Upload patch_models.py

patch_models.py

@@ -1,66 +1,217 @@
-"""Patch backend: add new models + fix OpenRouter routing.
-
-Routing rules for OpenRouter:
-- openai/google/* → OpenRouter
-- openai/deepseek/* → OpenRouter
-- openai/nvidia/* → OpenRouter
-- openai/poolside/* → OpenRouter
-- openai/sourceful/* → OpenRouter
-- openai/openrouter/* → OpenRouter
-- Everything else → HF Inference Router
-"""
+"""Patch backend: replace AVAILABLE_MODELS (remove old 5, add 16 new) + OpenRouter routing."""
+
 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 in agent.py ──
+# === Step 1: Replace _available_models() to return ONLY new models ===
 with open(AGENT_FILE) as f:
     content = f.read()
 
-target = "AVAILABLE_MODELS = _available_models()\n"
-patch = '''
-# === PATCH: Add models ===
-AVAILABLE_MODELS.extend([
-    {"id": "deepseek-ai/DeepSeek-V4-Pro", "label": "DeepSeek V4 Pro", "provider": "huggingface", "tier": "free", "recommended": True},
-    {"id": "deepseek-ai/DeepSeek-V4-Flash", "label": "DeepSeek V4 Flash", "provider": "huggingface", "tier": "free"},
-    {"id": "openai/gpt-5.5:fal-ai", "label": "GPT-5.5", "provider": "huggingface", "tier": "free"},
-    {"id": "Qwen/Qwen3-Coder-Next", "label": "Qwen3 Coder Next", "provider": "huggingface", "tier": "free", "recommended": True},
-    {"id": "google/gemma-3-1b-it", "label": "Gemma 3 1B", "provider": "huggingface", "tier": "free", "recommended": True},
-    {"id": "openai/google/gemini-2.0-flash-001", "label": "Gemini 2.0 Flash", "provider": "openrouter", "tier": "free"},
-    {"id": "openai/deepseek/deepseek-v4-flash:free", "label": "DeepSeek V4 Flash (OR)", "provider": "openrouter", "tier": "free"},
-    {"id": "openai/google/gemma-4-9b-it:free", "label": "Gemma 4 9B", "provider": "openrouter", "tier": "free", "recommended": True},
-    {"id": "openai/google/gemma-4-31b-it:free", "label": "Gemma 4 31B", "provider": "openrouter", "tier": "free", "recommended": True},
-    {"id": "openai/sourceful/riverflow-v2.5-fast:free", "label": "Riverflow V2.5 Fast", "provider": "openrouter", "tier": "free", "recommended": True},
-    {"id": "openai/sourceful/riverflow-v2.5-pro:free", "label": "Riverflow V2.5 Pro", "provider": "openrouter", "tier": "free", "recommended": True},
-    {"id": "openai/nvidia/nemotron-3-super-120b-a12b:free", "label": "Nemotron 3 Super 120B", "provider": "openrouter", "tier": "free"},
-    {"id": "openai/nvidia/nemotron-3-ultra-550b-a55b:free", "label": "Nemotron 3 Ultra 550B", "provider": "openrouter", "tier": "free", "recommended": True},
-    {"id": "openai/poolside/laguna-m.1:free", "label": "Laguna M.1", "provider": "openrouter", "tier": "free", "recommended": True},
-    {"id": "openai/poolside/laguna-xs.2:free", "label": "Laguna XS.2", "provider": "openrouter", "tier": "free"},
-])
-DEFAULT_FREE_MODEL_ID = "deepseek-ai/DeepSeek-V4-Pro"
-# === END PATCH ===
-'''
-
-if target in content:
-    content = content.replace(target, target + patch)
-    print("✅ Added models to AVAILABLE_MODELS")
+# Replace the default model constants first
+content = content.replace(
+    'DEFAULT_FREE_MODEL_ID = KIMI_K26_MODEL_ID',
+    'DEFAULT_FREE_MODEL_ID = "openai/openrouter/owl-alpha"'
+)
+
+# Replace _available_models() function body to return only new models
+old_func = '''def _available_models() -> list[dict[str, Any]]:
+    models = [
+        {
+            "id": DEFAULT_OPUS_MODEL_ID,
+            "label": "Claude Opus 4.8",
+            "provider": "huggingface",
+            "recommended": True,
+        },
+        {
+            "id": DEFAULT_GPT_MODEL_ID,
+            "label": "GPT-5.5",
+            "provider": "huggingface",
+        },
+        {
+            "id": DEFAULT_FREE_MODEL_ID,
+            "label": "Kimi K2.6",
+            "provider": "huggingface",
+        },
+        {
+            "id": MINIMAX_M27_MODEL_ID,
+            "label": "MiniMax M2.7",
+            "provider": "huggingface",
+        },
+        {
+            "id": GLM_51_MODEL_ID,
+            "label": "GLM 5.1",
+            "provider": "huggingface",
+        },
+        {
+            "id": DEEPSEEK_V4_PRO_MODEL_ID,
+            "label": "DeepSeek V4 Pro",
+            "provider": "huggingface",
+        },
+    ]
+    return models'''
+
+new_func = '''def _available_models() -> list[dict[str, Any]]:
+    models = [
+        {
+            "id": "openai/openrouter/owl-alpha",
+            "label": "Owl Alpha",
+            "provider": "openrouter",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "deepseek-ai/DeepSeek-V4-Pro",
+            "label": "DeepSeek V4 Pro",
+            "provider": "huggingface",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "deepseek-ai/DeepSeek-V4-Flash",
+            "label": "DeepSeek V4 Flash",
+            "provider": "huggingface",
+            "tier": "free",
+        },
+        {
+            "id": "openai/gpt-5.5:fal-ai",
+            "label": "GPT-5.5",
+            "provider": "huggingface",
+            "tier": "free",
+        },
+        {
+            "id": "Qwen/Qwen3-Coder-Next",
+            "label": "Qwen3 Coder Next",
+            "provider": "huggingface",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "google/gemma-3-1b-it",
+            "label": "Gemma 3 1B",
+            "provider": "huggingface",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "openai/google/gemini-2.0-flash-001",
+            "label": "Gemini 2.0 Flash",
+            "provider": "openrouter",
+            "tier": "free",
+        },
+        {
+            "id": "openai/deepseek/deepseek-v4-flash:free",
+            "label": "DeepSeek V4 Flash (OR)",
+            "provider": "openrouter",
+            "tier": "free",
+        },
+        {
+            "id": "openai/google/gemma-4-9b-it:free",
+            "label": "Gemma 4 9B",
+            "provider": "openrouter",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "openai/google/gemma-4-31b-it:free",
+            "label": "Gemma 4 31B",
+            "provider": "openrouter",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "openai/sourceful/riverflow-v2.5-fast:free",
+            "label": "Riverflow V2.5 Fast",
+            "provider": "openrouter",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "openai/sourceful/riverflow-v2.5-pro:free",
+            "label": "Riverflow V2.5 Pro",
+            "provider": "openrouter",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "openai/nvidia/nemotron-3-super-120b-a12b:free",
+            "label": "Nemotron 3 Super 120B",
+            "provider": "openrouter",
+            "tier": "free",
+        },
+        {
+            "id": "openai/nvidia/nemotron-3-ultra-550b-a55b:free",
+            "label": "Nemotron 3 Ultra 550B",
+            "provider": "openrouter",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "openai/poolside/laguna-m.1:free",
+            "label": "Laguna M.1",
+            "provider": "openrouter",
+            "tier": "free",
+            "recommended": True,
+        },
+        {
+            "id": "openai/poolside/laguna-xs.2:free",
+            "label": "Laguna XS.2",
+            "provider": "openrouter",
+            "tier": "free",
+        },
+    ]
+    return models'''
+
+if old_func in content:
+    content = content.replace(old_func, new_func)
+    print("OK: Replaced _available_models() with 16 new models")
 else:
-    print("⚠ AVAILABLE_MODELS = _available_models() not found!")
-    raise SystemExit(1)
+    print("WARN: Could not find old _available_models(), trying extend approach...")
+    # Fallback: clear the list and extend
+    target = "AVAILABLE_MODELS = _available_models()\n"
+    if target in content:
+        content = content.replace(
+            target,
+            target + "AVAILABLE_MODELS.clear()\n" +
+            "AVAILABLE_MODELS.extend([\n"
+            '    {"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-5.5:fal-ai", "label": "GPT-5.5", "provider": "huggingface", "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'
+            '    {"id": "openai/deepseek/deepseek-v4-flash:free", "label": "DeepSeek V4 Flash (OR)", "provider": "openrouter", "tier": "free"},\n'
+            '    {"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/sourceful/riverflow-v2.5-fast:free", "label": "Riverflow V2.5 Fast", "provider": "openrouter", "tier": "free", "recommended": True},\n'
+            '    {"id": "openai/sourceful/riverflow-v2.5-pro:free", "label": "Riverflow V2.5 Pro", "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'
+            '    {"id": "openai/poolside/laguna-xs.2:free", "label": "Laguna XS.2", "provider": "openrouter", "tier": "free"},\n'
+            "])\n"
+            'DEFAULT_FREE_MODEL_ID = "openai/openrouter/owl-alpha"\n'
+        )
+        print("OK: Used extend approach")
+    else:
+        print("FAIL: Cannot patch agent.py!")
+        import sys
+        sys.exit(1)
 
 try:
     ast.parse(content)
-    print("✅ agent.py syntax OK")
+    print("OK: agent.py syntax OK")
 except SyntaxError as e:
-    print(f"❌ Syntax error in agent.py: {e}")
+    print(f"FAIL: Syntax error in agent.py: {e}")
     raise
 
 with open(AGENT_FILE, "w") as f:
     f.write(content)
 
-# ── Step 2: Patch _resolve_llm_params to route OpenRouter models ──
+# === Step 2: Patch _resolve_llm_params for OpenRouter routing ===
 with open(LLM_PARAMS_FILE) as f:
     llm_content = f.read()
 
@@ -94,26 +245,24 @@ new_block = """    hf_model = normalized_model
 
 if old_block in llm_content:
     llm_content = llm_content.replace(old_block, new_block)
-    print("✅ Patched _resolve_llm_params: OpenRouter routing")
+    print("OK: Patched _resolve_llm_params")
+elif "openai/openrouter/" in llm_content:
+    print("OK: Already patched")
 else:
-    # Check if already patched
-    if "openai/sourceful/" in llm_content:
-        print("✅ Already patched")
-    else:
-        print("⚠ Could not find target block!")
-        idx = llm_content.find("hf_model = normalized_model")
-        if idx >= 0:
-            print(f"  Found at pos {idx}: {repr(llm_content[idx:idx+300])}")
-        raise SystemExit(1)
+    idx = llm_content.find("hf_model = normalized_model")
+    if idx >= 0:
+        print(f"WARN: Target block not found at idx {idx}")
+    import sys
+    sys.exit(1)
 
 try:
     ast.parse(llm_content)
-    print("✅ llm_params.py syntax OK")
+    print("OK: llm_params.py syntax OK")
 except SyntaxError as e:
-    print(f"❌ Syntax error in llm_params.py: {e}")
+    print(f"FAIL: Syntax error: {e}")
     raise
 
 with open(LLM_PARAMS_FILE, "w") as f:
     f.write(llm_content)
 
-print("✅ Backend patched — 15 models, sourceful/* routed to OpenRouter")
+print("OK: Backend patched - 16 models, owl-alpha default, OpenRouter routing")