Fix 403: free-tier models first in dropdown, default=gemma3:12b, friendly error messages for subscription/auth/rate-limit

llm.py

@@ -1,65 +1,66 @@
 """
 Ollama Cloud API client for script generation.
 Uses OpenAI-compatible API at https://ollama.com/v1
+
+Ollama Cloud tiers (as of 2026):
+  Free  — smaller models, 1 concurrent, light usage
+  Pro   ($20/mo) — all models, 3 concurrent, 50x usage
+  Max   ($100/mo) — all models, 10 concurrent, 250x usage
 """
 import json
 import re
 from openai import OpenAI
 
 
-# Current Ollama Cloud models (from https://ollama.com/search?c=cloud)
-# These run on Ollama's infrastructure — no local GPU needed
+# Models tagged by estimated tier access.
+# Top = likely Free tier (smaller models). Bottom = may need Pro/Max.
+# If a model returns 403, the user needs to upgrade or pick a Free-tier model.
 OLLAMA_CLOUD_MODELS = [
-    # --- Google ---
-    "gemma3:27b",
-    "gemma3:12b",
+    # --- 🟢 Likely Free tier (smaller models) ---
     "gemma3:4b",
+    "gemma3:12b",
+    "gemma3:27b",
+    "ministral-3:3b",
+    "ministral-3:8b",
+    "ministral-3:14b",
+    "gpt-oss:20b",
+    "rnj-1:8b",
+    "devstral-small-2:24b",
+    "nemotron-3-nano:30b",
+    # --- 🔵 May need Pro/Max subscription ---
     "gemma4:31b",
     "gemini-3-flash-preview",
-    # --- Alibaba Qwen ---
+    "deepseek-v4-flash",
+    "deepseek-v3.2",
+    "deepseek-v3.1:671b",
+    "gpt-oss:120b",
     "qwen3.5:397b",
     "qwen3-coder:480b",
     "qwen3-coder-next",
     "qwen3-next:80b",
     "qwen3-vl:235b",
-    # --- DeepSeek ---
-    "deepseek-v3.1:671b",
-    "deepseek-v3.2",
-    "deepseek-v4-flash",
-    # --- OpenAI ---
-    "gpt-oss:120b",
-    "gpt-oss:20b",
-    # --- Mistral ---
+    "qwen3-vl:235b-instruct",
     "mistral-large-3:675b",
     "devstral-2:123b",
-    "devstral-small-2:24b",
-    "ministral-3:14b",
-    "ministral-3:8b",
-    "ministral-3:3b",
-    # --- MiniMax ---
     "minimax-m2.7",
     "minimax-m2.5",
     "minimax-m2.1",
     "minimax-m2",
-    # --- Moonshot Kimi ---
     "kimi-k2.6",
     "kimi-k2.5",
     "kimi-k2:1t",
     "kimi-k2-thinking",
-    # --- NVIDIA ---
     "nemotron-3-super",
-    "nemotron-3-nano:30b",
-    # --- Z.ai GLM ---
     "glm-5.1",
     "glm-5",
     "glm-4.7",
     "glm-4.6",
-    # --- Deep Cogito ---
     "cogito-2.1:671b",
-    # --- Essential AI ---
-    "rnj-1:8b",
 ]
 
+# Default model — small, likely on free tier
+DEFAULT_MODEL = "gemma3:12b"
+
 
 def get_ollama_client(api_key: str, base_url: str = "https://ollama.com/v1") -> OpenAI:
     """Create an OpenAI-compatible client for Ollama cloud."""
@@ -70,7 +71,7 @@ def get_model_name(dropdown_value: str, custom_model: str) -> str:
     """Resolve model name: use custom input if provided, else dropdown."""
     if custom_model and custom_model.strip():
         return custom_model.strip()
-    return dropdown_value
+    return dropdown_value or DEFAULT_MODEL
 
 
 def extract_json_from_text(text: str) -> str:
@@ -90,6 +91,33 @@ def extract_json_from_text(text: str) -> str:
     return text
 
 
+def _friendly_api_error(e: Exception, model: str) -> str:
+    """Turn raw API errors into user-friendly messages."""
+    err_str = str(e)
+    if "403" in err_str or "subscription" in err_str.lower() or "upgrade" in err_str.lower():
+        return (
+            f"Model '{model}' requires a paid Ollama subscription.\n\n"
+            f"Options:\n"
+            f"  - Pick a Free-tier model (top of dropdown): gemma3:4b, gemma3:12b, gpt-oss:20b, etc.\n"
+            f"  - Upgrade your Ollama plan at https://ollama.com/upgrade\n"
+            f"  - Use a custom endpoint with your own model in the 'Custom Model Name' field."
+        )
+    if "401" in err_str or "unauthorized" in err_str.lower() or "invalid" in err_str.lower():
+        return (
+            f"Invalid Ollama API key. Please check your key at https://ollama.com/settings"
+        )
+    if "404" in err_str or "not found" in err_str.lower():
+        return (
+            f"Model '{model}' not found. Check the model name or try one from the dropdown."
+        )
+    if "rate" in err_str.lower() or "429" in err_str:
+        return (
+            f"Rate limit reached. Ollama free tier has limited usage. "
+            f"Wait a few minutes or upgrade at https://ollama.com/upgrade"
+        )
+    return f"Script generation failed: {err_str}"
+
+
 def generate_viral_script(
     topic: str, research_text: str, key_facts: list[str],
     model: str, api_key: str, additional_notes: str = "",
@@ -160,7 +188,6 @@ Return ONLY the JSON array, no other text. Example format:
 
         content = response.choices[0].message.content
 
-        # Handle models that use thinking tags (like qwen3, deepseek-r1)
         if "<think>" in content:
             content = re.sub(r'<think>[\s\S]*?</think>', '', content).strip()
 
@@ -193,7 +220,7 @@ Return ONLY the JSON array, no other text. Example format:
         }
 
     except Exception as e:
-        raise RuntimeError(f"Script generation failed: {str(e)}")
+        raise RuntimeError(_friendly_api_error(e, model))
 
 
 def generate_title(topic: str, model: str, api_key: str, base_url: str = "https://ollama.com/v1") -> str: