Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -4,19 +4,24 @@ import networkx as nx
|
|
4 |
from pyvis.network import Network
|
5 |
import tempfile
|
6 |
import openai
|
|
|
|
|
|
|
|
|
|
|
7 |
|
8 |
# ---------------------------
|
9 |
# Model Loading & Caching
|
10 |
# ---------------------------
|
11 |
@st.cache_resource(show_spinner=False)
|
12 |
def load_summarizer():
|
13 |
-
# Load a summarization pipeline from Hugging Face (
|
14 |
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
|
15 |
return summarizer
|
16 |
|
17 |
@st.cache_resource(show_spinner=False)
|
18 |
def load_text_generator():
|
19 |
-
# For
|
20 |
generator = pipeline("text-generation", model="gpt2")
|
21 |
return generator
|
22 |
|
@@ -24,61 +29,123 @@ summarizer = load_summarizer()
|
|
24 |
generator = load_text_generator()
|
25 |
|
26 |
# ---------------------------
|
27 |
-
#
|
28 |
# ---------------------------
|
|
|
|
|
|
|
|
|
|
|
|
|
29 |
def generate_ideas_with_openai(prompt, api_key):
|
|
|
|
|
|
|
|
|
30 |
openai.api_key = api_key
|
31 |
output_text = ""
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
|
|
|
|
|
|
49 |
return output_text
|
50 |
|
51 |
-
|
52 |
-
|
53 |
-
|
54 |
-
|
55 |
-
|
56 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
57 |
|
58 |
# ---------------------------
|
59 |
-
# Streamlit
|
60 |
# ---------------------------
|
61 |
-
st.title("Graph of AI Ideas Application")
|
62 |
|
|
|
63 |
st.sidebar.header("Configuration")
|
64 |
-
generation_mode = st.sidebar.selectbox("Select Idea Generation Mode",
|
65 |
-
|
66 |
openai_api_key = st.sidebar.text_input("OpenAI API Key (for GPT-3.5 Streaming)", type="password")
|
67 |
|
68 |
-
# --- Section 1:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
69 |
st.header("Research Paper Input")
|
70 |
paper_abstract = st.text_area("Enter the research paper abstract:", height=200)
|
71 |
|
72 |
if st.button("Generate Ideas"):
|
73 |
if paper_abstract.strip():
|
74 |
st.subheader("Summarized Abstract")
|
75 |
-
#
|
76 |
summary = summarizer(paper_abstract, max_length=100, min_length=30, do_sample=False)
|
77 |
summary_text = summary[0]['summary_text']
|
78 |
st.write(summary_text)
|
79 |
|
80 |
st.subheader("Generated Research Ideas")
|
81 |
-
# Build a prompt
|
82 |
prompt = (
|
83 |
f"Based on the following research paper abstract, generate innovative and promising research ideas for future work.\n\n"
|
84 |
f"Paper Abstract:\n{paper_abstract}\n\n"
|
@@ -89,7 +156,7 @@ if st.button("Generate Ideas"):
|
|
89 |
if not openai_api_key.strip():
|
90 |
st.error("Please provide your OpenAI API Key in the sidebar.")
|
91 |
else:
|
92 |
-
with st.spinner("Generating ideas using OpenAI GPT-3.5..."):
|
93 |
ideas = generate_ideas_with_openai(prompt, openai_api_key)
|
94 |
st.write(ideas)
|
95 |
else:
|
@@ -99,21 +166,16 @@ if st.button("Generate Ideas"):
|
|
99 |
else:
|
100 |
st.error("Please enter a research paper abstract.")
|
101 |
|
102 |
-
# --- Section
|
103 |
st.header("Knowledge Graph Visualization")
|
104 |
st.markdown(
|
105 |
-
"Simulate a knowledge graph by entering paper details and their citation relationships
|
106 |
-
"Enter details in CSV format: **PaperID,Title,CitedPaperIDs** (CitedPaperIDs separated by ';'). "
|
107 |
"Example:\n\n`1,Paper A,2;3`\n`2,Paper B,`\n`3,Paper C,2`"
|
108 |
)
|
109 |
papers_csv = st.text_area("Enter paper details in CSV format:", height=150)
|
110 |
|
111 |
if st.button("Generate Knowledge Graph"):
|
112 |
if papers_csv.strip():
|
113 |
-
import pandas as pd
|
114 |
-
from io import StringIO
|
115 |
-
|
116 |
-
# Process the CSV text input
|
117 |
data = []
|
118 |
for line in papers_csv.splitlines():
|
119 |
parts = line.split(',')
|
@@ -123,9 +185,8 @@ if st.button("Generate Knowledge Graph"):
|
|
123 |
cited = parts[2].strip()
|
124 |
cited_list = [c.strip() for c in cited.split(';') if c.strip()]
|
125 |
data.append({"paper_id": paper_id, "title": title, "cited": cited_list})
|
126 |
-
|
127 |
if data:
|
128 |
-
# Build a directed graph
|
129 |
G = nx.DiGraph()
|
130 |
for paper in data:
|
131 |
G.add_node(paper["paper_id"], title=paper["title"])
|
@@ -139,7 +200,7 @@ if st.button("Generate Knowledge Graph"):
|
|
139 |
net.add_node(node, label=node_data["title"])
|
140 |
for source, target in G.edges():
|
141 |
net.add_edge(source, target)
|
142 |
-
#
|
143 |
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".html")
|
144 |
net.write_html(temp_file.name)
|
145 |
with open(temp_file.name, 'r', encoding='utf-8') as f:
|
|
|
4 |
from pyvis.network import Network
|
5 |
import tempfile
|
6 |
import openai
|
7 |
+
import requests
|
8 |
+
import feedparser
|
9 |
+
import pandas as pd
|
10 |
+
from io import StringIO
|
11 |
+
import asyncio
|
12 |
|
13 |
# ---------------------------
|
14 |
# Model Loading & Caching
|
15 |
# ---------------------------
|
16 |
@st.cache_resource(show_spinner=False)
|
17 |
def load_summarizer():
|
18 |
+
# Load a summarization pipeline from Hugging Face (e.g., facebook/bart-large-cnn)
|
19 |
summarizer = pipeline("summarization", model="facebook/bart-large-cnn")
|
20 |
return summarizer
|
21 |
|
22 |
@st.cache_resource(show_spinner=False)
|
23 |
def load_text_generator():
|
24 |
+
# For demonstration, we load a text-generation model such as GPT-2
|
25 |
generator = pipeline("text-generation", model="gpt2")
|
26 |
return generator
|
27 |
|
|
|
29 |
generator = load_text_generator()
|
30 |
|
31 |
# ---------------------------
|
32 |
+
# Idea Generation Functions
|
33 |
# ---------------------------
|
34 |
+
def generate_ideas_with_hf(prompt):
|
35 |
+
# Use Hugging Face's text-generation pipeline (less creative than GPT‑3.5)
|
36 |
+
results = generator(prompt, max_length=150, num_return_sequences=1)
|
37 |
+
idea_text = results[0]['generated_text']
|
38 |
+
return idea_text
|
39 |
+
|
40 |
def generate_ideas_with_openai(prompt, api_key):
|
41 |
+
"""
|
42 |
+
Generates research ideas using OpenAI's GPT‑3.5 model with streaming.
|
43 |
+
This function uses the latest OpenAI SDK v1.0 which supports asynchronous API calls.
|
44 |
+
"""
|
45 |
openai.api_key = api_key
|
46 |
output_text = ""
|
47 |
+
|
48 |
+
async def stream_chat():
|
49 |
+
nonlocal output_text
|
50 |
+
# Asynchronously call the chat completion endpoint with streaming enabled.
|
51 |
+
response = await openai.ChatCompletion.acreate(
|
52 |
+
model="gpt-3.5-turbo",
|
53 |
+
messages=[
|
54 |
+
{"role": "system", "content": "You are an expert AI research assistant who generates innovative research ideas."},
|
55 |
+
{"role": "user", "content": prompt},
|
56 |
+
],
|
57 |
+
stream=True,
|
58 |
+
)
|
59 |
+
st_text = st.empty() # Placeholder for streaming output
|
60 |
+
async for chunk in response:
|
61 |
+
delta = chunk["choices"][0].get("delta", {})
|
62 |
+
text_piece = delta.get("content", "")
|
63 |
+
output_text += text_piece
|
64 |
+
st_text.text(output_text)
|
65 |
+
|
66 |
+
asyncio.run(stream_chat())
|
67 |
return output_text
|
68 |
|
69 |
+
# ---------------------------
|
70 |
+
# arXiv API Integration
|
71 |
+
# ---------------------------
|
72 |
+
def fetch_arxiv_results(query, max_results=5):
|
73 |
+
"""
|
74 |
+
Queries arXiv's free API to fetch relevant papers.
|
75 |
+
"""
|
76 |
+
base_url = "http://export.arxiv.org/api/query?"
|
77 |
+
search_query = "search_query=all:" + query
|
78 |
+
start = "0"
|
79 |
+
max_results = str(max_results)
|
80 |
+
query_url = f"{base_url}{search_query}&start={start}&max_results={max_results}"
|
81 |
+
response = requests.get(query_url)
|
82 |
+
if response.status_code == 200:
|
83 |
+
feed = feedparser.parse(response.content)
|
84 |
+
results = []
|
85 |
+
for entry in feed.entries:
|
86 |
+
title = entry.title
|
87 |
+
summary = entry.summary
|
88 |
+
published = entry.published
|
89 |
+
link = entry.link
|
90 |
+
authors = ", ".join(author.name for author in entry.authors)
|
91 |
+
results.append({
|
92 |
+
"title": title,
|
93 |
+
"authors": authors,
|
94 |
+
"published": published,
|
95 |
+
"summary": summary,
|
96 |
+
"link": link
|
97 |
+
})
|
98 |
+
return results
|
99 |
+
else:
|
100 |
+
return []
|
101 |
|
102 |
# ---------------------------
|
103 |
+
# Streamlit Application Layout
|
104 |
# ---------------------------
|
105 |
+
st.title("Graph of AI Ideas Application with arXiv Integration and OpenAI SDK v1.0")
|
106 |
|
107 |
+
# Sidebar Configuration
|
108 |
st.sidebar.header("Configuration")
|
109 |
+
generation_mode = st.sidebar.selectbox("Select Idea Generation Mode",
|
110 |
+
["Hugging Face Open Source", "OpenAI GPT-3.5 (Streaming)"])
|
111 |
openai_api_key = st.sidebar.text_input("OpenAI API Key (for GPT-3.5 Streaming)", type="password")
|
112 |
|
113 |
+
# --- Section 1: arXiv Paper Search ---
|
114 |
+
st.header("arXiv Paper Search")
|
115 |
+
arxiv_query = st.text_input("Enter a search query for arXiv papers:")
|
116 |
+
|
117 |
+
if st.button("Search arXiv"):
|
118 |
+
if arxiv_query.strip():
|
119 |
+
with st.spinner("Searching arXiv..."):
|
120 |
+
results = fetch_arxiv_results(arxiv_query, max_results=5)
|
121 |
+
if results:
|
122 |
+
st.subheader("arXiv Search Results:")
|
123 |
+
for idx, paper in enumerate(results):
|
124 |
+
st.markdown(f"**{idx+1}. {paper['title']}**")
|
125 |
+
st.markdown(f"*Authors:* {paper['authors']}")
|
126 |
+
st.markdown(f"*Published:* {paper['published']}")
|
127 |
+
st.markdown(f"*Summary:* {paper['summary']}")
|
128 |
+
st.markdown(f"[Read more]({paper['link']})")
|
129 |
+
st.markdown("---")
|
130 |
+
else:
|
131 |
+
st.error("No results found or an error occurred with the arXiv API.")
|
132 |
+
else:
|
133 |
+
st.error("Please enter a valid query for the arXiv search.")
|
134 |
+
|
135 |
+
# --- Section 2: Research Paper Input and Idea Generation ---
|
136 |
st.header("Research Paper Input")
|
137 |
paper_abstract = st.text_area("Enter the research paper abstract:", height=200)
|
138 |
|
139 |
if st.button("Generate Ideas"):
|
140 |
if paper_abstract.strip():
|
141 |
st.subheader("Summarized Abstract")
|
142 |
+
# Use the Hugging Face summarizer to capture key points
|
143 |
summary = summarizer(paper_abstract, max_length=100, min_length=30, do_sample=False)
|
144 |
summary_text = summary[0]['summary_text']
|
145 |
st.write(summary_text)
|
146 |
|
147 |
st.subheader("Generated Research Ideas")
|
148 |
+
# Build a combined prompt based on the abstract and its summary
|
149 |
prompt = (
|
150 |
f"Based on the following research paper abstract, generate innovative and promising research ideas for future work.\n\n"
|
151 |
f"Paper Abstract:\n{paper_abstract}\n\n"
|
|
|
156 |
if not openai_api_key.strip():
|
157 |
st.error("Please provide your OpenAI API Key in the sidebar.")
|
158 |
else:
|
159 |
+
with st.spinner("Generating ideas using OpenAI GPT-3.5 with SDK v1.0..."):
|
160 |
ideas = generate_ideas_with_openai(prompt, openai_api_key)
|
161 |
st.write(ideas)
|
162 |
else:
|
|
|
166 |
else:
|
167 |
st.error("Please enter a research paper abstract.")
|
168 |
|
169 |
+
# --- Section 3: Knowledge Graph Visualization ---
|
170 |
st.header("Knowledge Graph Visualization")
|
171 |
st.markdown(
|
172 |
+
"Simulate a knowledge graph by entering paper details and their citation relationships in CSV format: **PaperID,Title,CitedPaperIDs** (CitedPaperIDs separated by ';').\n\n"
|
|
|
173 |
"Example:\n\n`1,Paper A,2;3`\n`2,Paper B,`\n`3,Paper C,2`"
|
174 |
)
|
175 |
papers_csv = st.text_area("Enter paper details in CSV format:", height=150)
|
176 |
|
177 |
if st.button("Generate Knowledge Graph"):
|
178 |
if papers_csv.strip():
|
|
|
|
|
|
|
|
|
179 |
data = []
|
180 |
for line in papers_csv.splitlines():
|
181 |
parts = line.split(',')
|
|
|
185 |
cited = parts[2].strip()
|
186 |
cited_list = [c.strip() for c in cited.split(';') if c.strip()]
|
187 |
data.append({"paper_id": paper_id, "title": title, "cited": cited_list})
|
|
|
188 |
if data:
|
189 |
+
# Build a directed graph using NetworkX
|
190 |
G = nx.DiGraph()
|
191 |
for paper in data:
|
192 |
G.add_node(paper["paper_id"], title=paper["title"])
|
|
|
200 |
net.add_node(node, label=node_data["title"])
|
201 |
for source, target in G.edges():
|
202 |
net.add_edge(source, target)
|
203 |
+
# Save the interactive visualization to an HTML file and embed it in Streamlit
|
204 |
temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".html")
|
205 |
net.write_html(temp_file.name)
|
206 |
with open(temp_file.name, 'r', encoding='utf-8') as f:
|