Update app.py

app.py

@@ -64,18 +64,13 @@
 
 
 # app.py
-import os
-import re
-import gradio as gr
-import tempfile
+# app.py
+import os, re, gradio as gr, tempfile, json, requests, numpy as np
 from collections import defaultdict, Counter
 from pyvis.network import Network
 from rdflib import Graph, Namespace, URIRef
 from rdflib.namespace import RDF
-import json
-import numpy as np
 from sentence_transformers import SentenceTransformer, util
-import requests
 
 # -----------------------------
 # Hugging Face Inference helper
@@ -109,28 +104,27 @@ RELATION_PATTERNS = [
 ]
 
 def extract_graph(text):
-    # Use HF NER model (CPU-friendly)
     entities = re.findall(ENTITY_REGEX, text)
     edges = []
     for pat, label in RELATION_PATTERNS:
         for m in pat.finditer(text):
             src, _, dst = m.groups()
             edges.append((src, dst, label))
-            entities.extend([src,dst])
+            entities.extend([src, dst])
     return sorted(set(entities)), edges
 
 def render_pyvis(entities, edges):
-    net = Network(height="400px", width="100%", directed=True, bgcolor="#111", font_color="#eee")
+    net = Network(height="400px", width="100%", directed=True, bgcolor="#111", font_color="#eee", notebook=False, cdn_resources="in_line")
     net.barnes_hut(gravity=-20000, central_gravity=0.3, spring_length=160, spring_strength=0.05)
     for e in entities:
         net.add_node(e, label=e, title=e)
-    for s,d,r in edges:
-        net.add_edge(s,d,title=r,label=r)
+    for s, d, r in edges:
+        net.add_edge(s, d, title=r, label=r)
     with tempfile.TemporaryDirectory() as td:
         path = f"{td}/graph.html"
-        net.show(path)
-        with open(path,"r",encoding="utf-8") as f:
-            html=f.read()
+        net.write_html(path)
+        with open(path, "r", encoding="utf-8") as f:
+            html = f.read()
     return html
 
 # -----------------------------
@@ -140,7 +134,6 @@ ORG = Namespace("http://example.org/org/")
 REL = Namespace("http://example.org/rel/")
 LOC = Namespace("http://example.org/loc/")
 PER = Namespace("http://example.org/per/")
-
 PATTERNS = RELATION_PATTERNS
 
 def triples_from_text(text):
@@ -203,50 +196,76 @@ def hybrid_search(query, k=5):
 
 def synthesize_answer(query, hits):
     context="\n".join(ch for ch,_ in hits)
-    # Use small HF LLM for answer synthesis
     ans = hf_inference("google/flan-t5-small", f"Answer the question based on context:\nContext: {context}\nQuestion: {query}")
     return ans
 
+# -----------------------------
+# 4) NLP Pipeline (Summarize, Classify, NER)
+# -----------------------------
+def summarize_text(text):
+    return hf_inference("facebook/bart-large-cnn", text)
+
+def classify_text(text):
+    return hf_inference("facebook/bart-large-mnli", f"Classify: {text}")
+
+def ner_text(text):
+    return hf_inference("dslim/bert-base-NER", text, task="token-classification")
+
 # -----------------------------
 # Gradio UI
 # -----------------------------
 with gr.Blocks(title="All-in-One Graph Demo") as demo:
-    gr.Markdown("# All 4 Graph Demos in One")
-    
+    gr.Markdown("# 🧠 All 4 Graph + NLP Demos in One")
+
     inp_text = gr.Textbox(lines=10, label="Paste your paragraph")
+    question = gr.Textbox(label="Ask a question (for GraphRAG QA)")
     submit_btn = gr.Button("Process Everything")
-    
-    # 1) GraphGPT
-    graph_html = gr.HTML(label="GraphGPT Graph")
-    
-    # 2) KnowledgeGraphGPT
-    out_ttl = gr.Code(label="RDF Turtle")
-    out_jsonld = gr.Code(label="RDF JSON-LD")
-    
-    # 3) GraphRAG QA
-    stats = gr.Markdown()
-    question = gr.Textbox(label="Ask a question")
-    answer = gr.Textbox(label="Answer")
-    
+
+    with gr.Tab("GraphGPT"):
+        graph_html = gr.HTML()
+
+    with gr.Tab("KnowledgeGraphGPT"):
+        out_ttl = gr.Code(label="RDF Turtle")
+        out_jsonld = gr.Code(label="RDF JSON-LD")
+
+    with gr.Tab("GraphRAG"):
+        stats = gr.Markdown()
+        answer = gr.Textbox(label="Answer")
+
+    with gr.Tab("NLP Pipeline"):
+        summary_out = gr.Textbox(label="Summarization")
+        classify_out = gr.Textbox(label="Classification")
+        ner_out = gr.Textbox(label="NER")
+
     def pipeline_all(text, q):
-        # GraphGPT
+        # 1) GraphGPT
         ents, rels = extract_graph(text)
         html_graph = render_pyvis(ents, rels)
-        # KnowledgeGraphGPT
+
+        # 2) KnowledgeGraphGPT
         g = triples_from_text(text)
         ttl = g.serialize(format="turtle")
         jsonld = g.serialize(format="json-ld", indent=2)
-        # GraphRAG
-        texts_store.clear()
-        texts_store.append(text)
+
+        # 3) GraphRAG
+        texts_store.clear(); texts_store.append(text)
         build_index(texts_store)
-        hits = hybrid_search(q)
-        ans = synthesize_answer(q, hits)
+        hits = hybrid_search(q if q.strip() else "What is this text about?")
+        ans = synthesize_answer(q if q.strip() else "What is this text about?", hits)
         stats_txt = f"GraphRAG Chunks: {len(chunks)}, Graph edges: {len(edges)}"
-        return html_graph, ttl, jsonld, stats_txt, ans
-    
-    submit_btn.click(pipeline_all, inputs=[inp_text, question], 
-                     outputs=[graph_html, out_ttl, out_jsonld, stats, answer])
+
+        # 4) NLP Pipeline
+        summary = summarize_text(text)
+        classification = classify_text(text)
+        entities = ner_text(text)
+
+        return html_graph, ttl, jsonld, stats_txt, ans, summary, classification, entities
+
+    submit_btn.click(
+        pipeline_all,
+        inputs=[inp_text, question],
+        outputs=[graph_html, out_ttl, out_jsonld, stats, answer, summary_out, classify_out, ner_out]
+    )
 
 if __name__=="__main__":
-    demo.launch()
+    demo.launch(server_name="0.0.0.0", server_port=7860)