Upload patch_models.py

patch_models.py

@@ -77,7 +77,7 @@ new_func = '''def _available_models() -> list[dict[str, Any]]:
         },
         {
             "id": "openai/gpt-oss-120b",
-            "label": "GPT-5.5",
+            "label": "gpt-oss-120b",
             "provider": "openrouter",
             "tier": "free",
         },
@@ -130,7 +130,7 @@ new_func = '''def _available_models() -> list[dict[str, Any]]:
         },
         {
             "id": "openai/cohere/north-mini-code:free",
-            "label": "Riverflow V2.5 Pro",
+            "label": "north-mini-code:free",
             "provider": "openrouter",
             "tier": "free",
             "recommended": True,
@@ -178,7 +178,7 @@ else:
             '    {"id": "openai/openrouter/owl-alpha", "label": "Owl Alpha", "provider": "openrouter", "tier": "free", "recommended": True},\n'
             '    {"id": "deepseek-ai/DeepSeek-V4-Pro", "label": "DeepSeek V4 Pro", "provider": "huggingface", "tier": "free", "recommended": True},\n'
             '    {"id": "deepseek-ai/DeepSeek-V4-Flash", "label": "DeepSeek V4 Flash", "provider": "huggingface", "tier": "free"},\n'
-            '    {"id": "openai/gpt-oss-120b", "label": "GPT-5.5", "provider": "openrouter", "tier": "free"},\n'
+            '    {"id": "openai/gpt-oss-120b", "label": "gpt-oss-120b", "provider": "openrouter", "tier": "free"},\n'
             '    {"id": "Qwen/Qwen3-Coder-Next", "label": "Qwen3 Coder Next", "provider": "huggingface", "tier": "free", "recommended": True},\n'
             '    {"id": "google/gemma-3-1b-it", "label": "Gemma 3 1B", "provider": "huggingface", "tier": "free", "recommended": True},\n'
             '    {"id": "openai/google/gemini-2.0-flash-001", "label": "Gemini 2.0 Flash", "provider": "openrouter", "tier": "free"},\n'
@@ -186,7 +186,7 @@ else:
             '    {"id": "openai/google/gemma-4-9b-it:free", "label": "Gemma 4 9B", "provider": "openrouter", "tier": "free", "recommended": True},\n'
             '    {"id": "openai/google/gemma-4-31b-it:free", "label": "Gemma 4 31B", "provider": "openrouter", "tier": "free", "recommended": True},\n'
             '    {"id": "openai/tencent/hy3-preview", "label": "Hy3 Preview", "provider": "openrouter", "tier": "paid", "recommended": True},\n'
-            '    {"id": "openai/cohere/north-mini-code:free", "label": "Riverflow V2.5 Pro", "provider": "openrouter", "tier": "free", "recommended": True},\n'
+            '    {"id": "openai/cohere/north-mini-code:free", "label": "north-mini-code:free", "provider": "openrouter", "tier": "free", "recommended": True},\n'
             '    {"id": "openai/nvidia/nemotron-3-super-120b-a12b:free", "label": "Nemotron 3 Super 120B", "provider": "openrouter", "tier": "free"},\n'
             '    {"id": "openai/nvidia/nemotron-3-ultra-550b-a55b:free", "label": "Nemotron 3 Ultra 550B", "provider": "openrouter", "tier": "free", "recommended": True},\n'
             '    {"id": "openai/poolside/laguna-m.1:free", "label": "Laguna M.1", "provider": "openrouter", "tier": "free", "recommended": True},\n'
@@ -211,8 +211,9 @@ with open(AGENT_FILE, "w") as f:
     f.write(content)
 
 # === Step 2: Patch _resolve_llm_params for OpenRouter routing ===
-# IMPORTANT: Keep the "openai/" prefix when sending to OpenRouter!
-# LiteLLM needs it to identify the provider as OpenAI-compatible.
+# IMPORTANT: Strip the "openai/" prefix before sending to OpenRouter!
+# The "openai/" prefix is needed by LiteLLM for provider detection,
+# but OpenRouter expects bare model slugs like "openrouter/owl-alpha".
 with open(LLM_PARAMS_FILE) as f:
     llm_content = f.read()
 
@@ -228,11 +229,12 @@ old_block = """    hf_model = normalized_model
 new_block = """    hf_model = normalized_model
     api_key = _resolve_hf_router_token(session_hf_token)
 
-    # === PATCH: Route all openai/-prefixed models to OpenRouter ===
-    # Match ANY model starting with openai/ — all of them go to OpenRouter
+    # === PATCH: Route openai/-prefixed models to OpenRouter ===
+    # Strip the "openai/" prefix — OpenRouter expects bare slugs like "openrouter/owl-alpha"
     if normalized_model.startswith("openai/"):
+        actual_model = normalized_model[len("openai/"):]
         return {
-            "model": normalized_model,
+            "model": actual_model,
             "api_base": "https://openrouter.ai/api/v1",
             "api_key": os.environ.get("OPENROUTER_API_KEY") or api_key or "",
         }
@@ -246,55 +248,26 @@ new_block = """    hf_model = normalized_model
 
 if old_block in llm_content:
     llm_content = llm_content.replace(old_block, new_block)
-    print("OK: Patched _resolve_llm_params (full block replace)")
-elif "openai/gpt-oss-120b" in llm_content or "gpt-oss-120b" in llm_content:
-    print("OK: Already patched with gpt-oss-120b support")
+    print("OK: Patched _resolve_llm_params (strips openai/ prefix)")
 else:
-    # Try updating just the routing if already patched
-    idx = llm_content.find("if normalized_model.startswith(")
+    # Already patched — update the OR routing to strip the prefix
+    idx = llm_content.find("if normalized_model.startswith(\"openai/\"):")
     if idx >= 0:
-        # Already has some patch — simplify to catch-all openai/ -> OR
-        if "openrouter" in llm_content[idx:idx+200]:
-            # Find and replace the OR condition block
-            or_start = llm_content.find("# === PATCH:", idx)
-            if or_start >= 0:
-                or_end = llm_content.find("# === END PATCH", or_start)
-                if or_end >= 0:
-                    # Find the full OR routing block up to the return
-                    or_block_end = llm_content.find("\n    # Don't double-prefix", or_end)
-                    if or_block_end < 0:
-                        or_block_end = llm_content.find("\n    params = {", or_end)
-                    
-                    if or_block_end >= 0:
-                        old_or_block = llm_content[or_start:or_block_end]
-                    else:
-                        old_or_block = llm_content[or_start:or_end+15]
-                else:
-                    old_or_block = llm_content[idx:llm_content.find("params = {", idx)]
-            else:
-                old_or_block = llm_content[idx:llm_content.find("params = {", idx)]
-            
-            new_or_block = '''    # === PATCH: Route all openai/-prefixed models to OpenRouter ===
-    # Match ANY model starting with openai/ — all of them go to OpenRouter
-    if normalized_model.startswith("openai/"):
-        return {
-            "model": normalized_model,
-            "api_base": "https://openrouter.ai/api/v1",
-            "api_key": os.environ.get("OPENROUTER_API_KEY") or api_key or "",
-        }
-    # === END PATCH ===
-
-    params = {'''
+        # Find the return block
+        return_start = llm_content.find('"model":', idx)
+        if return_start >= 0:
+            line_start = llm_content.rfind('\n', 0, return_start) + 1
+            line_end = llm_content.find('\n', line_start)
+            old_model_line = llm_content[line_start:line_end]
             
-            if old_or_block != new_or_block:
-                llm_content = llm_content.replace(old_or_block, new_or_block)
-                print("OK: Simplified OR routing to catch-all openai/ prefix")
+            # Check if it already strips prefix
+            if "normalized_model[len" in old_model_line or "actual_model" in old_model_line:
+                print("OK: Already strips openai/ prefix")
             else:
-                print("OK: Routing already up to date")
-        else:
-            print("FAIL: No existing OR patch found!")
-            import sys
-            sys.exit(1)
+                # Replace the model line to strip prefix
+                new_model_line = '            "model": normalized_model[len("openai/"):],'
+                llm_content = llm_content.replace(old_model_line, new_model_line)
+                print("OK: Fixed to strip openai/ prefix before sending to OpenRouter")
     else:
         print("FAIL: Could not find routing condition!")
         import sys
@@ -310,4 +283,4 @@ except SyntaxError as e:
 with open(LLM_PARAMS_FILE, "w") as f:
     f.write(llm_content)
 
-print("OK: Backend patched - All openai/-prefixed models route to OpenRouter")
+print("OK: Backend patched - 16 models, OR strips openai/ prefix, GPT-5.5→gpt-oss-120b, Riverflow→north-mini-code:free")