Update app.py

app.py

@@ -245,108 +245,171 @@ from tools.final_answer import FinalAnswerTool
 from Gradio_UI import GradioUI
 
 
+# @tool
+# def my_custom_tool(arg1: str, arg2: int) -> str:
+#     """
+#     Search and retrieve publicly available medical datasets from Hugging Face based on any medical-related keyword.
+
+#     Args:
+#         arg1: A keyword related to medical data (e.g., 'cancer', 'diabetes', 'CT scan', 'radiology', 'dermoscopy').
+#         arg2: The maximum number of datasets to retrieve.
+
+#     Returns:
+#         A numbered list (top N) of dataset names matching the search query.
+#     """
+#     try:
+#         keyword = arg1.strip().lower()
+#         limit = int(arg2)
+
+#        # Define a comprehensive list of medically relevant terms
+#         medical_terms = [
+#             # Anatomy / Body Parts
+#             "skin", "brain", "lung", "chest", "abdomen", "spine", "bone", "heart", "liver", "kidney",
+#             "bladder", "stomach", "colon", "rectum", "esophagus", "pancreas", "breast", "ear", "eye",
+#             "retina", "tooth", "teeth", "tongue", "jaw", "neck", "wrist", "hand", "leg", "arm",
+#             "shoulder", "pelvis",
+
+#             # Diseases / Conditions
+#             "cancer", "tumor", "stroke", "diabetes", "pneumonia", "covid", "asthma", "eczema",
+#             "melanoma", "hypertension", "alzheimer", "parkinson", "arthritis", "scoliosis",
+#             "epilepsy", "glaucoma", "ulcer", "hepatitis", "leukemia", "lymphoma", "tuberculosis",
+#             "anemia", "obesity", "depression", "anxiety", "bipolar", "autism", "adhd", "ptsd",
+#             "psychosis", "schizophrenia",
+
+#             # Imaging Modalities
+#             "mri", "ct", "xray", "x-ray", "ultrasound", "pet", "fmri", "mammo", "angiography",
+#             "radiography", "echocardiogram", "spect", "dermoscopy", "colonoscopy", "endoscopy",
+#             "biopsy", "histopathology",
+
+#             # Medical Specialties
+#             "radiology", "pathology", "oncology", "cardiology", "neurology", "dermatology",
+#             "dentistry", "ophthalmology", "urology", "orthopedics", "gastroenterology",
+#             "pulmonology", "nephrology", "psychiatry", "pediatrics", "geriatrics",
+#             "infectious disease",
+
+#             # Symptoms / Signs
+#             "lesion", "infection", "fever", "pain", "inflammation", "rash", "headache", "swelling",
+#             "cough", "seizure", "dizziness", "vomiting", "diarrhea", "nausea", "fatigue", "itching",
+
+#             # Common Specific Diseases
+#             "breast cancer", "prostate cancer", "lung cancer", "skin cancer", "colon cancer",
+#             "brain tumor", "liver cancer", "cervical cancer", "bladder cancer", "thyroid cancer",
+
+#             # Procedures / Interventions
+#             "surgery", "chemotherapy", "radiation", "transplant", "dialysis", "intubation",
+#             "stenting", "ventilation", "vaccination", "anesthesia", "rehabilitation", "prosthetics",
+#             "orthotics",
+
+#             # Lab Tests / Biomarkers
+#             "blood test", "cbc", "glucose", "hemoglobin", "cholesterol", "biomarker", "urinalysis",
+#             "pcr", "serology", "antibody", "antigen",
+
+#             # Clinical Settings / Roles
+#             "icu", "hospital", "emergency", "clinical notes", "nursing", "physician", "patient",
+#             "medical record", "electronic health record", "ehr", "vitals",
+
+#             # Age-based Terms
+#             "pediatric", "neonatal", "infant", "child", "adolescent", "geriatrics", "elderly",
+
+#             # Epidemiology / Public Health
+#             "epidemiology", "prevalence", "incidence", "mortality", "public health", "health disparity",
+#             "risk factor", "social determinant",
+
+#             # Pharmacology / Medications
+#             "drug", "medication", "pharmacology", "side effect", "adverse event", "dose", "tablet",
+#             "vaccine", "clinical trial", "placebo"
+#         ]
+
+
+#         if not any(term in keyword for term in medical_terms):
+#             return f"No medical datasets found for '{arg1}'. Please try another medical term."
+
+#         # Query Hugging Face API
+#         try:
+#             response = requests.get(
+#                 f"https://huggingface.co/api/datasets?search={keyword}&limit={limit}",
+#                 timeout=10
+#             )
+#             response.raise_for_status()
+#             datasets = response.json()
+#         except Exception:
+#             # Offline fallback
+#             datasets = [{"id": f"example/{keyword}-dataset-{i+1}"} for i in range(limit)]
+
+#         if not datasets:
+#             return f"No datasets found for '{arg1}'."
+
+#         # Format results neatly with numbered bullets
+#         formatted = "\n".join(
+#             [f"- Dataset {i+1}: {ds.get('id', 'Unknown')}" for i, ds in enumerate(datasets[:limit])]
+#         )
+#         return f"Medical datasets related to '{arg1}':\n{formatted}"
+
+#     except Exception as e:
+#         return f"Error searching medical datasets for '{arg1}': {str(e)}"
+from smolagents import tool
+import requests
+
 @tool
-def my_custom_tool(arg1: str, arg2: int) -> str:
+def my_custom_tool(keyword: str, top_n: int) -> str:
     """
-    Search and retrieve publicly available medical datasets from Hugging Face based on any medical-related keyword.
+    Search and retrieve publicly available medical datasets from Hugging Face based on a given keyword.
 
     Args:
-        arg1: A keyword related to medical data (e.g., 'cancer', 'diabetes', 'CT scan', 'radiology', 'dermoscopy').
-        arg2: The maximum number of datasets to retrieve.
+        keyword: A medical-related keyword (e.g., 'cancer', 'heart', 'xray', 'radiology', 'dermoscopy').
+        top_n: The maximum number of datasets to retrieve.
 
     Returns:
-        A numbered list (top N) of dataset names matching the search query.
+        A numbered list (top N) of Hugging Face dataset names matching the search query.
+        If no relevant results are found, a helpful message is returned.
     """
     try:
-        keyword = arg1.strip().lower()
-        limit = int(arg2)
+        keyword = keyword.strip().lower()
+
+        # Guardrail: Prevent using dataset IDs instead of search keywords
+        if "/" in keyword:
+            return (
+                f"'{keyword}' looks like a dataset ID, not a keyword. "
+                f"Please provide a general medical keyword (e.g., 'heart', 'cancer', 'xray')."
+            )
 
-       # Define a comprehensive list of medically relevant terms
+        # Define medically relevant terms
         medical_terms = [
-            # Anatomy / Body Parts
-            "skin", "brain", "lung", "chest", "abdomen", "spine", "bone", "heart", "liver", "kidney",
-            "bladder", "stomach", "colon", "rectum", "esophagus", "pancreas", "breast", "ear", "eye",
-            "retina", "tooth", "teeth", "tongue", "jaw", "neck", "wrist", "hand", "leg", "arm",
-            "shoulder", "pelvis",
-
-            # Diseases / Conditions
-            "cancer", "tumor", "stroke", "diabetes", "pneumonia", "covid", "asthma", "eczema",
-            "melanoma", "hypertension", "alzheimer", "parkinson", "arthritis", "scoliosis",
-            "epilepsy", "glaucoma", "ulcer", "hepatitis", "leukemia", "lymphoma", "tuberculosis",
-            "anemia", "obesity", "depression", "anxiety", "bipolar", "autism", "adhd", "ptsd",
-            "psychosis", "schizophrenia",
-
-            # Imaging Modalities
-            "mri", "ct", "xray", "x-ray", "ultrasound", "pet", "fmri", "mammo", "angiography",
-            "radiography", "echocardiogram", "spect", "dermoscopy", "colonoscopy", "endoscopy",
-            "biopsy", "histopathology",
-
-            # Medical Specialties
-            "radiology", "pathology", "oncology", "cardiology", "neurology", "dermatology",
-            "dentistry", "ophthalmology", "urology", "orthopedics", "gastroenterology",
-            "pulmonology", "nephrology", "psychiatry", "pediatrics", "geriatrics",
-            "infectious disease",
-
-            # Symptoms / Signs
-            "lesion", "infection", "fever", "pain", "inflammation", "rash", "headache", "swelling",
-            "cough", "seizure", "dizziness", "vomiting", "diarrhea", "nausea", "fatigue", "itching",
-
-            # Common Specific Diseases
-            "breast cancer", "prostate cancer", "lung cancer", "skin cancer", "colon cancer",
-            "brain tumor", "liver cancer", "cervical cancer", "bladder cancer", "thyroid cancer",
-
-            # Procedures / Interventions
-            "surgery", "chemotherapy", "radiation", "transplant", "dialysis", "intubation",
-            "stenting", "ventilation", "vaccination", "anesthesia", "rehabilitation", "prosthetics",
-            "orthotics",
-
-            # Lab Tests / Biomarkers
-            "blood test", "cbc", "glucose", "hemoglobin", "cholesterol", "biomarker", "urinalysis",
-            "pcr", "serology", "antibody", "antigen",
-
-            # Clinical Settings / Roles
-            "icu", "hospital", "emergency", "clinical notes", "nursing", "physician", "patient",
-            "medical record", "electronic health record", "ehr", "vitals",
-
-            # Age-based Terms
-            "pediatric", "neonatal", "infant", "child", "adolescent", "geriatrics", "elderly",
-
-            # Epidemiology / Public Health
-            "epidemiology", "prevalence", "incidence", "mortality", "public health", "health disparity",
-            "risk factor", "social determinant",
-
-            # Pharmacology / Medications
-            "drug", "medication", "pharmacology", "side effect", "adverse event", "dose", "tablet",
-            "vaccine", "clinical trial", "placebo"
+            "skin", "brain", "lung", "chest", "heart", "cancer", "tumor", "radiology",
+            "xray", "ultrasound", "pathology", "ct", "mri", "covid", "stroke", "diabetes",
+            "melanoma", "lesion", "infection", "fever", "pain", "inflammation", "rash",
+            "biopsy", "histopathology", "blood", "icu", "ehr", "patient", "vitals",
         ]
 
-
+        # Validate the keyword
         if not any(term in keyword for term in medical_terms):
-            return f"No medical datasets found for '{arg1}'. Please try another medical term."
+            return f"No medical datasets found for '{keyword}'. Please try another medical term."
 
-        # Query Hugging Face API
+        # Query Hugging Face datasets API
         try:
             response = requests.get(
-                f"https://huggingface.co/api/datasets?search={keyword}&limit={limit}",
+                f"https://huggingface.co/api/datasets?search={keyword}&limit={top_n}",
                 timeout=10
             )
             response.raise_for_status()
             datasets = response.json()
         except Exception:
-            # Offline fallback
-            datasets = [{"id": f"example/{keyword}-dataset-{i+1}"} for i in range(limit)]
+            # Fallback in case of API issues or no internet
+            datasets = [{"id": f"example/{keyword}-dataset-{i+1}"} for i in range(top_n)]
 
+        # Handle case when no datasets are found
         if not datasets:
-            return f"No datasets found for '{arg1}'."
+            return f"No Hugging Face datasets found for '{keyword}'."
 
-        # Format results neatly with numbered bullets
+        # Format output as a clean, numbered list
         formatted = "\n".join(
-            [f"- Dataset {i+1}: {ds.get('id', 'Unknown')}" for i, ds in enumerate(datasets[:limit])]
+            [f"- Dataset {i+1}: {ds.get('id', 'Unknown')}" for i, ds in enumerate(datasets[:top_n])]
         )
-        return f"Medical datasets related to '{arg1}':\n{formatted}"
+
+        return f"Top {len(datasets[:top_n])} Hugging Face datasets related to '{keyword}':\n{formatted}"
 
     except Exception as e:
-        return f"Error searching medical datasets for '{arg1}': {str(e)}"
+        return f"Error while searching Hugging Face datasets for '{keyword}': {str(e)}"
 
 
 @tool