Improve logging and url fetching

app.py

@@ -4,6 +4,10 @@ from huggingface_hub import HfApi
 import requests
 import re
 import concurrent.futures
+import logging
+from tenacity import retry, stop_after_attempt, wait_exponential, retry_if_exception_type
+
+logging.basicConfig(level=logging.WARNING, format="%(asctime)s - %(levelname)s - %(message)s")
 
 # Configure the Streamlit page
 st.set_page_config(page_title="Heretic Models Explorer", page_icon="🔥", layout="wide")
@@ -16,6 +20,18 @@ st.markdown(
     "allowing you to sort and compare them easily. Click on any column header to sort!"
 )
 
+@retry(
+    retry=retry_if_exception_type((requests.exceptions.ConnectionError, requests.exceptions.Timeout, requests.exceptions.HTTPError)),
+    stop=stop_after_attempt(3),
+    wait=wait_exponential(multiplier=1, min=2, max=10)
+)
+def fetch_url(url):
+    response = requests.get(url, timeout=5)
+    # Raise HTTPError for rate limits (429) or server errors (5xx) to trigger a retry
+    if response.status_code in (429, 500, 502, 503, 504):
+        response.raise_for_status()
+    return response
+
 # Fetch model's readme.md file and attempt to parse refusal and KL divergence data.
 def fetch_model_info(model):
     """Fetches the README.md for a given model and dynamically extracts metrics."""
@@ -28,7 +44,7 @@ def fetch_model_info(model):
     # Download the README.md via raw URL for high-speed fetching
     url = f"https://huggingface.co/{model_id}/raw/main/README.md"
     try:
-        response = requests.get(url, timeout=5)
+        response = fetch_url(url)
         if response.status_code == 200:
             readme_text = response.text
             
@@ -38,8 +54,8 @@ def fetch_model_info(model):
             if kl_match:
                 try:
                     kl_div = float(kl_match.group(1))
-                except ValueError:
-                    pass
+                except ValueError as e:
+                    logging.warning(f"Could not parse KL divergence for {model_id}: {e}")
                     
             # Extract Refusals using Regex
             # Matches formats like: "Refusals | 15/100" or "15 / 100"
@@ -54,13 +70,18 @@ def fetch_model_info(model):
                         initial_refusals_str = f"{fractions[0][0]}/{fractions[0][1]}"
                     elif 'initial' not in line_lower and refusals_str is None:
                         refusals_str = f"{fractions[0][0]}/{fractions[0][1]}"
-                        refusal_rate = (int(fractions[0][0]) / int(fractions[0][1])) * 100 
-                        
+                        try:
+                            refusal_rate = (int(fractions[0][0]) / int(fractions[0][1])) * 100 
+                        except (ValueError, ZeroDivisionError) as e:
+                            logging.warning(f"Could not parse refusals for {model_id}: {e}")
+                            continue
                         if len(fractions) >= 2 and initial_refusals_str is None:
                             initial_refusals_str = f"{fractions[1][0]}/{fractions[1][1]}"
                             
-    except Exception:
-        pass
+    except requests.exceptions.RequestException as e:
+        logging.warning(f"Error fetching README for {model_id}: {e}")
+    except Exception as e:
+        logging.error(f"Unexpected error parsing README for {model_id}: {e}")
 
     # Skip models without explicitly stated KL divergence or refusal data
     if kl_div is None or refusals_str is None:
@@ -82,7 +103,7 @@ def get_model_data(model):
     model_id = model.id
     url = f"https://huggingface.co/{model_id}/raw/main/reproduce/reproduce.json"
     try:
-        response = requests.get(url, timeout=5)
+        response = fetch_url(url)
         if response.status_code == 200:
             data = response.json()
             metrics = data.get("metrics")
@@ -105,8 +126,12 @@ def get_model_data(model):
                         "Downloads": getattr(model, 'downloads', 0),
                         "URL": f"https://huggingface.co/{model_id}"
                     }
-    except Exception:
-        pass
+    except requests.exceptions.RequestException as e:
+        logging.warning(f"Error fetching JSON for {model_id}: {e}")
+    except (ValueError, KeyError, TypeError) as e:
+        logging.warning(f"Formatting error in JSON for {model_id}: {e}")
+    except Exception as e:
+        logging.error(f"Unexpected error processing JSON for {model_id}: {e}")
     return fetch_model_info(model)
 
 @st.cache_data(ttl=3600, show_spinner=False) # Cache for 1 hour

requirements.txt

@@ -3,3 +3,4 @@ streamlit
 huggingface_hub
 pandas
 requests
+tenacity

src/streamlit_app.py

@@ -4,6 +4,10 @@ from huggingface_hub import HfApi
 import requests
 import re
 import concurrent.futures
+import logging
+from tenacity import retry, stop_after_attempt, wait_exponential, retry_if_exception_type
+
+logging.basicConfig(level=logging.WARNING, format="%(asctime)s - %(levelname)s - %(message)s")
 
 # Configure the Streamlit page
 st.set_page_config(page_title="Heretic Models Explorer", page_icon="🔥", layout="wide")
@@ -16,6 +20,18 @@ st.markdown(
     "allowing you to sort and compare them easily. Click on any column header to sort!"
 )
 
+@retry(
+    retry=retry_if_exception_type((requests.exceptions.ConnectionError, requests.exceptions.Timeout, requests.exceptions.HTTPError)),
+    stop=stop_after_attempt(3),
+    wait=wait_exponential(multiplier=1, min=2, max=10)
+)
+def fetch_url(url):
+    response = requests.get(url, timeout=5)
+    # Raise HTTPError for rate limits (429) or server errors (5xx) to trigger a retry
+    if response.status_code in (429, 500, 502, 503, 504):
+        response.raise_for_status()
+    return response
+
 # Fetch model's readme.md file and attempt to parse refusal and KL divergence data.
 def fetch_model_info(model):
     """Fetches the README.md for a given model and dynamically extracts metrics."""
@@ -28,7 +44,7 @@ def fetch_model_info(model):
     # Download the README.md via raw URL for high-speed fetching
     url = f"https://huggingface.co/{model_id}/raw/main/README.md"
     try:
-        response = requests.get(url, timeout=5)
+        response = fetch_url(url)
         if response.status_code == 200:
             readme_text = response.text
             
@@ -38,8 +54,8 @@ def fetch_model_info(model):
             if kl_match:
                 try:
                     kl_div = float(kl_match.group(1))
-                except ValueError:
-                    pass
+                except ValueError as e:
+                    logging.warning(f"Could not parse KL divergence for {model_id}: {e}")
                     
             # Extract Refusals using Regex
             # Matches formats like: "Refusals | 15/100" or "15 / 100"
@@ -54,13 +70,18 @@ def fetch_model_info(model):
                         initial_refusals_str = f"{fractions[0][0]}/{fractions[0][1]}"
                     elif 'initial' not in line_lower and refusals_str is None:
                         refusals_str = f"{fractions[0][0]}/{fractions[0][1]}"
-                        refusal_rate = (int(fractions[0][0]) / int(fractions[0][1])) * 100 
-                        
+                        try:
+                            refusal_rate = (int(fractions[0][0]) / int(fractions[0][1])) * 100 
+                        except (ValueError, ZeroDivisionError) as e:
+                            logging.warning(f"Could not parse refusals for {model_id}: {e}")
+                            continue
                         if len(fractions) >= 2 and initial_refusals_str is None:
                             initial_refusals_str = f"{fractions[1][0]}/{fractions[1][1]}"
                             
-    except Exception:
-        pass
+    except requests.exceptions.RequestException as e:
+        logging.warning(f"Error fetching README for {model_id}: {e}")
+    except Exception as e:
+        logging.error(f"Unexpected error parsing README for {model_id}: {e}")
 
     # Skip models without explicitly stated KL divergence or refusal data
     if kl_div is None or refusals_str is None:
@@ -82,7 +103,7 @@ def get_model_data(model):
     model_id = model.id
     url = f"https://huggingface.co/{model_id}/raw/main/reproduce/reproduce.json"
     try:
-        response = requests.get(url, timeout=5)
+        response = fetch_url(url)
         if response.status_code == 200:
             data = response.json()
             metrics = data.get("metrics")
@@ -105,8 +126,12 @@ def get_model_data(model):
                         "Downloads": getattr(model, 'downloads', 0),
                         "URL": f"https://huggingface.co/{model_id}"
                     }
-    except Exception:
-        pass
+    except requests.exceptions.RequestException as e:
+        logging.warning(f"Error fetching JSON for {model_id}: {e}")
+    except (ValueError, KeyError, TypeError) as e:
+        logging.warning(f"Formatting error in JSON for {model_id}: {e}")
+    except Exception as e:
+        logging.error(f"Unexpected error processing JSON for {model_id}: {e}")
     return fetch_model_info(model)
 
 @st.cache_data(ttl=3600, show_spinner=False) # Cache for 1 hour