Upload app.py with huggingface_hub

app.py

@@ -176,7 +176,7 @@ HEADER = """<div style="background:linear-gradient(135deg,#0a0f2e 0%,#1a0a0a 100
 <svg width="100" height="28" viewBox="0 0 120 32"><polyline points="0,16 20,16 26,4 30,28 34,2 38,26 44,16 120,16" fill="none" stroke="#c1121f" stroke-width="2.5" stroke-linecap="round"/></svg>
 <div style="font-size:2em;font-weight:900;letter-spacing:2px;"><span style="color:#ffffff;">Cardio</span><span style="color:#c1121f;">Lab</span><span style="color:#ffffff;"> AI</span></div>
 <svg width="100" height="28" viewBox="0 0 120 32" style="transform:scaleX(-1);"><polyline points="0,16 20,16 26,4 30,28 34,2 38,26 44,16 120,16" fill="none" stroke="#c1121f" stroke-width="2.5" stroke-linecap="round"/></svg></div>
-<div style="color:#9ca3af;font-size:0.68em;letter-spacing:2px;text-transform:uppercase;">Fine-tuned on 16 SJSU Papers | RAG | Llama 3.3 70B | 5 AI Models</div></div>
+<div style="color:#9ca3af;font-size:0.68em;letter-spacing:2px;text-transform:uppercase;">RAG + Fine-tuned | BioGPT | ClinicalTrials | Weekly Updates | 5 AI Models</div></div>
 <div style="display:flex;align-items:center;gap:14px;">
 <div style="text-align:right;"><div style="color:#9ca3af;font-size:0.68em;text-transform:uppercase;">Research Pillars</div>
 <div style="color:#ffffff;font-size:0.72em;margin-top:3px;">MHV  CKD  FSI</div>
@@ -240,6 +240,170 @@ def get_pubmed_chat(query, n=3):
         return chr(10).join(["https://pubmed.ncbi.nlm.nih.gov/"+i for i in ids]) if ids else ""
     except: return ""
 
+
+# ── PHASE C: BIOGPT + CLINICALTRIALS + WEEKLY UPDATE ──────────────
+
+def search_biogpt(query):
+    """Search BioGPT — trained on 15M PubMed papers via HuggingFace API"""
+    if not HF_TOKEN: return ""
+    try:
+        headers = {"Authorization": "Bearer "+HF_TOKEN}
+        # Use BioGPT for biomedical question answering
+        payload = {"inputs": query+" [SEP] Answer based on biomedical literature:"}
+        r = requests.post(
+            "https://api-inference.huggingface.co/models/microsoft/BioGPT-Large-PubMedQA",
+            headers=headers, json=payload, timeout=20
+        )
+        if r.status_code == 200:
+            result = r.json()
+            if isinstance(result, list) and len(result) > 0:
+                text = result[0].get("generated_text","")
+                # Extract just the answer part
+                if "[SEP]" in text:
+                    text = text.split("[SEP]")[-1].strip()
+                return text[:400] if text else ""
+        return ""
+    except: return ""
+
+def search_clinical_trials(query, n=5):
+    """Search ClinicalTrials.gov for heart valve and CKD trials"""
+    try:
+        r = requests.get(
+            "https://clinicaltrials.gov/api/v2/studies",
+            params={
+                "query.term": query,
+                "filter.overallStatus": "RECRUITING|COMPLETED",
+                "pageSize": n,
+                "format": "json",
+                "fields": "NCTId,BriefTitle,OverallStatus,Phase,StartDate,Condition"
+            },
+            timeout=12
+        )
+        if r.status_code != 200: return []
+        studies = r.json().get("studies",[])
+        results = []
+        for s in studies:
+            proto = s.get("protocolSection",{})
+            ident = proto.get("identificationModule",{})
+            status = proto.get("statusModule",{})
+            nct = ident.get("nctId","")
+            title = ident.get("briefTitle","")
+            phase = status.get("phase","")
+            overall = status.get("overallStatus","")
+            if nct and title:
+                results.append({
+                    "nct": nct,
+                    "title": title,
+                    "status": overall,
+                    "phase": phase,
+                    "url": "https://clinicaltrials.gov/study/"+nct
+                })
+        return results
+    except: return []
+
+def get_weekly_pubmed_update(topics=None):
+    """Get papers published in last 7 days on CardioLab topics"""
+    if topics is None:
+        topics = [
+            "mechanical heart valve thrombogenicity",
+            "microfluidic creatinine CKD diagnosis",
+            "PIV hemodynamics prosthetic valve",
+            "Mock Circulatory Loop cardiac",
+            "bileaflet valve fluid structure interaction"
+        ]
+    all_new = []
+    try:
+        from datetime import datetime, timedelta
+        week_ago = (datetime.now() - timedelta(days=7)).strftime("%Y/%m/%d")
+        for topic in topics:
+            r = requests.get(
+                "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi",
+                params={
+                    "db":"pubmed",
+                    "term":topic,
+                    "mindate":week_ago,
+                    "datetype":"pdat",
+                    "retmax":3,
+                    "retmode":"json",
+                    "sort":"date"
+                },
+                timeout=10
+            )
+            ids = r.json()["esearchresult"]["idlist"]
+            for pmid in ids:
+                all_new.append({
+                    "pmid": pmid,
+                    "topic": topic,
+                    "url": "https://pubmed.ncbi.nlm.nih.gov/"+pmid
+                })
+        return all_new
+    except: return []
+
+def full_research_search(query, search_model="Llama 3.3 70B (Best)"):
+    """Complete search across ALL sources including Phase C additions"""
+    if not query.strip(): return "Please enter a research topic."
+
+    model_id = CHAT_MODELS.get(search_model, "llama-3.3-70b-versatile")
+    expanded = expand_query_ai(query, model_id) if GROQ_KEY else query
+
+    # All search sources
+    pubmed = fetch_pubmed(expanded, n=6)
+    scholar = fetch_scholar(expanded, n=5)
+    europe = fetch_europe_pmc(expanded, n=4)
+    trials = search_clinical_trials(query, n=4)
+    weekly = get_weekly_pubmed_update()
+    biogpt_answer = search_biogpt(query)
+
+    # Format output
+    out = "QUERY: "+query+chr(10)
+    out += "AI EXPANDED: "+expanded+chr(10)
+    out += "SOURCES: PubMed + Scholar + EuropePMC + ClinicalTrials + SJSU + BioGPT"+chr(10)
+    out += "="*50+chr(10)+chr(10)
+
+    # BioGPT answer first
+    if biogpt_answer:
+        out += "BIOGPT ANSWER (trained on 15M PubMed papers):"+chr(10)
+        out += biogpt_answer+chr(10)+chr(10)
+        out += "="*50+chr(10)+chr(10)
+
+    # PubMed results
+    if pubmed:
+        out += "PUBMED ("+str(len(pubmed))+" papers):"+chr(10)
+        for p in pubmed[:6]:
+            out += p["title"][:85]+" ("+p["year"]+")"+chr(10)
+            out += "  "+p["url"]+chr(10)+chr(10)
+
+    # Scholar results
+    if scholar:
+        out += "SEMANTIC SCHOLAR ("+str(len(scholar))+" papers):"+chr(10)
+        for p in scholar[:5]:
+            out += p["title"][:85]+" ("+p["year"]+")"
+            if p["citations"] not in ("N/A","","0"): out += " | "+p["citations"]+" citations"
+            out += chr(10)+"  "+p["url"]+chr(10)+chr(10)
+
+    # Clinical trials
+    if trials:
+        out += "CLINICALTRIALS.GOV ("+str(len(trials))+" trials):"+chr(10)
+        for t in trials:
+            out += t["title"][:80]+" | "+t["status"]+" | "+t.get("phase","")+" "+chr(10)
+            out += "  "+t["url"]+chr(10)+chr(10)
+
+    # Weekly updates
+    weekly_relevant = [w for w in weekly if any(
+        kw in query.lower() for kw in ["valve","heart","ckd","creatinine","piv","tgt","mcl"]
+    )]
+    if weekly_relevant:
+        out += "NEW THIS WEEK (last 7 days):"+chr(10)
+        for w in weekly_relevant[:5]:
+            out += "  "+w["url"]+" ["+w["topic"][:40]+"]"+chr(10)
+
+    # SJSU ScholarWorks
+    out += chr(10)+"SJSU SCHOLARWORKS:"+chr(10)
+    out += "  https://scholarworks.sjsu.edu/do/search/?q="+requests.utils.quote(query)+"&context=6781027"
+
+    return out
+
+
 def research_chat(message, history, chat_model="Llama 3.3 70B (Best)"):
     if not message.strip(): return "", history
     paper_context, paper_results = search_papers(message, n=4)
@@ -593,14 +757,14 @@ with gr.Blocks(title="CardioLab AI - SJSU", css=CSS) as demo:
             voice_clear.click(lambda: [], outputs=voice_chatbot)
 
         with gr.Tab("Papers"):
-            gr.Markdown("### Search PubMed + Semantic Scholar + SJSU ScholarWorks")
+            gr.Markdown("### Search PubMed + Scholar + EuropePMC + ClinicalTrials.gov + SJSU + BioGPT (15M papers)")
             with gr.Row():
                 search_input = gr.Textbox(placeholder="e.g. bileaflet mechanical heart valve thrombogenicity hemodynamics", label="Research Topic", scale=3)
                 search_model_dd = gr.Dropdown(choices=list(CHAT_MODELS.keys()), value="Llama 3.3 70B (Best)", label="AI Model", scale=1)
                 search_btn = gr.Button("Search", variant="primary", scale=1)
             search_output = gr.Textbox(label="Results", lines=22)
-            search_btn.click(quick_search, inputs=[search_input, search_model_dd], outputs=search_output)
-            search_input.submit(quick_search, inputs=[search_input, search_model_dd], outputs=search_output)
+            search_btn.click(full_research_search, inputs=[search_input, search_model_dd], outputs=search_output)
+            search_input.submit(full_research_search, inputs=[search_input, search_model_dd], outputs=search_output)
 
         with gr.Tab("PIV CSV"):
             with gr.Row():
@@ -668,6 +832,6 @@ with gr.Blocks(title="CardioLab AI - SJSU", css=CSS) as demo:
                     gr.Button("Analyze TGT",variant="primary").click(tgt_manual,inputs=[t1,t2,t3,t4,t5],outputs=out2)
 
     gr.HTML("""<div style="text-align:center;padding:10px;border-top:1px solid #e5e7eb;background:#f9fafb;">
-    <span style="color:#9ca3af;font-size:0.75em;">CardioLab AI v36 | SJSU Biomedical Engineering | Fine-tuned on 16 SJSU Papers | RAG + Custom Model | Inspired by <a href="https://github.com/snap-stanford/Biomni" style="color:#c1121f;">Biomni Stanford</a> | <a href="https://github.com/pranatechsol/Cardio-Lab-Ai" style="color:#0057a8;">GitHub</a> | Apache 2.0 | $0 Cost</span></div>""")
+    <span style="color:#9ca3af;font-size:0.75em;">CardioLab AI v37 | SJSU Biomedical Engineering | Fine-tuned on 16 SJSU Papers | RAG + Custom Model | Inspired by <a href="https://github.com/snap-stanford/Biomni" style="color:#c1121f;">Biomni Stanford</a> | <a href="https://github.com/pranatechsol/Cardio-Lab-Ai" style="color:#0057a8;">GitHub</a> | Apache 2.0 | $0 Cost</span></div>""")
 
 demo.launch()