Upload patch_models.py

patch_models.py

@@ -1,48 +1,52 @@
-"""Patch backend: add new model IDs to AVAILABLE_MODELS + fix title generation."""
-import re, os
+"""Patch backend: add new model IDs to model_ids.py + agent.py models list + title gen."""
 
+# === Step 1: Patch model_ids.py — add new model constants ===
+IDS_FILE = "/app/agent/core/model_ids.py"
+with open(IDS_FILE, "r") as f:
+    content = f.read()
+
+# Add new model ID constants after existing ones
+old_models_block = 'CLAUDE_OPUS_48_MODEL_ID = "anthropic/claude-opus-4.8:fal-ai"\nGPT_55_MODEL_ID = "openai/gpt-5.5:fal-ai"\nKIMI_K27_CODE_MODEL_ID = "moonshotai/Kimi-K2.7-Code:novita"\nMINIMAX_M3_MODEL_ID = "MiniMaxAI/MiniMax-M3:novita"\nGLM_52_MODEL_ID = "zai-org/GLM-5.2:novita"\nDEEPSEEK_V4_PRO_MODEL_ID = "deepseek-ai/DeepSeek-V4-Pro:novita"'
+
+new_models_block = 'CLAUDE_OPUS_48_MODEL_ID = "openrouter/owl-alpha"\nGPT_55_MODEL_ID = "openai/gpt-5.5:fal-ai"\nKIMI_K27_CODE_MODEL_ID = "deepseek-ai/DeepSeek-V4-Pro"\nMINIMAX_M3_MODEL_ID = "deepseek-ai/DeepSeek-V4-Flash"\nGLM_52_MODEL_ID = "openai/deepseek/deepseek-v4-flash"\nDEEPSEEK_V4_PRO_MODEL_ID = "nvidia/nemotron-3-super-120b-a12b:free"\n\n# === PATCH: New model IDs ===\nGEMMA_4_MODEL_ID = "google/gemma-4-31B-it"\nTENCENT_HY3_MODEL_ID = "openai/tencent/hy3:free"\nGEMMA_3_1B_MODEL_ID = "google/gemma-3-1b-it"\nQWEN3_CODER_NEXT_MODEL_ID = "Qwen/Qwen3-Coder-Next"\nGEMINI_FLASH_MODEL_ID = "openai/google/gemini-2.0-flash-001"'
+
+content = content.replace(old_models_block, new_models_block)
+
+# Also update the HOSTED_MODEL_IDS set
+old_hosted = "HOSTED_MODEL_IDS = {\n    CLAUDE_OPUS_48_MODEL_ID,\n    GPT_55_MODEL_ID,\n    KIMI_K27_CODE_MODEL_ID,\n    MINIMAX_M3_MODEL_ID,\n    GLM_52_MODEL_ID,\n    DEEPSEEK_V4_PRO_MODEL_ID,\n}"
+new_hosted = "HOSTED_MODEL_IDS = {\n    CLAUDE_OPUS_48_MODEL_ID,\n    GPT_55_MODEL_ID,\n    KIMI_K27_CODE_MODEL_ID,\n    MINIMAX_M3_MODEL_ID,\n    GLM_52_MODEL_ID,\n    DEEPSEEK_V4_PRO_MODEL_ID,\n    GEMMA_4_MODEL_ID,\n    TENCENT_HY3_MODEL_ID,\n    GEMMA_3_1B_MODEL_ID,\n    QWEN3_CODER_NEXT_MODEL_ID,\n    GEMINI_FLASH_MODEL_ID,\n}"
+content = content.replace(old_hosted, new_hosted)
+
+with open(IDS_FILE, "w") as f:
+    f.write(content)
+print("✅ model_ids.py patched")
+
+# === Step 2: Patch agent.py — expand AVAILABLE_MODELS ===
 AGENT_FILE = "/app/backend/routes/agent.py"
 with open(AGENT_FILE, "r") as f:
     content = f.read()
 
-# Add new models to AVAILABLE_MODELS
-target = "AVAILABLE_MODELS = _available_models()\n"
-patch = '''
-# === PATCH: Add HF Inference + OpenRouter models ===
-import os as _os
-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/deepseek/deepseek-v4-flash", "label": "DeepSeek V4 Flash (OR)", "provider": "openrouter", "tier": "free"},
-    {"id": "nvidia/nemotron-3-super-120b-a12b:free", "label": "Nemotron 3 Super 120B", "provider": "openrouter", "tier": "free"},
-    {"id": "google/gemma-4-31B-it", "label": "Gemma 4 31B", "provider": "huggingface", "tier": "free", "recommended": True},
-    {"id": "openai/tencent/hy3:free", "label": "Tencent HY3", "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": "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"},
-])
-PREMIUM_MODEL_IDS = set()  # No premium gate
-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")
-
-# Fix title generation - use huggingface/ prefix for litellm
-old_title = 'model="openai/openai/gpt-oss-120b:cerebras",\n            api_base="https://router.huggingface.co/v1",\n            api_key=api_key,'
-new_title = 'model="huggingface/deepseek-ai/DeepSeek-V4-Pro",\n            api_key=api_key,'
-if old_title in content:
-    content = content.replace(old_title, new_title)
-    content = content.replace('            reasoning_effort="low",\n', '')
-    print("✅ Title generation patched")
+# Update imports to include new model IDs
+old_import = "from agent.core.model_ids import (\n    CLAUDE_OPUS_48_MODEL_ID,\n    DEEPSEEK_V4_PRO_MODEL_ID,\n    GLM_52_MODEL_ID,\n    GPT_55_MODEL_ID,\n    KIMI_K27_CODE_MODEL_ID,\n    MINIMAX_M3_MODEL_ID,\n    strip_huggingface_model_prefix,\n)"
+new_import = "from agent.core.model_ids import (\n    CLAUDE_OPUS_48_MODEL_ID,\n    DEEPSEEK_V4_PRO_MODEL_ID,\n    GLM_52_MODEL_ID,\n    GPT_55_MODEL_ID,\n    KIMI_K27_CODE_MODEL_ID,\n    MINIMAX_M3_MODEL_ID,\n    GEMMA_4_MODEL_ID,\n    TENCENT_HY3_MODEL_ID,\n    GEMMA_3_1B_MODEL_ID,\n    QWEN3_CODER_NEXT_MODEL_ID,\n    GEMINI_FLASH_MODEL_ID,\n    strip_huggingface_model_prefix,\n)"
+content = content.replace(old_import, new_import)
+
+# Expand _available_models()
+old_available = 'def _available_models() -> list[dict[str, Any]]:\n    models = [\n        {\n            "id": CLAUDE_OPUS_48_MODEL_ID,\n            "label": "Claude Opus 4.8",\n        },\n        {\n            "id": DEFAULT_GPT_MODEL_ID,\n            "label": "GPT-5.5",\n        },\n        {\n            "id": KIMI_K27_CODE_MODEL_ID,\n            "label": "Kimi K2.7 Code",\n        },\n        {\n            "id": MINIMAX_M3_MODEL_ID,\n            "label": "MiniMax M3",\n        },\n        {\n            "id": DEFAULT_MODEL_ID,\n            "label": "GLM 5.2",\n            "recommended": True,\n        },\n        {\n            "id": DEEPSEEK_V4_PRO_MODEL_ID,\n            "label": "DeepSeek V4 Pro",\n        },\n    ]\n    return models'
+new_available = 'def _available_models() -> list[dict[str, Any]]:\n    models = [\n        {\n            "id": CLAUDE_OPUS_48_MODEL_ID,\n            "label": "Owl Alpha",\n        },\n        {\n            "id": DEFAULT_GPT_MODEL_ID,\n            "label": "GPT-5.5",\n        },\n        {\n            "id": KIMI_K27_CODE_MODEL_ID,\n            "label": "DeepSeek V4 Pro",\n        },\n        {\n            "id": MINIMAX_M3_MODEL_ID,\n            "label": "DeepSeek V4 Flash",\n        },\n        {\n            "id": DEFAULT_MODEL_ID,\n            "label": "DeepSeek V4 Flash",\n            "recommended": True,\n        },\n        {\n            "id": DEEPSEEK_V4_PRO_MODEL_ID,\n            "label": "Nemotron 3 Super 120B",\n        },\n        {\n            "id": GEMMA_4_MODEL_ID,\n            "label": "Gemma 4 31B",\n            "recommended": True,\n        },\n        {\n            "id": TENCENT_HY3_MODEL_ID,\n            "label": "Tencent HY3",\n        },\n        {\n            "id": GEMMA_3_1B_MODEL_ID,\n            "label": "Gemma 3 1B",\n            "recommended": True,\n        },\n        {\n            "id": QWEN3_CODER_NEXT_MODEL_ID,\n            "label": "Qwen3 Coder Next",\n            "recommended": True,\n        },\n        {\n            "id": GEMINI_FLASH_MODEL_ID,\n            "label": "Gemini 2.0 Flash",\n        },\n    ]\n    return models'
+content = content.replace(old_available, new_available)
+
+# Update title generation — the /title endpoint uses gpt-oss-120b:cerebras as default
+# Change it to use our HF router model
+old_title = '"openai/gpt-oss-120b:cerebras",'
+new_title = '"huggingface/deepseek-ai/DeepSeek-V4-Pro",'
+content = content.replace(old_title, new_title)
 
+# Validate syntax
 import ast
 ast.parse(content)
 print("✅ Syntax OK")
 
 with open(AGENT_FILE, "w") as f:
     f.write(content)
-print("✅ Backend patched")
+print("✅ agent.py patched")