app.py

import os
from typing import List, Optional, Union
import gradio as gr
import spacy
from spacy.tokens import Doc, Span
from relik import Relik
from relik.inference.data.objects import TaskType, RelikOutput
from relik.retriever.pytorch_modules import GoldenRetriever
from relik.retriever.indexers.inmemory import InMemoryDocumentIndex
from pyvis.network import Network

# RELIK Models Setup
wikipedia_retriever = GoldenRetriever("relik-ie/encoder-e5-base-v2-wikipedia", device="cuda")
wikipedia_index = InMemoryDocumentIndex.from_pretrained("relik-ie/encoder-e5-base-v2-wikipedia-index", index_precision="bf16", device="cuda")
wikidata_retriever = GoldenRetriever("relik-ie/encoder-e5-small-v2-wikipedia-relations", device="cuda")
wikidata_index = InMemoryDocumentIndex.from_pretrained("relik-ie/encoder-e5-small-v2-wikipedia-relations-index", index_precision="bf16", device="cuda")

relik_models = {
    "sapienzanlp/relik-entity-linking-large": Relik.from_pretrained(
        "sapienzanlp/relik-entity-linking-large", device="cuda", index=wikipedia_index, retriever=wikipedia_retriever,
        reader_kwargs={"dataset_kwargs": {"use_nme": True}}
    ),
    "relik-ie/relik-relation-extraction-small": Relik.from_pretrained(
        "relik-ie/relik-relation-extraction-small", index=wikidata_index, device="cuda", retriever=wikidata_retriever
    )
}

def get_span_annotations(response, doc):
    spans = []
    for span in response.spans:
        spans.append(Span(doc, span.start, span.end, span.label))
    colors = {span.label_: '#ff5733' for span in spans}  # Simple fixed color for demonstration
    return spans, colors

def generate_graph(spans, response, colors):
    g = Network(width="720px", height="600px", directed=True)
    for ent in spans:
        g.add_node(ent.text, label=ent.text, color=colors[ent.label_], size=15)
    seen_rels = set()
    for rel in response.triplets:
        if (rel.subject.text, rel.object.text, rel.label) in seen_rels:
            continue
        g.add_edge(rel.subject.text, rel.object.text, label=rel.label)
        seen_rels.add((rel.subject.text, rel.object.text, rel.label))
    html = g.generate_html()
    return f"""<iframe style="width: 100%; height: 600px;margin:0 auto" srcdoc='{html.replace("'", '"')}'></iframe>"""

def text_analysis(Text, Model, Relation_Threshold, Window_Size, Window_Stride):
    if Model not in relik_models:
        raise ValueError(f"Model {Model} not found.")
    relik = relik_models[Model]
    nlp = spacy.blank("xx")
    annotated_text = relik(Text, annotation_type="word", relation_threshold=Relation_Threshold, window_size=Window_Size, window_stride=Window_Stride)
    doc = Doc(nlp.vocab, words=[token.text for token in annotated_text.tokens])
    spans, colors = get_span_annotations(annotated_text, doc)
    doc.spans["sc"] = spans
    display_el = spacy.displacy.render(doc, style="span", options={"colors": colors}).replace("\n", " ")
    display_el = display_el.replace("border-radius: 0.35em;", "border-radius: 0.35em; white-space: nowrap;").replace("span style", "span id='el' style")
    display_re = generate_graph(spans, annotated_text, colors) if annotated_text.triplets else ""
    return display_el, display_re

theme = gr.themes.Base(primary_hue="rose", secondary_hue="rose", text_size="lg")
css = """
h1 { text-align: center; display: block; }
mark { color: black; }
#el { white-space: nowrap; }
"""

with gr.Blocks(fill_height=True, css=css, theme=theme) as demo:
    gr.Markdown("# ReLiK with P-FAF Integration")
    gr.Interface(
        text_analysis,
        [
            gr.Textbox(label="Input Text", placeholder="Enter sentence here..."),
            gr.Dropdown(list(relik_models.keys()), value="sapienzanlp/relik-entity-linking-large", label="Relik Model"),
            gr.Slider(minimum=0, maximum=1, step=0.05, value=0.5, label="Relation Threshold"),
            gr.Slider(minimum=16, maximum=128, step=16, value=32, label="Window Size"),
            gr.Slider(minimum=8, maximum=64, step=8, value=16, label="Window Stride")
        ],
        [gr.HTML(label="Entities"), gr.HTML(label="Relations")],
        examples=[
            ["Michael Jordan was one of the best players in the NBA."],
            ["Noam Chomsky is a renowned linguist and cognitive scientist."]
        ],
        allow_flagging="never"
    )
    if __name__ == "__main__":
        demo.launch()