to modify Tika by replacing tesseract-ocr-eng with tesseract-ocr-ara, which resulted in successful operation within Open WebUI. However, when the file contains English text, the system is unable to read it. Additionally, there is an issue when multiple languages are present in the file.
Therefore, I suggest implementing a feature in Open WebUI that allows users to specify the languages present in files or images before sending them to the Tika server, ensuring more accurate results.
When uploading an image directly to Open WebUI and querying the language model about it, the system is unable to read the content. In contrast, if the image is embedded within a PDF file, it is processed successfully with OCR, particularly for English text.
I recommend enhancing the support for processing standalone images in Open WebUI, enabling the system to apply OCR directly to images without the need to embed them in PDF files.
Thank you once again for your dedication, and I look forward to future improvements that will enhance multilingual support and image processing capabilities in Open WebUI.