Update app.py

app.py

@@ -268,8 +268,6 @@ def call_gemini_for_edl(transcript_data, story_prompt, api_key):
         st.error("Gemini API Key is missing.")
         return None
         
-    url = f"https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-flash:generateContent?key={api_key}"
-    
     system_prompt = (
         "You are an expert Documentary Senior Editor. Use the provided transcript JSON "
         "(which includes Speaker IDs and word-level timestamps) to create a condensed story. "
@@ -305,49 +303,64 @@ def call_gemini_for_edl(transcript_data, story_prompt, api_key):
         "Content-Type": "application/json"
     }
     
-    max_retries = 3
-    for attempt in range(max_retries):
-        try:
-            res = requests.post(url, json=payload, headers=headers)
-            
-            # Handle Rate Limiting (429) specifically
-            if res.status_code == 429:
-                if attempt < max_retries - 1:
-                    wait_time = 10 * (attempt + 1)
-                    st.warning(f"Rate limit hit. Retrying in {wait_time} seconds...")
-                    time.sleep(wait_time)
-                    continue
-                else:
-                    st.error("Google API Rate Limit Exceeded (429). Please wait a few minutes before trying again, or upgrade your API tier.")
-                    return None
+    # Cascading Fallback List - If one fails, it instantly tries the next.
+    models_to_try = [
+        "gemini-2.5-flash",
+        "gemini-3.1-flash", 
+        "gemini-1.5-flash",
+        "gemini-2.5-pro"
+    ]
+    
+    # Try the final/stable endpoint first, fallback to beta if features aren't promoted yet
+    endpoints_to_try = ["v1", "v1beta"]
+    last_error = ""
+    
+    for model_name in models_to_try:
+        for api_version in endpoints_to_try:
+            url = f"https://generativelanguage.googleapis.com/{api_version}/models/{model_name}:generateContent?key={api_key}"
             
-            # Instead of generic failure, show the EXACT error message from Google
-            if not res.ok:
-                error_details = res.text
-                try:
-                    error_details = res.json().get('error', {}).get('message', res.text)
-                except Exception:
-                    pass
-                st.error(f"Google API Rejected Request ({res.status_code}): {error_details}")
-                return None
+            try:
+                res = requests.post(url, json=payload, headers=headers)
                 
-            result_json = res.json()
-            raw_text = result_json['candidates'][0]['content']['parts'][0]['text']
-            cleaned_text = clean_json_response(raw_text)
-            return json.loads(cleaned_text)
-            
-        except json.JSONDecodeError as e:
-            st.error(f"Failed to parse Gemini's output as JSON: {e}")
-            return None
-        except requests.exceptions.RequestException as e:
-            if attempt < max_retries - 1:
-                time.sleep(5)
+                # Success
+                if res.status_code == 200:
+                    result_json = res.json()
+                    raw_text = result_json['candidates'][0]['content']['parts'][0]['text']
+                    cleaned_text = clean_json_response(raw_text)
+                    return json.loads(cleaned_text)
+                    
+                # Handle High Demand (503) or Rate Limits (429)
+                elif res.status_code in [503, 429]:
+                    last_error = f"Model {model_name} busy ({res.status_code}) on {api_version}."
+                    st.warning(f"Debug: {last_error} Skipping to next model...")
+                    break # Break inner loop: No point trying v1beta if v1 is rate limited. Move to next model.
+                    
+                # Handle endpoints that don't exist (404) or API rejections (400)
+                else:
+                    try:
+                        error_details = res.json().get('error', {}).get('message', res.text)
+                    except Exception:
+                        error_details = res.text
+                    last_error = f"Model {model_name} rejected request ({res.status_code}) on {api_version}: {error_details}"
+                    st.warning(f"Debug: {last_error} Trying fallback endpoint/model...")
+                    continue # Try v1beta, or the next model
+                    
+            except json.JSONDecodeError:
+                last_error = f"Model {model_name} returned invalid JSON on {api_version}."
+                st.warning(f"Debug: {last_error} Trying fallback endpoint/model...")
+                continue
+            except requests.exceptions.RequestException as e:
+                last_error = f"Network error connecting to {model_name} on {api_version}: {e}"
+                st.warning(f"Debug: {last_error} Trying fallback endpoint/model...")
+                continue
+            except Exception as e:
+                last_error = f"Error with {model_name} on {api_version}: {e}"
+                st.warning(f"Debug: {last_error} Trying fallback endpoint/model...")
                 continue
-            st.error(f"Network Error connecting to Gemini API: {e}")
-            return None
-        except Exception as e:
-            st.error(f"Senior Editor AI Error: {e}")
-            return None
+
+    # If the loop finishes without returning, all models failed
+    st.error(f"All Gemini models failed or are currently experiencing high demand. Last error: {last_error}")
+    return None
 
 # --- Streamlit UI ---
 st.set_page_config(page_title="Junior Editor", layout="wide")
@@ -527,3 +540,6 @@ if uploaded_file:
                         
                         st.download_button(f"Download .{ext.upper()}", data=final_output, file_name=f"junior_editor_cut.{ext}")
 
+
+
+