added huggingface tags

gtk installation fixed.

Early errors failing to launch the app solved.

python path addedto the system for running the app.

save error logs from run.bat

missing dependency added.

compilation wheel for llama.cpp on Windows.

folder reference to `requirements.txt` path.

Python version fix.

Fixed Python installation in Windows. Character continuation error.

Fixing admin permissions for installing python.

increased the max tokens generation for avoiding final rendering errors.

Generation between the token limits of the model is fixed.

Possibility to receive multiple pdfs. All images and pdfs uploaded are processed to extract their contents for the chunking process and the latter vector database creation.

Multiple queries option. You can specify which of those queries will correspond to flashcards or to summaries.

Changed OCR methology. Opted for EasyOCR since PaddleOCR was having conflicts with Intel cards.

Update for easy installation on Windows.

Use gradio 5.33.0 with python 3.10.12

Fix python_version format to 3.10.12

Set python_version to 3.10 to fix audioop compatibility

Fix gradio version for Python 3.13 compatibility

Use precompiled llama-cpp-python wheel to avoid OOM during build

README description of the repository. Claude md file for memory.

Truncation logic for avoiding unparseable framgents

Implementation of the Gradio app. The app is able to create the flashcards and summaries of some controlled examples.

Explanation of the Gradio app implementation for the generation of the Summaries and flashcards PDFs

Update of multilingual response. Making sure spanish queries are answered in spanish.

evaluation of the quality of results of the RAG pipeline using the faithfullness, answer relevancy and context precision metrics. The model is performing acceptable with a low score on the context retreival. Additional metrics may be added such as retrieval recall in the future for better assessment.

Notebook showing the RAG pipeline implementation with the creation of the structured outputs following the output schema for flashcards and summaries.

gitignore update

Study of chunking methodologies and what is the best strategy for the current use cases we are interested in.

equation detection possible feature. In experimental state. Needs some more work.

document ingestion notebook applying Paddle OCR model for image-to-text parsing.

Initial commit