Upload patch_models.py

patch_models.py

@@ -1,10 +1,17 @@
-"""Patch backend: add new model IDs to AVAILABLE_MODELS + fix OpenRouter routing."""
+"""Patch backend: add new model IDs to AVAILABLE_MODELS + fix OpenRouter routing.
+
+OpenRouter models:
+- openai/openrouter/owl-alpha — FREE by default, no :free suffix
+- openai/google/gemini-2.0-flash-001 — PAID, no :free
+- openai/*/...:free — free tier models, need :free suffix
+"""
 import ast
+import os
 
 AGENT_FILE = "/app/backend/routes/agent.py"
 LLM_PARAMS_FILE = "/app/agent/core/llm_params.py"
 
-# ── Step 1: Patch AVAILABLE_MODELS ──
+# ── Step 1: Patch AVAILABLE_MODELS in agent.py ──
 with open(AGENT_FILE) as f:
     content = f.read()
 
@@ -18,7 +25,7 @@ AVAILABLE_MODELS.extend([
     {"id": "openai/nvidia/nemotron-3-super-120b-a12b:free", "label": "Nemotron 3 Super 120B", "provider": "openrouter", "tier": "free"},
     {"id": "google/gemma-3-1b-it", "label": "Gemma 3 1B", "provider": "huggingface", "tier": "free", "recommended": True},
     {"id": "Qwen/Qwen3-Coder-Next", "label": "Qwen3 Coder Next", "provider": "huggingface", "tier": "free", "recommended": True},
-    {"id": "openai/openrouter/owl-alpha:free", "label": "Owl Alpha", "provider": "openrouter", "tier": "free", "recommended": True},
+    {"id": "openai/openrouter/owl-alpha", "label": "Owl Alpha", "provider": "openrouter", "tier": "free", "recommended": True},
     {"id": "openai/google/gemini-2.0-flash-001", "label": "Gemini 2.0 Flash", "provider": "openrouter", "tier": "free"},
     {"id": "openai/google/gemma-4-31b-it:free", "label": "Gemma 4 31B", "provider": "openrouter", "tier": "free", "recommended": True},
     {"id": "openai/nvidia/nemotron-3-ultra-550b-a55b:free", "label": "Nemotron 3 Ultra 550B", "provider": "openrouter", "tier": "free", "recommended": True},
@@ -62,7 +69,8 @@ new_block = """    hf_model = normalized_model
 
     # === PATCH: Route OpenRouter models to openrouter.ai ===
     # OpenRouter expects the model name WITHOUT the openai/ prefix.
-    # E.g. "openai/openrouter/owl-alpha:free" → send "openrouter/owl-alpha:free"
+    # E.g. "openai/openrouter/owl-alpha" -> send "openrouter/owl-alpha"
+    # No :free suffix for Owl Alpha (free by default).
     _or_prefixes = ("openai/google/", "openai/deepseek/", "openai/nvidia/", "openai/poolside/", "openai/openrouter/")
     if normalized_model.startswith(_or_prefixes):
         # Strip the "openai/" prefix so OpenRouter gets the correct model ID
@@ -85,6 +93,11 @@ if old_block in llm_content:
     print("✅ Patched _resolve_llm_params: OpenRouter routing with stripped prefix")
 else:
     print("⚠ Could not find target block in llm_params.py")
+    # Debug: find the actual block
+    idx = llm_content.find("hf_model = normalized_model")
+    if idx >= 0:
+        print(f"Found 'hf_model = normalized_model' at pos {idx}")
+        print(repr(llm_content[idx:idx+300]))
 
 try:
     ast.parse(llm_content)
@@ -97,4 +110,6 @@ with open(LLM_PARAMS_FILE, "w") as f:
     f.write(llm_content)
 
 print("✅ Backend patched")
-print("  - Owl Alpha: sends 'openrouter/owl-alpha:free' to openrouter.ai")
+print("  - Owl Alpha: sends 'openrouter/owl-alpha' to openrouter.ai (NO :free)")
+print("  - Gemini 2.0 Flash: sends 'google/gemini-2.0-flash-001' to openrouter.ai (NO :free)")
+print("  - Other OR free models: include :free suffix")