Configure Arabic PDF Sentiment

I already have OpenWebUI up and running with a Retrieval-Augmented Generation (RAG) pipeline, but it is still limited to English sources. What I need now is for the same interface to accept Arabic-language PDFs, retrieve the text accurately, and classify each chunk of content as positive, neutral, or negative. Here is what has to happen: the workflow must ingest uploaded Arabic PDFs, extract the text reliably (right-to-left layout and diacritics included), hand that text to a sentiment model, then display a simple three-way result inside OpenWebUI. I do not require granular scores—just the clear Intermediate triad of positive / neutral / negative for each passage. OpenWebUI is containerised, so the solution should plug into the existing Docker setup with minimal extra services. Python, LangChain, PyPDF2, or similar libraries are fine as long as they run efficiently with Arabic script. If you prefer a hosted model (e.g., an Arabic-tuned sentiment endpoint) that can also work; latency just needs to stay reasonable. Deliverables • Updated containers / scripts that let OpenWebUI read Arabic PDFs • Sentiment component returning positive, neutral, or negative per extracted chunk • Clear README or inline comments explaining how to redeploy or extend the setup Once I can upload a sample PDF and see the three-way sentiment output inside the UI, the job is done.

Project ID: 39754805