samwjya commited on
Commit
44b3247
Β·
1 Parent(s): 76f8bb5

update ui

Browse files
Files changed (1) hide show
  1. ui/app.py +87 -75
ui/app.py CHANGED
@@ -1,15 +1,19 @@
1
- import sys
2
  import os
3
 
4
- sys.path.append(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
5
-
6
- from dotenv import load_dotenv
7
- load_dotenv()
8
 
 
9
  import streamlit as st
10
- from groq import Groq
11
- from retriever import fetch_papers
12
- from reranker import load_model, rerank, AVAILABLE_MODELS, DEFAULT_MODEL
 
 
 
 
 
 
13
 
14
  st.set_page_config(page_title="PaperRank", layout="wide")
15
 
@@ -21,10 +25,48 @@ TIER_CONFIG = [
21
  ]
22
 
23
 
24
- @st.cache_resource
25
- def get_model(model_name: str):
26
- return load_model(model_name)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27
 
 
28
 
29
  def tier_badge(score: float) -> str:
30
  if score >= 0.90:
@@ -43,8 +85,6 @@ def tier_badge(score: float) -> str:
43
 
44
  # ── Sidebar ────────────────────────────────────────────────────────────────────
45
  with st.sidebar:
46
- # st.title("PaperRank")
47
- # st.caption("CrossEncoder reranking on live academic papers.")
48
  st.title("PaperRank")
49
  st.caption("Research Paper Finder")
50
  st.divider()
@@ -85,46 +125,39 @@ with st.sidebar:
85
  if run and query.strip():
86
  st.session_state.pop("summary", None)
87
  st.session_state.pop("summary_query", None)
88
- with st.spinner(f"[1/3] Fetching up to {limit} papers from OpenAlex..."):
89
  try:
90
- baseline_papers = fetch_papers(query, limit=limit)
91
  except Exception as e:
92
- st.error(f"Fetch failed: {e}")
93
  st.stop()
94
 
95
- st.success(f"[1/3] Retrieved {len(baseline_papers)} papers with abstracts.")
96
- st.session_state["baseline_papers"] = baseline_papers
97
  st.session_state["active_query"] = query
98
-
99
- with st.spinner(f"[2/3] Reranking {len(baseline_papers)} papers with {model_name}..."):
100
- model = get_model(model_name)
101
- st.session_state["reranked_papers"] = rerank(query, baseline_papers, model)
102
- st.session_state["active_model"] = model_name
103
-
104
- # Rerun so the year slider updates to the actual bounds before results render
105
  st.rerun()
106
 
107
 
108
- # ── Pipeline: re-rerank when model changes (no new fetch needed) ───────────────
109
  if (
110
  not run
111
- and "baseline_papers" in st.session_state
112
  and st.session_state.get("active_model") != model_name
113
  ):
114
  with st.spinner(f"Switching model to {model_name}..."):
115
- model = get_model(model_name)
116
- st.session_state["reranked_papers"] = rerank(
117
- st.session_state["active_query"],
118
- st.session_state["baseline_papers"],
119
- model,
120
- )
121
- st.session_state["active_model"] = model_name
 
 
122
  st.rerun()
123
 
124
 
125
  # ── Main area ──────────────────────────────────────────────────────────────────
126
- # st.title("PaperRank")
127
- # st.caption("CrossEncoder reranking on live academic papers.")
128
  st.title("PaperRank")
129
  st.caption("Research Paper Finder")
130
  if "reranked_papers" not in st.session_state:
@@ -146,7 +179,6 @@ def in_year_range(paper: dict) -> bool:
146
 
147
  filtered_reranked = [p for p in reranked_papers if in_year_range(p)]
148
 
149
- # Tier counts from the full year-filtered reranked list (not just top k)
150
  tier_counts = {label: 0 for label, _, _ in TIER_CONFIG}
151
  for paper in filtered_reranked:
152
  score = paper.get("ce_score", 0.0)
@@ -155,7 +187,6 @@ for paper in filtered_reranked:
155
  tier_counts[label] += 1
156
  break
157
 
158
- # Rank-change lookups (CE rank vs original OpenAlex rank)
159
  oa_rank_by_id = {p["paperId"]: p["openalex_rank"] for p in filtered_reranked if p.get("paperId")}
160
 
161
  # ── Header ─────────────────────────────────────────────────────────────────────
@@ -193,27 +224,14 @@ if summary_exists:
193
  )
194
  else:
195
  if st.button("Generate Research Summary", type="secondary"):
196
- top5 = filtered_reranked[:5]
197
- papers_text = "\n".join(
198
- f"{i}. {p.get('title', 'N/A')}: {p.get('abstract') or 'No abstract available'}"
199
- for i, p in enumerate(top5, 1)
200
- )
201
- prompt = (
202
- f"Based on these top 5 papers retrieved for the query: {active_query}\n\n"
203
- f"Papers:\n{papers_text}\n\n"
204
- "Write a concise 150 word synthesis of the key themes and "
205
- "findings relevant to the query. Focus on what the papers "
206
- "collectively say about the topic."
207
- )
208
  with st.spinner("Generating research summary..."):
209
- client = Groq(api_key=os.getenv("GROQ_API_KEY"))
210
- response = client.chat.completions.create(
211
- model="llama-3.3-70b-versatile",
212
- messages=[{"role": "user", "content": prompt}],
213
- max_tokens=300,
214
- )
215
- st.session_state["summary"] = response.choices[0].message.content
216
- st.session_state["summary_query"] = active_query
217
  st.rerun()
218
 
219
 
@@ -230,7 +248,7 @@ def render_ce_card(paper: dict) -> None:
230
  if oa_r and oa_r != rank:
231
  diff = abs(oa_r - rank)
232
  arrow = '↑' if rank < oa_r else '↓'
233
- note = f"{title } (#{oa_r} β†’ #{rank}){arrow}{diff}"
234
  else:
235
  note = ""
236
 
@@ -251,22 +269,16 @@ def render_ce_card(paper: dict) -> None:
251
  )
252
  else:
253
  if st.button("Summarize", key=f"summarize_{pid}", help="Generate a 2-3 sentence summary"):
254
- full_title = paper.get("title", "N/A")
255
- abstract = paper.get("abstract") or "No abstract available."
256
- prompt = (
257
- "Summarize this academic paper in 2-3 sentences focusing "
258
- "on the key contribution and findings:\n"
259
- f"Title: {full_title}\n"
260
- f"Abstract: {abstract}"
261
- )
262
  with st.spinner("Generating summary..."):
263
- client = Groq(api_key=os.getenv("GROQ_API_KEY"))
264
- response = client.chat.completions.create(
265
- model="llama-3.3-70b-versatile",
266
- messages=[{"role": "user", "content": prompt}],
267
- max_tokens=150,
268
- )
269
- st.session_state[summary_key] = response.choices[0].message.content
 
 
270
  st.rerun()
271
 
272
 
@@ -280,7 +292,7 @@ if not top_reranked:
280
  for paper in top_reranked:
281
  render_ce_card(paper)
282
 
283
- # ── Expander: remaining papers (rank k+1 to end of filtered list) ──────────────
284
  rest_reranked = filtered_reranked[k:]
285
 
286
  if rest_reranked:
 
 
1
  import os
2
 
3
+ from dotenv import load_dotenv, find_dotenv
4
+ load_dotenv(find_dotenv())
 
 
5
 
6
+ import requests
7
  import streamlit as st
8
+
9
+ BACKEND_URL = os.getenv("BACKEND_URL", "http://localhost:8000")
10
+
11
+ AVAILABLE_MODELS = [
12
+ "BAAI/bge-reranker-v2-m3",
13
+ "cross-encoder/ms-marco-MiniLM-L-6-v2",
14
+ "cross-encoder/ms-marco-electra-base",
15
+ ]
16
+ DEFAULT_MODEL = "cross-encoder/ms-marco-MiniLM-L-6-v2"
17
 
18
  st.set_page_config(page_title="PaperRank", layout="wide")
19
 
 
25
  ]
26
 
27
 
28
+ # ── API helpers ────────────────────────────────────────────────────────────────
29
+
30
+ def _call_search(query: str, limit: int, model_name: str) -> list[dict]:
31
+ resp = requests.post(
32
+ f"{BACKEND_URL}/search",
33
+ json={"query": query, "limit": limit, "model_name": model_name},
34
+ timeout=120,
35
+ )
36
+ resp.raise_for_status()
37
+ return resp.json()["papers"]
38
+
39
+
40
+ def _call_summarize(title: str, abstract: str) -> str:
41
+ resp = requests.post(
42
+ f"{BACKEND_URL}/summarize",
43
+ json={"title": title, "abstract": abstract},
44
+ timeout=30,
45
+ )
46
+ resp.raise_for_status()
47
+ return resp.json()["summary"]
48
+
49
+
50
+ def _call_research_summary(query: str, papers: list[dict]) -> str:
51
+ resp = requests.post(
52
+ f"{BACKEND_URL}/research-summary",
53
+ json={"query": query, "papers": papers},
54
+ timeout=30,
55
+ )
56
+ resp.raise_for_status()
57
+ return resp.json()["summary"]
58
+
59
+
60
+ def _api_error(e: Exception) -> str:
61
+ if isinstance(e, requests.ConnectionError):
62
+ return f"Cannot reach backend at {BACKEND_URL}. Is the API server running?"
63
+ if isinstance(e, requests.HTTPError) and e.response is not None:
64
+ detail = e.response.json().get("detail", str(e))
65
+ return f"Backend error: {detail}"
66
+ return f"Request failed: {e}"
67
+
68
 
69
+ # ── UI helpers ─────────────────────────────────────────────────────────────────
70
 
71
  def tier_badge(score: float) -> str:
72
  if score >= 0.90:
 
85
 
86
  # ── Sidebar ────────────────────────────────────────────────────────────────────
87
  with st.sidebar:
 
 
88
  st.title("PaperRank")
89
  st.caption("Research Paper Finder")
90
  st.divider()
 
125
  if run and query.strip():
126
  st.session_state.pop("summary", None)
127
  st.session_state.pop("summary_query", None)
128
+ with st.spinner(f"Fetching and reranking with {model_name}..."):
129
  try:
130
+ papers = _call_search(query, limit, model_name)
131
  except Exception as e:
132
+ st.error(_api_error(e))
133
  st.stop()
134
 
135
+ st.session_state["reranked_papers"] = papers
 
136
  st.session_state["active_query"] = query
137
+ st.session_state["active_model"] = model_name
 
 
 
 
 
 
138
  st.rerun()
139
 
140
 
141
+ # ── Re-search when model changes without a new Run click ───────────────────────
142
  if (
143
  not run
144
+ and "reranked_papers" in st.session_state
145
  and st.session_state.get("active_model") != model_name
146
  ):
147
  with st.spinner(f"Switching model to {model_name}..."):
148
+ try:
149
+ papers = _call_search(
150
+ st.session_state["active_query"], limit, model_name
151
+ )
152
+ except Exception as e:
153
+ st.error(_api_error(e))
154
+ st.stop()
155
+ st.session_state["reranked_papers"] = papers
156
+ st.session_state["active_model"] = model_name
157
  st.rerun()
158
 
159
 
160
  # ── Main area ──────────────────────────────────────────────────────────────────
 
 
161
  st.title("PaperRank")
162
  st.caption("Research Paper Finder")
163
  if "reranked_papers" not in st.session_state:
 
179
 
180
  filtered_reranked = [p for p in reranked_papers if in_year_range(p)]
181
 
 
182
  tier_counts = {label: 0 for label, _, _ in TIER_CONFIG}
183
  for paper in filtered_reranked:
184
  score = paper.get("ce_score", 0.0)
 
187
  tier_counts[label] += 1
188
  break
189
 
 
190
  oa_rank_by_id = {p["paperId"]: p["openalex_rank"] for p in filtered_reranked if p.get("paperId")}
191
 
192
  # ── Header ─────────────────────────────────────────────────────────────────────
 
224
  )
225
  else:
226
  if st.button("Generate Research Summary", type="secondary"):
 
 
 
 
 
 
 
 
 
 
 
 
227
  with st.spinner("Generating research summary..."):
228
+ try:
229
+ summary = _call_research_summary(active_query, filtered_reranked[:5])
230
+ except Exception as e:
231
+ st.error(_api_error(e))
232
+ st.stop()
233
+ st.session_state["summary"] = summary
234
+ st.session_state["summary_query"] = active_query
 
235
  st.rerun()
236
 
237
 
 
248
  if oa_r and oa_r != rank:
249
  diff = abs(oa_r - rank)
250
  arrow = '↑' if rank < oa_r else '↓'
251
+ note = f"{title} (#{oa_r} β†’ #{rank}){arrow}{diff}"
252
  else:
253
  note = ""
254
 
 
269
  )
270
  else:
271
  if st.button("Summarize", key=f"summarize_{pid}", help="Generate a 2-3 sentence summary"):
 
 
 
 
 
 
 
 
272
  with st.spinner("Generating summary..."):
273
+ try:
274
+ paper_summary = _call_summarize(
275
+ paper.get("title", "N/A"),
276
+ paper.get("abstract") or "No abstract available.",
277
+ )
278
+ except Exception as e:
279
+ st.error(_api_error(e))
280
+ st.stop()
281
+ st.session_state[summary_key] = paper_summary
282
  st.rerun()
283
 
284
 
 
292
  for paper in top_reranked:
293
  render_ce_card(paper)
294
 
295
+ # ── Expander: remaining papers ──────────────────────────────────────────────────
296
  rest_reranked = filtered_reranked[k:]
297
 
298
  if rest_reranked: