app.py

import gradio as gr
from utils import submit_gradio_module, load_retrieval_results
from fuzzywuzzy import fuzz

HEADER = """<div style="text-align: center; margin-bottom: 20px;">
    <h1>The Arabic RAG Leaderboard</h1>
    <p style="font-size: 14px; color: #888;">The only leaderboard you will require for your RAG needs 🏆</p>
</div>
"""

ABOUT_SECTION = """
## About

The Arabic RAG Leaderboard is designed to evaluate and compare the performance of Retrieval-Augmented Generation (RAG) models on a set of retrieval and generative tasks. By leveraging a comprehensive evaluation framework, the leaderboard provides a detailed assessment of a model's ability to retrieve relevant information and generate accurate, coherent, and contextually appropriate responses.

### Why Focus on RAG Models?

The Arabic RAG Leaderboard is specifically designed to assess **RAG models**, which combine retrieval mechanisms with generative capabilities to enhance the quality and relevance of generated content. These models are particularly useful in scenarios where access to up-to-date and contextually relevant information is crucial. While foundational models can be evaluated, the primary focus is on RAG models that excel in both retrieval and generation tasks.

### How to Submit Your Model?

Navigate to the submission section below to submit your RAG model from the HuggingFace Hub for evaluation. Ensure that your model is public and the submitted metadata (precision, revision, #params) is accurate.

### Contact

For any inquiries or assistance, feel free to reach out through the community tab at [Navid-AI Community](https://huggingface.co/spaces/Navid-AI/The-Arabic-Rag-Leaderboard/discussions) or via [email](mailto:[email protected]).
"""

CITATION_BUTTON_LABEL = """
Copy the following snippet to cite these results
"""

CITATION_BUTTON_TEXT = """
@misc{AraGen,
  author = {Mohaned A. Rashad, Hamza Shahid},
  title = {The Arabic RAG Leaderboard},
  year = {2025},
  publisher = {Navid-AI},
  howpublished = "url{https://huggingface.co/spaces/Navid-AI/The-Arabic-Rag-Leaderboard}"
}
"""

df = load_retrieval_results()
df[["Model"]] = df[["Model"]].applymap(lambda x: f'<a href="https://huggingface.co/{x}" target="_blank">{x}</a>')
print(df)

def search_leaderboard(model_name):
    if not model_name:
        return df
    
    threshold = 95  # You can adjust this value to make the search more or less strict

    def calculate_similarity(row):
        similarity = fuzz.partial_ratio(model_name.lower(), row['model'].lower())
        return similarity if similarity >= threshold else 0

    # Add a new column for similarity scores
    df['similarity'] = df.apply(calculate_similarity, axis=1)

    # Filter and sort the dataframe
    filtered_df = df[df['similarity'] > 0].sort_values('similarity', ascending=False)

    # Remove the similarity column before returning
    filtered_df = filtered_df.drop('similarity', axis=1)

    return filtered_df


def main():
    columns_to_show = ["Model", "Model Size", "Embedding Dimension", "Max Tokens", "Web Search Dataset (MRR)", "Web Search Dataset (nDCG@k=None)"]
    with gr.Blocks() as demo:
        gr.HTML(HEADER)
        
        with gr.Tabs():
            with gr.Tab("Retrieval"):
                with gr.Tabs():
                    with gr.Tab("Leaderboard"):
                        with gr.Row():
                            search_box_retrieval = gr.Textbox(
                                placeholder="Search for models...", 
                                label="Search", 
                                interactive=True
                            )

                        retrieval_leaderboard = gr.Dataframe(
                            df[columns_to_show],
                            datatype="markdown",
                            interactive=False
                        )
                        
                        # Submit the search box and the leaderboard
                        search_box_retrieval.submit(
                            search_leaderboard,
                            inputs=search_box_retrieval,
                            outputs=retrieval_leaderboard
                        )
                        
                    submit_gradio_module("Retriever")

            with gr.Tab("Reranking"):
                with gr.Tabs():
                    with gr.Tab("Leaderboard"):
                        search_box_reranker = gr.Textbox(
                                placeholder="Search for models...", 
                                label="Search", 
                                interactive=True
                            )
                        
                        reranker_leaderboard = gr.Dataframe(
                            df[columns_to_show],
                            interactive=False,
                            )

                        search_box_reranker.submit(
                            search_leaderboard,
                            inputs=search_box_reranker,
                            outputs=reranker_leaderboard
                        )
                    submit_gradio_module("Reranker")
                    
            # with gr.Tab("LLM Context Answering"):
            #     with gr.Tabs():
            #         with gr.Tab("Leaderboard"):
            #             pass
            #         submit_gradio_module("LLM")

            with gr.Row():
                with gr.Accordion("📙 Citation", open=False):
                    gr.Textbox(
                        value=CITATION_BUTTON_TEXT,
                        label=CITATION_BUTTON_LABEL,
                        lines=20,
                        elem_id="citation-button",
                        show_copy_button=True,
                    )

    demo.launch()

if __name__ == "__main__":
    main()