Create app.py

app.py

@@ -0,0 +1,213 @@
+# ================================================================
+# 🧬 PathogenAgent: Agentic AI for Pathogen Genomic Analysis
+# Hugging Face Space – Gradio + LangChain + BioGPT
+# ================================================================
+
+import gradio as gr
+import requests
+import json
+import re
+from Bio import Entrez, SeqIO
+from io import StringIO
+import torch
+from transformers import AutoTokenizer, AutoModelForCausalLM
+
+# ================================================================
+# 1. تنظیمات
+# ================================================================
+
+Entrez.email = "public@example.com"
+Entrez.tool = "PathogenAgent"
+
+# بارگذاری BioGPT (مدل شما یا مدل پایه)
+MODEL_NAME = "Sepideh2027/biogpt-clinvar-finetuned"  # یا "microsoft/biogpt"
+tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
+model = AutoModelForCausalLM.from_pretrained(MODEL_NAME)
+model.eval()
+
+if torch.cuda.is_available():
+    model = model.to("cuda")
+
+# ================================================================
+# 2. ابزارها (Tools)
+# ================================================================
+
+def search_pubmed(query):
+    """جستجوی PubMed"""
+    try:
+        url = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi"
+        params = {"db": "pubmed", "term": query, "retmax": 3, "retmode": "json"}
+        resp = requests.get(url, params=params, timeout=10)
+        ids = resp.json().get("esearchresult", {}).get("idlist", [])
+        if not ids:
+            return "No articles found."
+        
+        url2 = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi"
+        params2 = {"db": "pubmed", "id": ",".join(ids[:3]), "retmode": "json"}
+        resp2 = requests.get(url2, params=params2, timeout=10)
+        results = []
+        for uid, rec in resp2.json().get("result", {}).items():
+            if uid == "uids": continue
+            results.append(f"- {rec.get('title', 'No title')}")
+        return "📄 **PubMed Articles:**\n" + "\n".join(results)
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+def search_genbank(organism):
+    """دریافت توالی ژنومی از GenBank"""
+    try:
+        handle = Entrez.esearch(db="nucleotide", term=f"{organism}[Organism] AND complete genome", retmax=1)
+        record = Entrez.read(handle)
+        handle.close()
+        if not record["IdList"]:
+            return f"No sequence found for '{organism}'."
+        seq_id = record["IdList"][0]
+        handle = Entrez.efetch(db="nucleotide", id=seq_id, rettype="fasta", retmode="text")
+        seq_data = handle.read()
+        handle.close()
+        seq = SeqIO.read(StringIO(seq_data), "fasta")
+        return f"🧬 **GenBank:**\n- {organism}\n- Length: {len(seq.seq):,} bp\n- {str(seq.seq[:200])}..."
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+def search_clinvar(variant):
+    """دریافت اطلاعات بالینی از ClinVar"""
+    try:
+        url = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esearch.fcgi"
+        params = {"db": "clinvar", "term": variant, "retmax": 2, "retmode": "json"}
+        resp = requests.get(url, params=params, timeout=10)
+        ids = resp.json().get("esearchresult", {}).get("idlist", [])
+        if not ids:
+            return f"No ClinVar records for '{variant}'."
+        url2 = "https://eutils.ncbi.nlm.nih.gov/entrez/eutils/esummary.fcgi"
+        params2 = {"db": "clinvar", "id": ",".join(ids[:2]), "retmode": "json"}
+        resp2 = requests.get(url2, params=params2, timeout=10)
+        results = []
+        for uid, rec in resp2.json().get("result", {}).items():
+            if uid == "uids": continue
+            sig = rec.get("clinicalsignificance", [{}])[0].get("description", "Unknown")
+            results.append(f"- {rec.get('title', '')} → **{sig}**")
+        return "🧬 **ClinVar:**\n" + "\n".join(results)
+    except Exception as e:
+        return f"Error: {str(e)}"
+
+# ================================================================
+# 3. Agent Planner (ساده اما هوشمند)
+# ================================================================
+
+def pathogen_agent(query):
+    """
+    Agent اصلی که بر اساس سوال کاربر تصمیم می‌گیرد از کدام ابزارها استفاده کند
+    """
+    query_lower = query.lower()
+    
+    # تصمیم‌گیری ساده بر اساس کلمات کلیدی
+    if "pubmed" in query_lower or "article" in query_lower or "paper" in query_lower:
+        return search_pubmed(query)
+    
+    elif "genbank" in query_lower or "genome" in query_lower or "sequence" in query_lower:
+        # استخراج نام موجودیت
+        words = query.split()
+        organism = " ".join(words[-2:]) if len(words) > 2 else query
+        return search_genbank(organism)
+    
+    elif "clinvar" in query_lower or "variant" in query_lower or "mutation" in query_lower:
+        # استخراج نام واریانت
+        return search_clinvar(query)
+    
+    else:
+        # ترکیب چند ابزار
+        pubmed_result = search_pubmed(query)
+        genbank_result = search_genbank(query)
+        clinvar_result = search_clinvar(query)
+        
+        # خلاصه‌سازی با BioGPT
+        summary = f"""
+        **Query:** {query}
+        
+        **PubMed Summary:**
+        {pubmed_result}
+        
+        **GenBank Summary:**
+        {genbank_result}
+        
+        **ClinVar Summary:**
+        {clinvar_result}
+        
+        ---
+        **Generated by PathogenAgent v1.0**
+        """
+        
+        return summary
+
+# ================================================================
+# 4. BioGPT برای تفسیر نهایی (اختیاری)
+# ================================================================
+
+def interpret_with_biogpt(text):
+    """خلاصه‌سازی نهایی با BioGPT"""
+    prompt = f"Summarize the following genomic information:\n{text}\nSummary:"
+    inputs = tokenizer(prompt, return_tensors="pt", truncation=True, max_length=512)
+    if torch.cuda.is_available():
+        inputs = {k: v.to("cuda") for k, v in inputs.items()}
+    with torch.no_grad():
+        outputs = model.generate(**inputs, max_new_tokens=150, temperature=0.3)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+# ================================================================
+# 5. رابط کاربری Gradio
+# ================================================================
+
+def process_query(query):
+    if not query.strip():
+        return "⚠️ Please enter a pathogen name, gene, or variant."
+    
+    # اجرای Agent
+    result = pathogen_agent(query)
+    
+    # استفاده از BioGPT برای خلاصه‌سازی (اختیاری)
+    if len(result) > 500:
+        summary = interpret_with_biogpt(result)
+        return f"{result}\n\n---\n**🧠 AI Summary:**\n{summary}"
+    
+    return result
+
+# رابط Gradio
+with gr.Blocks(title="PathogenAgent - Agentic AI for Genomics", theme=gr.themes.Soft()) as demo:
+    gr.Markdown("""
+    # 🧬 PathogenAgent: Agentic AI for Pathogen Genomic Analysis
+    
+    **Ask about any pathogen, gene, or variant.**  
+    The agent will search PubMed, GenBank, and ClinVar to provide a comprehensive report.
+    
+    ---
+    """)
+    
+    with gr.Row():
+        with gr.Column(scale=4):
+            query_input = gr.Textbox(
+                label="🔬 Enter your query",
+                placeholder="e.g., SARS-CoV-2 spike protein mutations",
+                lines=2
+            )
+            submit_btn = gr.Button("🚀 Analyze", variant="primary")
+        
+        with gr.Column(scale=6):
+            output = gr.Markdown(label="📊 Report")
+    
+    submit_btn.click(
+        fn=process_query,
+        inputs=query_input,
+        outputs=output
+    )
+    
+    gr.Markdown("""
+    ---
+    ### 📌 Examples
+    - `SARS-CoV-2 genome`
+    - `BRCA1 c.68_69delAG`
+    - `Ebola virus recent outbreaks`
+    - `TP53 R175H`
+    """)
+
+demo.launch()